Interesting, I’m not big on AI but I have thought often it would be nice to have an ‘agent’ that monitors ebay or other classifieds sites for items based on a natural language description.
Something like “I want an old mini PC to use as a home server, it should have roughly these specs and cost under this amount”, and then an LLM would run some searches every day, parse the results and send me a message if something comes up.
It’s pretty easy to get alerts for when items are available for a certain price if you know the exact item you want, but on eBay and classifieds sites, I usually just want something in a rough ballpark, and the best way to find that is come back and check every day looking through searches.
I don’t really see any value in having the AI do the purchase itself though.
>eBay and classifieds sites, I usually just want something in a rough ballpark, and the best way to find that is come back and check every day looking through searches.
You already can do that on ebay. Whatever terms get you ballpark results can become a search alert. You don't need to know the items you want exactly. And you get a handy email with the results of that alert when they hit, which you can scroll over in about ten seconds.
The purchasing itself can important for it to jump on a great price. Maybe it finds what you're looking for at 1a while you're sleeping for example. Also if this were a business and you were going to resell it the AI could also create the listing as soon as the item is purchased.
This reminds me so much of an old World of Warcraft addon i used in 2005-6 or so… I believe it was called ‘bottom feeder’ or something.
Basically, you would leave your character logged in sitting at the auction house. It would observe auctions for a while, and generate pricing data and sales data. Then, you would enable automatic mode, and it would automatically bid/buy any item that someone put up for sale if the price was much lower than normal.
You would leave it running overnight, or whatever, then come back to go pick up all the items it bought, and then you would go back to the auction house and sell all your items you bought at the correct price.
Basically, you would see buy auctions created by people who didn’t realize what the correct price should be and sold too cheaply. Since this was an automated system, you could beat any human to take advantage of the deal.
I made a ton of in game currency doing this.
After a few months they changed the auction rules to prevent this… add-ons could no longer directly bid on items, and you had to sit there and click “buy” whenever the script found a good deal. This severely limited the amount you could make with the script.
Basically this mirrors the eBay timeline, with the same reasons I am guessing… eBay (like WoW) doesn’t want bots collecting arbitrage.
I essentially do this but on a state surplus auctions site. It's just a scheduled action which searches for something, e.g. old Lego kits, once a week. Usually nothing comes up but at least once there are kits I know about it.
Does the tools and features ebay already has not meet this need?
Can't you set up a saved search that ebay will notify you of?
>“I want an old mini PC to use as a home server, it should have roughly these specs and cost under this amount”
This is a bad example because at pretty much all times, there is sufficient inventory for you to find the actual item you want, so you don't need the "agent" to repeatedly check. In instances where there is limited inventory, saved searches have been the reliable solution for decades. It's how niche youtube channels have acquired niche hardware forever.
I found this interesting to think about. On the surface, this could fit the requirements. However, in the past decade, I feel a dynamic where my relationship with tech companies like eBay are actually antagonistic, rather than cooperative. They (and American society) have become so extractive and user-hostile that I have no trust their features are designed to accomplish my goals. Instead, they are designed to accomplish the company's objectives and it's only by coincidence if the customer is satisfied.
I thought botting on ebay has been forbidden since forever, LLM or otherwise. This isn't so much a policy change as an affirmation that existing policy still applies.
Yeah that's very possible, I have never built anything with LLMs and I'm not a heavy user so I'm not sure how feasible it is.
I do think I would already get value from a least worse option every day, a sort of 'digest', so I don't have remember, and to look through results myself. I think it's a best case for LLM use for me, there is no harm at all in false negatives or positives, there are no significant stakes and I think the vagueness / unpredictability of the output is an advantage, it might find something that I had not even considered (like for my example: here is a used laptop with roughly these specs, it could also be a good home home server, something like that).
> I never understood why eBay set things up to enable sniping.
I've seen studies on auction method that suggest the difference in the final price (between explicit end time and the more traditional extend-until-bidding-stops method) is negligible for online auctions except in a few special cases. This is a marked difference from the expectation that, like a real-world in-person auction, the extending deadline might encourage further spur-of-the-moment bidding.
Whether it makes much difference to the final price or not is immaterial though if the buyers believe that it does. This is one of the (several) reasons why eBay won out against similar competition in the early days: buyers felt they were getting a better deal by being able to snipe so favoured eBay with more of their attention and this brought more sellers to the platform (which attracted more buyers, and so on round the loop). It is telling that to deal with the extra load imposed on the system by external bots refreshing pages and putting in automated bids, instead of switching to an extending auction model they implemented what is almost a built-in sniping feature.
Auction sites have to be very careful (or just very lucky) in their messaging, to convince both sellers and buyers that they are getting a good deal - any major change to how eBay works could upset the balance that they currently have in that regard and start a flood in the other direction (the more people leave, the more other people will think about leaving) to the building toward critical mass that was how they won out in those early days.
In my experience, most purely online auctions, other than eBay, do work that way. Numerous auction houses, for example, including essentially all the major ones, have their auctions online now: when they are hybrid, that involves online live bidding where an online bid will cause the auctioneer in the room to keep the lot open for more bids; when they are "timed" or "online only", times are extended in some way on bids near the deadline. It does, in fact, work much better. There is still an advantage to bidding very late: there is no disadvantage, and it lowers bids in cases of irrational or imperfect opposing bidders. But it limits that process to something that can be done by hand.
eBay really seems to be the only auctioneer using the snipable process it uses.
This is how the popular car auction site bringatrailer.com bidding process works for cars sold on their site too, a quirk of which is that it makes watching the end of the auctions live online kinda fun, especially given the discussions that break out in comment section on each car up for sale while folks nervously watch the current candidate for the final bid cool down.
Much like your example, in the two minutes before the end of the auction, every new bid placed extends the auction clock by another two minutes, the winner hasn't won the auction until two minutes have passed with no further counter bids.
Completely agree with this. It's very odd that the eBay algorithm has a hard stop, as it directly discourages the price from rising when the bidding is hottest, and is absolutely susceptible to sniping.
> Bidding is open as long as there have been bids in the past 5 minutes.
You also have to proportionally raise the bid increment, or you'll have people bidding $1 up at the end of every 5 minute period in order to exhaust and frustrate the person they're bidding against. Their opponent's only choices are 1) to mirror the $1 raises, which could go on for ages (then sleep and automation become issues), or 2) to make a big jump hoping that they jump past their opponents limit.
In the case of 2) the dollar bidder's limit could be +$10, and there's no rational way for their opponent to choose the amount for a big jump other than jumping to their limit. Meaning that they just wipe away all of their potential bargain and get it for their valuation. Leaves a sour taste; feels like they're bidding against themselves.
As somebody who auctions things, I use a required greater than "≈10%" of the current bid amount bid increment, and the auction doesn't end on an item until there hasn't been a bid for 10 minutes. Works great. "≈10%" means to just drop the last digit of the current bid to know the minimum next bid. Then I can set the auction to end an hour before I really want it to end; if it hasn't ended by then it's because I mispriced something and the right people found it.
You capture all the value you can, and it runs completely unattended. You just need a way to timestamp bids and broadcast that timestamp e.g. a forum post.
This how Trade Me (NZ auction site) works: any bid in the last 2 minutes delays the close time to 2 minutes after the bid. That can happen repeatedly, and I've seen it go on for over 20 minutes on highly contended auctions. It works well.
There is no need for that. They only need to implement a closing auction like stock markets. But eBay hasn't done anything since the 1990's except raise fees.
Guessing here - but they are probably relying on game theory / auction theory. They have a built in "sniping bot" - by allowing you to type your highest price, and it will auto-bid for you until that price.
The fear of being sniped encourages you to bid your maximum value, and not just wait and see if you can sneak in a lower bid. This is what all auction sites want.
Another eBay precursor auction side, onsale.com, had the same setup. The auction ended at X date/time or five or ten minutes (I forget which) after the last bid was made.
What's the point of sniping bots when eBay has automatic bidding? Counter-sniping is essentially built-in, if your price ceiling is higher then a snipers then you're guaranteed to win even if they bid at the last millisecond.
This was my belief for many years, but then I tried sniping (with the same prices I was putting as my maximum bid before!) and my success rate skyrocketed and the prices I was paying dropped.
It seems that despite repeated reminders and explanations, there are three groups of people using eBay "incorrectly" that make the sniping strategy viable:
1) People who do not understand proxy bidding and think that they "need" to repeatedly bid in increments.
2) People who are irrational about their price ceiling and are willing to bid above their price ceiling because they want to "win".
3) People who want to drive up the price either to deprive others of a good deal, or to drive up the price on behalf of the seller by starting a bidding war with the two above groups.
From a sellers perspective it is common to deal with buyers who won't pay because they paid "more than they wanted", although this is against the eBay ToS and a bid is a contract to purchase the item, because there are few consequences for not doing so.
For some reason, auctions with more bidders seem to attract more bidders, whereas auctions with zero bids seem to go unnoticed. I wonder if this has to do with eBay's search ranking algorithm or some other irrational behavior that I don't understand. At any rate, bidding with 5 or less seconds left to go seems to defeat the above behaviors. I find it distasteful and irrational but it works so I put up with it.
eBay's reputation and trust network is really what makes it a viable product at this point. Given how unreliable Facebook Marketplace buyers are and how many scams are present, I would hesitate to conduct any major transactions beyond a local area.
Snipers essentially convert the ascending-bid proxy auction used in eBay into a Vickrey second-price sealed bid auction, allowing a buyer to not reveal their preferences to other participants. In theory, with rational participants, this shouldn't have any effect on revenue. In practice, buyers do not always understand auction mechanics and delay setting the highest price they're willing to pay until they are outbid. If they're outbid 3 seconds before the deadline, they lost.
Establishing the price ceiling is difficult, though. You might arbitrarily set it as $23, but be sniped at $23.30. The sniper bot only needs to bid that small increment over your arbitrary ceiling.
Can you really say that $23 was your hard limit, or would you have paid $23.40? Unless you're buying something also available at retail, nobody can be that accurate in foresight.
Sniping removes the 'contemplation window' to reconsider your bid.
- bidding more than once and allowing time for others to counter bid drives up the price through competition for the item. Sniping also removes the temptation to counter bid, rather than to stick to your maximum bid.
- not sniping allows the seller to do ghost bidding, letting them discover your maximum price (including counter bidding). Here someone always out bid you (the ghost bidder) but the seller says the winner didn't complete the sale so offers it to you at your highest bid.
From what I understand, the reasoning behind the snipe method of bidding is to avoid showing to other bidders that there is interest, leading to the, supposed, outcome of more likely being the only bidder and thereby receiving the item at the sellers starting bid price (or slightly above) rather than at the "max one was willing to pay" price.
I've bought hundreds of things on ebay over the years and I've never understood the issue with "sniping".
Sure, I've been outbid at the last moment. Losing an auction is always a little frustrating. But if I was willing to pay that price I should have bid it myself. Feels fair enough?
And I prefer to use sniping bots because they let me revise my bid all the way up until the auction ends. If I put a bid on something and then sleep on it and decide I don’t actually want to pay that much, I can lower my bid or cancel it. If I bid with eBay directly then I loose that flexibility. It has nothing to do with trying to outsmart people or be sneaky.
why would you bid the highest price you can afford in an auction? the seller agreed to auction the thing; they could have just offered it for a set price.
This is most likely the reason. I could see a lot of "buy for me bot" users deciding that they really did not mean that color shirt (or some other reason) when they asked it to buy a "brand X shirt in size Y" and forgot to tell the bot what colors they would accept as options and did not realize the bot might buy an "electric purple" (or some other color they dislike) shirt because it was not constrained in color choice.
I personally didn't understand why people still snipe on eBay even though they already have an automated bidding system and the reason is that you don't want to fall victim to nibblers. Nibblers are people who bid a low amount that raises the price but is unlikely to win the auction. I.e. someone bids 30€ on an item that you would bid 50€ on. This raises the price to 30€ because of your automatic bids. The nibbler then bids 35€ to see if that was enough and it still wasn't, losing you 5€ from early bidding. If the nibbler thought he could get the item for 30€, you would only have to pay 30.50€ to beat him. The other reason is that you don't want to lock up your money since a long auction timer means you can't start bidding on the next auction in case you lose.
This was my thought as well, sniping bots have been around for as long as ebay has. Perhaps though, the sniping bots don't cause as much load on ebay's infrastructure?
I think AI is going to level the playing field with all these bots that have been used for things like this (including scalpers for those low supply/high demand items), and retailers will (hopefully) have no choice but to address the issue once everyone starts to use/abuse them.
Banning AI agents is the new "banning mobile browsers." Companies tried that too in the early smartphone era - remember when sites blocked mobile user agents to force desktop views?
The businesses that win will be the ones that build AI-agent-friendly interfaces, not the ones that try to ban them. eBay is protecting their ad revenue and impulse-buy funnel in the short term, but they're ceding the future to whoever figures out how to make agent-compatible commerce work.
Every product and platform will eventually have an "agent API" alongside their human UI. The only question is who builds it first.
> remember when sites blocked mobile user agents to force desktop views?
You mean today, when web apps refuse to render to mobile browser agents, forcing redirect to mobile apps, but work fine when toggled to a desktop agent?
I think more likely we will unlock browser agents and no company will develop an agent api. They will have a user facing website used by agents or humans the same way.
This also completely sidesteps any actions Ebay decides. It will have all credentials for them and mfa. To ebay this will look identical to the user with no real way to stop it.
If they don't currently see a way to make agent commerce work, the smart move is to stay alive long enough for someone else to figure it out then buy them or simply copy it. And if it never works out this will have been far cheaper in the medium run.
I do not endorse this approach but it is well established in the tech industry.
Tried selling on eBay as a regular Joe lately? Item sold for roughly $190 and I lost $45 in fees - I didn't even have a premium ad or pay for any of the boosting.
The problem is with items that have a national market but not a local one. For example - there may be very few local buyers who will pay a decent price for a vintage slide rule, but many on eBay. My general strategy is to list on FBM first for the eBay price that I hope to get, and then accept offers down to 75% of the price. If I don't get any bites after about a month I switch to eBay.
This. I was selling an obscure book once. I doubt there is anyone local that would be interested in it. It was sold on eBay within a week.
Same for a half functioning Xbox. No "normal" person would want that. But apparently, on eBay, something like a dozen people took serious interest in it, and it was sold in a few days in "parts only" condition. For sure I didn't like how much the transaction fee I paid, but at least I got rid of it for a decent amount of money.
At least in the UK, I don't lose any of the selling price to fees, 0%.
The buyer has additional buying fees on their side and postage is included in the final price.
As someone who hasn't sold on eBay in a looooong time but was thinking about it for some stuff I haven't been able to sell on Marketplace, their pages and pages of fee structuring were intimidating. What was the breakdown of that $45, if you don't mind sharing?
Of course you can run anything you want on your computer. But it is also ebay's right to decide whether or not its computer will allow requests from your computer.
> But it is also ebay's right to decide whether or not its computer will allow requests from your computer.
That is dangerous thinking right there: Ebay does not have rights.
Of course ebay may do it anyway, and it may take time for justice to correct things, but it is not Right, nor their right, to violate law even to protect themselves.
One pattern I keep noticing is that when the future gets harder to predict, the first visible response is not innovation but a tightening of legal and risk frameworks. Platforms start hardening contracts and banning edge behaviors because internal models can no longer reliably track downstream consequences. A subtle form of constraint collapse where rules substitute for orientation.
Feedback still flows through metrics and policies, but it no longer carries enough of a cue to guide real learning, so it gets inverted into compliance and arbitration instead. Risk management becomes the substitute for understanding, and when context collapses, meaning drifts.
Is so far as you don't have to use the site, that's true, but they are legally enforceable, and you could absolutely be sued for breaking them if you upset eBay enough.
I’ve been working on a small experimental gateway that sits between agents and customer-facing execution paths and forces decisions through policy + approval before anything irreversible happens.
v2 I just shipped adds:
• risk scoring on drafts
• policy path traces
• approval chain previews
• highlighted spans showing what triggered the block
• admin review endpoints
The motivation is exactly what people are pointing at here: once agents can transact, marketplaces end up banning them unless there’s a way to pause, inspect, and assign responsibility at execution time.
Curious what failure modes you’d want intercepted first if eBay or Amazon ever exposed agent purchase APIs.
Anecdotally, I've noticed an uptick in my eBay feeds recently of items being immediately relisted after having supposedly sold. This has always happened occasionally, but within the past couple weeks, I've noticed it happening for like 25% of my watched items. I'm wondering if bots are buying things for which the owners are then canceling when it comes time to pay.
I've also had to return a few items for which eBay's AI-generated description was wrong in ways that couldn't be verified in the product's images. I can only imagine the increase in canceled/returned orders from all the different AI features and bots.
What is the use case for LLM agent shoppers? I can't imagine delegating the purchase of a used item to an AI (I'd be okay with AI identifying the best deals for me to review). This must be something for people who are doing something at scale like flipping items on Ebay or drop shipping.
I imagine this type of automation existed before LLM agents came along - what do they add? Is it just the ability to evaluate the product description? Item quality is already listed as a categorical variable.
"Hey, ChatGPT/Grok/GeneriBot4000, please watch for a great deal on a 1982 stratocaster guitar - must be in good or better condition, $600 or less, and if you see it, go ahead and buy it without confirmation"
Ongoing tasks, arbitrage for mispriced postings in ways that aren't currently exploited that LLMs make feasible - by banning auto-buy, maybe they're attempting to delineate between human seeming behavior and automation, and giving AI permission to buy looks too much like a real person?
I have decent tech company salary but I don't even buy $10 books without checking everything. This week I almost bought a wrong book (manually) because how similar the title is. Automating stuff with AI is interesting, but I don't want the hassle of getting surprised and handling returns, if the item can be returned at all, especially on eBay.
I genuinely wonder, would you do that, really? Sure 600$ is not the end of the world for certain countries, but neither it is a sum I'm willing to just lose on random. What if the electronic parrot buys from an obvious counterfeit vendor or obvious scammer? Or what if it buys you a stratocaster but different? Or a random 1982 guitar? What if it ignores 600$? Or what if it buys 600$ item with 300$ shipping and 500$ customs from god knows where?
I've seen enough by now and I know that some people will just unleash LLMs on anything without almost no oversight. We can already see people use agentic IDEs with "do all the shit" flag, they would probably easily add finances to the list of automation.
Yeah I guess that makes sense for some people. I'm just not in a financial position where I'd let an AI buy a $600 used guitar without me taking a look at it first.
My mistake, you're completely correct, perhaps even more-correct than the wonderful flavor of Mococoa drink, with all-natural cocoa beans from the upper slopes of Mount Nicaragua. No artificial sweeteners!
Drop shippers who arbitrage between major and minor ecommerce platforms need to maintain their listings, re-price things, etc. They don't care if the AI gets it wrong sometimes as long as they more than make back the cost of deploying it.
So now imagine ten thousand of these jerks telling their AI of choice "hey go scrape everything you can and re-list it for 10% more". That's a lot of load on the platforms at both ends for listings that are unlikely to generate many sales.
But that also seems like a very inefficient way to accomplish this automation task from the drop shippers side too. What do you gain from the LLM that non LLM automation couldn't do more cost effectively?
How do ticket scalpers make money? It's an automation war. You can run arbitrage strategies at scale if you can scrape markets with bots that understand unstructured data. Even if trades go wrong sometimes it can be profitable on average.
"Hey ChatGPT I want to build my own personal cloud storage computer, buy all the hardware for me then walk me through building and configuring it. My budget is $600, try to get the best deals and make sure that all the parts are compatible. I'm fine with used parts as long as they're a good deal and are in working order."
"Buy good quality for less than market value and don't get scammed" is so far beyond AGI complete that I cannot fathom how people think it will be reliably doable by LLMs
I think it's telling though how everyone's example of how an AI agent can help you shop is "I want it to take advantage of someone else's mistake or knowledge asymmetry to make me free money".
Are you all just embarrassed used car salespeople?
Can’t charge for something if you’re giving it away for free.
Data’s the only moat left. Companies like stack overflow need to build revenue streams from AI or they will cease to exist.
By banning bots and then licensing some kind of access, eBay can protect itself from merely being a listing point that no human actually visits. Tailwind and their adverts via docs model, eBay and its promoted listings model, we’re going to see businesses adapt or die on this.
I might be out of the loop, but are agents actually out there buying stuff from "unwilling" vendors at any significant scale? I thought that was still mostly limited to opt-in partnerships with retailers. Still, eBay might be anticipating the issues you mentioned and trying to get ahead of them.
Not commonly known (I work in this space), but yes.
Agents are being used to automate things like non-cash account balance arbitrage, stacking and abusing marketing promotions, triangulated purchasing schemes, and purchase-refund arbitrage schemes at an increasingly large scale.
They may have an inkling that the big LLM companies will want to pay for future/past data... I imagine either Google or OpenAI has something predictive and shopping-related in the books.
Right -- this seems more of a protective measure than something they will proactively enforce.
If you have a well-behaved agent that uses a browser to buy on eBay, I doubt that will cause issues. But if it leads to issues, they can point to that clause instead of having to help repair the issues caused by someone else's software.
And we enter the predicted cycle of "new thing that's going to Connect Everything"
Just like faxes, the internet and the world wide web, sure, this new thing Could connect everything, but that's not nearly as much of a technical problem as a sociopolitical one. Same as it ever was.
It may just be to stop third parties from creating a whole business out of "shop for me" AI bots. Individual users getting away with it might not be a problem, but with it being against ToS, it'd be a lot more shaky to build a business out of it.
In fact, it may just be that eBay wants to be the business selling AI "buy for me" agents.
This is a case where it may be that people are outsourcing shitty user experiences to an AI.
I’m not a huge ebayer but I’m usually watching one or two auctions at a time. The problem is that you can’t disallow marketing notifications. So, if I want to be usefully alerted for a new item in a search, or that I’ve been outbid, or the imminent end of an auction, I’ll also be getting notifications and emails about all kinds of shit I don’t care about. $5 off coupons (that only apply to 8 items that I don’t want). “You might like this!” notifications (spoiler: I never do). Group buying times (who cares?).
So I either disable ALL notifications (and have an LLM write a script that crawls searches manually and much more appropriately notifies me on its own), or I enable notifications and get a bunch of trash spam.
As it relates to specifically to buying, we’ve known for a long time that we’re all up against some kind of bot that’s timed the exact last moment and amount to outbid us. It’s no fun.
I’ve been an eBay user since 1998 and it’s been on a very slow roll of enshittification since then.
Make your experience better for humans and maybe we’d be less inclined to outsource negative experiences to AI.
example of focused leadership - a commenter already noted how wondering listings drive revenue
if it was some Bozo executive as we see at most tech companies - they would be advocating to implement the Open Agentic Commerce whatever being pushed by google while not noticing its killing their own company
This does make me wonder. I see on HN (and hello if you see my comment) people who use screen scrapers or screen readers to read and use the web. I would be REALLY interested to know how many of these users use any of these newer AI browsers like Comet, I forget what the one from ChatGPT is called, but I know as a regular user I can make Comet do automated things like price comparisons across tabs and websites. I could totally see the immense value in someone who relies on a screen reader to access the web having access to an AI powered browser, but I don't know that any of them are designed with these users in mind necessarily.
My question then becomes, does this policy violate the ADA for those users in particular? UIf it doesn't today, should it tomorrow? Especially if these AI browsers actually become viable for those users. Will there be a future where if you're protected by ADA you can be cleared to use a more automated browser? I would imagine a sane rule for such an exception would require you to fully identify yourself to the website in order to prevent abuse by bots pretending to need that type of access (the good old "trust me bro" problem). Or maybe they get to use it but it becomes more rate limited to the average user speed or whatever.
Dumbest thing in the world, not wanting buyers who are ready to complete transaction the moment they find what they want. When my car broke down and cost too much to repair I described what I need - low mileage, big trunk, 40+ MPG, under certain price, close to dealership where my broken car sat in service. That I gave the query to Grok 4 Expert (because it does heavy web scraping), found a 2020 Ford Escape Hybrid and drove it away two hours later, because rationally speaking missing a lot of work in this economy is a bigger financial risk than missing $3000 on manual bargain hunting vs a good AI hunt. If any dealers blocked scaping on sales pages, too bad for them. Speaking of which, any good bargain / secondhand market AI agent friendly API?
No, but when instituting bullshit policies or trying to regulate natural/normal behavior for selfish gain, it helps you if you can enforce the policy, otherwise people will just ignore it.
eBay is hyper aggressive about fingerprinting, they will catch things like it trivially. Browsers leak all sorts of information like what sockets are open on localhost, making yourself look like an actual person is very challenging to someone motivated to detect you.
LLMs don't need browser automation though. Multimodal models with vision input can operate a real computer with "real" user inputs over USB, where the computer itself returns a real, plausible browser fingerprint because it is a real browser being operated by something that behaves humanly.
This ban isn't about 'fairness' or bot protection; it's about protecting the Impression Funnel.
Marketplaces like eBay are designed to monetize 'Wandering Attention'—sponsored listings, 'customers also bought', and sidebar ads.
An AI Agent represents 'Laser Focused Attention'. It executes a transaction with zero wandering. It effectively turns the marketplace into a commoditized backend database / dumb pipe.
From a Growth/Unit Economics perspective, an AI Agent is a nightmare customer. It has zero probability of impulse buying and generates zero ad revenue. They are banning them to save their business model, not their inventory.
I suspect they might lose money on returns too, which are probably more likely if an AI misunderstands what the buyer wants or misjudges quality or can’t detect fake listings etc
I don’t know if there are other ways eBay could lose money on returns. But my single data point: the very first thing I sold on eBay (a manual lever espresso machine) got returned because the buyer clearly didn’t know how to use it, and claimed it was defective. And because eBay has a money back guarantee, they just reached their hands into my back account and withdrew the earnings from the sale + the shipping costs for the delivery to the buyer + the shipping fees for the return. They even kept their listing fee and the sales tax. So… I don’t think eBay stands to lose money directly from returns. Maybe they risk pissing too many sellers off with an increased rate of this horrific experience?!
Another reason I could think of is Security. There is a bunch of cheating goes on there. As a seller I lost my laptop to a scammer. Seller paid be until I shipped & cancelled the transaction. Buyer asked me ship it to their son’s address. Since I didn’t use buyer’s address registered in their eBay account eBay/Paypal didn’t pay me either. AI accelerates these scams.
Or probably even wrose, it actually shifts the attention and the wandering. That phase will happen inside the LLM, where the LLM decides which link to suggest, i.e. whoever pays the LLM the most. And worse yet, that will apply not just for products, but for platforms, so if amazon pays chatgpt more than ebay does, there goes your sale.
An AI-Agent browsing eBay for a "widget" for a given individual will also likely not be browsing eBay's advertising listings (sponsored and promoted listings [1]) which would potentially equate to a loss in ad revenue for eBay. So there is likely a "protect the advertising moat" aspect to their "ban" as well.
[1] Given how hard eBay pushes sellers to purchase the sponsored and promoted listing tiers (at an additional fee of course) implies they make some nice revenue stream from these advertisements.
This assumes the LLM ecosystem stays centralized. Open source models running locally or on user-controlled infrastructure flip this - the agent works for you, not for whoever pays the model provider.
The race is already happening: open weights models are getting good enough that "your personal shopping agent" doesn't need to phone home to a company with ad incentives. The future probably looks more like ad blockers than ad platforms - agents that aggressively optimize for user preferences, not platform revenue.
I'd say it mostly has to do with limiting their own liability and reputational damage if an AI bot "hallucinates" and places hundreds of incorrect orders and sellers get hit with negative ratings and refund requests due to no fault of their own.
Stated more cynically, many platforms have an interest in attention hijacking. Done well, agents' 'laser focused attention' could help users avoid wasting time (wandering attention) and money (impulse buys). This is a good thing, even if it dings revenue of some existing platforms. If a company's business model is impulse buying and ad revenue (this isn't eBay IMO), then good riddance.
Great analysis of the real motivation here. But this feels like the record labels trying to ban MP3 players. You can protect the impression funnel today, but the trajectory is clear - consumers will increasingly delegate purchasing decisions to agents, and the platforms that adapt will capture that flow.
The marketplace that builds "agent-friendly" commerce (verified listings, structured data, transparent pricing, API access) becomes the default backend for AI shopping. The one that bans agents becomes a legacy system humans have to manually navigate when the agent can't help.
eBay's current business model may be a "nightmare customer" for AI agents, but that's a problem with the business model, not with the agents.
>The marketplace that builds "agent-friendly" commerce (verified listings, structured data, transparent pricing, API access) becomes the default backend for AI shopping.
I'd like to believe this, but claims like this have been made since the early days of internet commerce. After all, it's not hard to specify structured data about items and run queries against it. But it largely has not materialized outside of a few special suppliers.
You can't actually search Amazon or eBay or Wayfair for things with specified dimensions or characteristics. You can, however, find lots of listings for things like "Gzsbaby 6 Piece Jumbo Dinosaur Toys for Kids 3-5 and Toddlers, Large Soft Dinosaur Toys for Lovers - Perfect Party Favors, Birthday Gifts "
Perhaps this time is different? But why is it different? What economic incentives will lead to good structured data and transparent pricing, rather than whatever the AI equivalent of glurge/slop listings is?
I don't think the 'wandering attention' dissapears, rather it is pushed to the LLM product. It's more of a competitive transfer from the incumbent product category to the new one, it's not that the new product category 'fixes' the 'problem'
If I am able to find what I’m looking for and purchase it via AI why must I be subjected to advertising & promotion of items in have zero interest in? Amazon has made finding what I want painful. I suspect eBay is just as painful (I see to work there).
This reply has hallmarks of AI slop. Green name, 2x its not X its Y, em dashes.
If you want to be a productive member try commenting what you put into your prompt instead of the slop that comes out.
Also classic ai drivel: This is about protecting the business model, not their inventory. My brother in AI, eBay doesn't have inventory. They're a platform.
I didn't catch that on first read, but I see why you'd say that. LLMs are ridiculous in the constant usage "it's not X it's Y" -- It's in almost every response from Opus 4.5. "It's not X it's Y" is ruined for regular writing.
I'm also skeptical of anything that claims to reliably detect AI writing. FWIW, I plugged the comment into Pangram Labs, which claims to be the most reliably and seems to have worked well before. It categorized the comment as 100% human written with medium confidence.
That's a cynical take, so it will probably get upvoted, but what are you basing it on?
Ebay is a pretty eclectic marketplace and I can think of a number of possible reasons that have little to do with ads. For example, they may be worried about high error rates, and thus buyer and seller dissatisfaction. If I instruct an agent to buy X, eBay is almost never interchangeable with Amazon or Target.
They have no problem surfacing their listings on Google Shopping.
But, ads directly correspond to revenue stream, and a loss of ad "impressions" would result in a reduced revenue stream, so a "protect the advertising" response is not at all unusual to consider as a portion of their (eBay's) reasoning for this ban.
Given how hard they push sellers to purchase their "extra cost listing enhancements" (i.e., purchase to have your listings show in the "advertisement" spots) it appears that they may make a decent revenue stream from these advertising angles. An AI-agent could find listings without going through the advertising displays and as such cut into this revenue stream.
> In connection with using or accessing our Services you agree to comply with this User Agreement, our policies, our terms, and all applicable laws, rules, and regulations, and you will not...
> use any robot, spider, scraper, data mining tools, data gathering and extraction tools, or other automated means (including, without limitation buy-for-me agents, LLM-driven bots, or any end-to-end flow that *attempts to place orders without human review*) to access our Services for any purpose, except with the prior express permission of eBay;
Hasn't eBay's traffic been 80% bots since day one? I haven't participated in an auction in forever because even 20 years ago you were guaranteed to get sniped by a bot on anything except actual garbage.
Interesting, I’m not big on AI but I have thought often it would be nice to have an ‘agent’ that monitors ebay or other classifieds sites for items based on a natural language description.
Something like “I want an old mini PC to use as a home server, it should have roughly these specs and cost under this amount”, and then an LLM would run some searches every day, parse the results and send me a message if something comes up.
It’s pretty easy to get alerts for when items are available for a certain price if you know the exact item you want, but on eBay and classifieds sites, I usually just want something in a rough ballpark, and the best way to find that is come back and check every day looking through searches.
I don’t really see any value in having the AI do the purchase itself though.
>eBay and classifieds sites, I usually just want something in a rough ballpark, and the best way to find that is come back and check every day looking through searches.
You already can do that on ebay. Whatever terms get you ballpark results can become a search alert. You don't need to know the items you want exactly. And you get a handy email with the results of that alert when they hit, which you can scroll over in about ten seconds.
Sure, sure, and next you'll tell me there's a way to know which side of the car the gas cap is on.
2 replies →
The purchasing itself can important for it to jump on a great price. Maybe it finds what you're looking for at 1a while you're sleeping for example. Also if this were a business and you were going to resell it the AI could also create the listing as soon as the item is purchased.
This reminds me so much of an old World of Warcraft addon i used in 2005-6 or so… I believe it was called ‘bottom feeder’ or something.
Basically, you would leave your character logged in sitting at the auction house. It would observe auctions for a while, and generate pricing data and sales data. Then, you would enable automatic mode, and it would automatically bid/buy any item that someone put up for sale if the price was much lower than normal.
You would leave it running overnight, or whatever, then come back to go pick up all the items it bought, and then you would go back to the auction house and sell all your items you bought at the correct price.
Basically, you would see buy auctions created by people who didn’t realize what the correct price should be and sold too cheaply. Since this was an automated system, you could beat any human to take advantage of the deal.
I made a ton of in game currency doing this.
After a few months they changed the auction rules to prevent this… add-ons could no longer directly bid on items, and you had to sit there and click “buy” whenever the script found a good deal. This severely limited the amount you could make with the script.
Basically this mirrors the eBay timeline, with the same reasons I am guessing… eBay (like WoW) doesn’t want bots collecting arbitrage.
> the AI could also create the listing as soon as the item is purchased.
There are businesses doing the other way round: list a bunch of stuff, then once an order is placed find the item to fulfil it with.
I can't think of a scenario in which me buying some used stuff on ebay is that serious to be honest, having the AI buy would be a huge risk as well.
> Also if this were a business and you were going to resell it the AI could also create the listing as soon as the item is purchased.
That sounds like an awful grift, but good point, people might use such a system for that.
4 replies →
I essentially do this but on a state surplus auctions site. It's just a scheduled action which searches for something, e.g. old Lego kits, once a week. Usually nothing comes up but at least once there are kits I know about it.
Does the tools and features ebay already has not meet this need?
Can't you set up a saved search that ebay will notify you of?
>“I want an old mini PC to use as a home server, it should have roughly these specs and cost under this amount”
This is a bad example because at pretty much all times, there is sufficient inventory for you to find the actual item you want, so you don't need the "agent" to repeatedly check. In instances where there is limited inventory, saved searches have been the reliable solution for decades. It's how niche youtube channels have acquired niche hardware forever.
I found this interesting to think about. On the surface, this could fit the requirements. However, in the past decade, I feel a dynamic where my relationship with tech companies like eBay are actually antagonistic, rather than cooperative. They (and American society) have become so extractive and user-hostile that I have no trust their features are designed to accomplish my goals. Instead, they are designed to accomplish the company's objectives and it's only by coincidence if the customer is satisfied.
I thought botting on ebay has been forbidden since forever, LLM or otherwise. This isn't so much a policy change as an affirmation that existing policy still applies.
I would expect a naive implementation would give you a "least worse" option everyday and can't judge when it is "good enough"
Afterall, that's what most people would be when asked to make decisions for others without context.
Making the agent understand your requirements would be quite a bit of work.
Yeah that's very possible, I have never built anything with LLMs and I'm not a heavy user so I'm not sure how feasible it is.
I do think I would already get value from a least worse option every day, a sort of 'digest', so I don't have remember, and to look through results myself. I think it's a best case for LLM use for me, there is no harm at all in false negatives or positives, there are no significant stakes and I think the vagueness / unpredictability of the output is an advantage, it might find something that I had not even considered (like for my example: here is a used laptop with roughly these specs, it could also be a good home home server, something like that).
This would be helpful only if you were the only person using it.
This was the premise of ChatGPT Pulse
So scraping bots and “buy for me” bots are bad, but the incredibly annoying sniping bots are OK? That sure feels like a double standard.
I never understood why eBay set things up to enable sniping.
Many years ago, there was an auction site called uBid. They had the sane rule: Bidding is open as long as there have been bids in the past 5 minutes.
So the end date could be January 24th, 3pm, but if someone bids at 2:58pm, the deadline is extended to 3:05pm. And it keeps going.
You know, like how auctions in the real world work.
> I never understood why eBay set things up to enable sniping.
I've seen studies on auction method that suggest the difference in the final price (between explicit end time and the more traditional extend-until-bidding-stops method) is negligible for online auctions except in a few special cases. This is a marked difference from the expectation that, like a real-world in-person auction, the extending deadline might encourage further spur-of-the-moment bidding.
Whether it makes much difference to the final price or not is immaterial though if the buyers believe that it does. This is one of the (several) reasons why eBay won out against similar competition in the early days: buyers felt they were getting a better deal by being able to snipe so favoured eBay with more of their attention and this brought more sellers to the platform (which attracted more buyers, and so on round the loop). It is telling that to deal with the extra load imposed on the system by external bots refreshing pages and putting in automated bids, instead of switching to an extending auction model they implemented what is almost a built-in sniping feature.
Auction sites have to be very careful (or just very lucky) in their messaging, to convince both sellers and buyers that they are getting a good deal - any major change to how eBay works could upset the balance that they currently have in that regard and start a flood in the other direction (the more people leave, the more other people will think about leaving) to the building toward critical mass that was how they won out in those early days.
3 replies →
In my experience, most purely online auctions, other than eBay, do work that way. Numerous auction houses, for example, including essentially all the major ones, have their auctions online now: when they are hybrid, that involves online live bidding where an online bid will cause the auctioneer in the room to keep the lot open for more bids; when they are "timed" or "online only", times are extended in some way on bids near the deadline. It does, in fact, work much better. There is still an advantage to bidding very late: there is no disadvantage, and it lowers bids in cases of irrational or imperfect opposing bidders. But it limits that process to something that can be done by hand.
eBay really seems to be the only auctioneer using the snipable process it uses.
7 replies →
This is how the popular car auction site bringatrailer.com bidding process works for cars sold on their site too, a quirk of which is that it makes watching the end of the auctions live online kinda fun, especially given the discussions that break out in comment section on each car up for sale while folks nervously watch the current candidate for the final bid cool down.
Much like your example, in the two minutes before the end of the auction, every new bid placed extends the auction clock by another two minutes, the winner hasn't won the auction until two minutes have passed with no further counter bids.
> https://bringatrailer.com/how-bat-works/
1 reply →
Completely agree with this. It's very odd that the eBay algorithm has a hard stop, as it directly discourages the price from rising when the bidding is hottest, and is absolutely susceptible to sniping.
> Bidding is open as long as there have been bids in the past 5 minutes.
You also have to proportionally raise the bid increment, or you'll have people bidding $1 up at the end of every 5 minute period in order to exhaust and frustrate the person they're bidding against. Their opponent's only choices are 1) to mirror the $1 raises, which could go on for ages (then sleep and automation become issues), or 2) to make a big jump hoping that they jump past their opponents limit.
In the case of 2) the dollar bidder's limit could be +$10, and there's no rational way for their opponent to choose the amount for a big jump other than jumping to their limit. Meaning that they just wipe away all of their potential bargain and get it for their valuation. Leaves a sour taste; feels like they're bidding against themselves.
As somebody who auctions things, I use a required greater than "≈10%" of the current bid amount bid increment, and the auction doesn't end on an item until there hasn't been a bid for 10 minutes. Works great. "≈10%" means to just drop the last digit of the current bid to know the minimum next bid. Then I can set the auction to end an hour before I really want it to end; if it hasn't ended by then it's because I mispriced something and the right people found it.
You capture all the value you can, and it runs completely unattended. You just need a way to timestamp bids and broadcast that timestamp e.g. a forum post.
This how Trade Me (NZ auction site) works: any bid in the last 2 minutes delays the close time to 2 minutes after the bid. That can happen repeatedly, and I've seen it go on for over 20 minutes on highly contended auctions. It works well.
7 replies →
There is no need for that. They only need to implement a closing auction like stock markets. But eBay hasn't done anything since the 1990's except raise fees.
10 replies →
Guessing here - but they are probably relying on game theory / auction theory. They have a built in "sniping bot" - by allowing you to type your highest price, and it will auto-bid for you until that price.
The fear of being sniped encourages you to bid your maximum value, and not just wait and see if you can sneak in a lower bid. This is what all auction sites want.
4 replies →
That's how whatnot does it as well (what eBay badly copied as eBay live)
Another eBay precursor auction side, onsale.com, had the same setup. The auction ended at X date/time or five or ten minutes (I forget which) after the last bid was made.
Wow. I remember Ubid, and I bought several things from it. It would become problematic when folks would bid up things far beyond the starting price.
AFAIK eBay does do this but I’m not sure if it’s only certain categories or it’s configurable by seller. I’ve definitely seen it happen
Pretty much every auction platform I've seen, except eBay, extends listings by a few minutes every time a bid is received.
This is how Yahoo! Japan Auctions works. If an auction receives a bid in the last few minutes, it is automatically extended.
It works quite well!
Japanese auction sites work this way too.
What's the point of sniping bots when eBay has automatic bidding? Counter-sniping is essentially built-in, if your price ceiling is higher then a snipers then you're guaranteed to win even if they bid at the last millisecond.
This was my belief for many years, but then I tried sniping (with the same prices I was putting as my maximum bid before!) and my success rate skyrocketed and the prices I was paying dropped.
It seems that despite repeated reminders and explanations, there are three groups of people using eBay "incorrectly" that make the sniping strategy viable: 1) People who do not understand proxy bidding and think that they "need" to repeatedly bid in increments. 2) People who are irrational about their price ceiling and are willing to bid above their price ceiling because they want to "win". 3) People who want to drive up the price either to deprive others of a good deal, or to drive up the price on behalf of the seller by starting a bidding war with the two above groups.
From a sellers perspective it is common to deal with buyers who won't pay because they paid "more than they wanted", although this is against the eBay ToS and a bid is a contract to purchase the item, because there are few consequences for not doing so.
For some reason, auctions with more bidders seem to attract more bidders, whereas auctions with zero bids seem to go unnoticed. I wonder if this has to do with eBay's search ranking algorithm or some other irrational behavior that I don't understand. At any rate, bidding with 5 or less seconds left to go seems to defeat the above behaviors. I find it distasteful and irrational but it works so I put up with it.
eBay's reputation and trust network is really what makes it a viable product at this point. Given how unreliable Facebook Marketplace buyers are and how many scams are present, I would hesitate to conduct any major transactions beyond a local area.
10 replies →
Snipers essentially convert the ascending-bid proxy auction used in eBay into a Vickrey second-price sealed bid auction, allowing a buyer to not reveal their preferences to other participants. In theory, with rational participants, this shouldn't have any effect on revenue. In practice, buyers do not always understand auction mechanics and delay setting the highest price they're willing to pay until they are outbid. If they're outbid 3 seconds before the deadline, they lost.
1 reply →
Establishing the price ceiling is difficult, though. You might arbitrarily set it as $23, but be sniped at $23.30. The sniper bot only needs to bid that small increment over your arbitrary ceiling.
Can you really say that $23 was your hard limit, or would you have paid $23.40? Unless you're buying something also available at retail, nobody can be that accurate in foresight.
Sniping removes the 'contemplation window' to reconsider your bid.
53 replies →
Sniping is the only way to bid for two reasons:
- bidding more than once and allowing time for others to counter bid drives up the price through competition for the item. Sniping also removes the temptation to counter bid, rather than to stick to your maximum bid.
- not sniping allows the seller to do ghost bidding, letting them discover your maximum price (including counter bidding). Here someone always out bid you (the ghost bidder) but the seller says the winner didn't complete the sale so offers it to you at your highest bid.
1 reply →
The act of bidding itself shows interest and raises the price.
3 replies →
From what I understand, the reasoning behind the snipe method of bidding is to avoid showing to other bidders that there is interest, leading to the, supposed, outcome of more likely being the only bidder and thereby receiving the item at the sellers starting bid price (or slightly above) rather than at the "max one was willing to pay" price.
1 reply →
Auto bid isn't the same as sniping. Sniping hides information about demand. Auto bid can't hide information as soon as there is another bidder.
1 reply →
I have lost most of my bids to bots. Bots will literally bit at hh:59:59. The ceiling value doesn't work unless you bid way above the asking price.
2 replies →
I've bought hundreds of things on ebay over the years and I've never understood the issue with "sniping".
Sure, I've been outbid at the last moment. Losing an auction is always a little frustrating. But if I was willing to pay that price I should have bid it myself. Feels fair enough?
And I prefer to use sniping bots because they let me revise my bid all the way up until the auction ends. If I put a bid on something and then sleep on it and decide I don’t actually want to pay that much, I can lower my bid or cancel it. If I bid with eBay directly then I loose that flexibility. It has nothing to do with trying to outsmart people or be sneaky.
I run up the prices in less competitive auctions just for fun occasionally, especially if I think someone is getting too good a deal.
12 replies →
why would you bid the highest price you can afford in an auction? the seller agreed to auction the thing; they could have just offered it for a set price.
5 replies →
Buy for me bots results in more returns, cancellations, item not as described and other problems ebay and sellers have to deal with
This is most likely the reason. I could see a lot of "buy for me bot" users deciding that they really did not mean that color shirt (or some other reason) when they asked it to buy a "brand X shirt in size Y" and forgot to tell the bot what colors they would accept as options and did not realize the bot might buy an "electric purple" (or some other color they dislike) shirt because it was not constrained in color choice.
I personally didn't understand why people still snipe on eBay even though they already have an automated bidding system and the reason is that you don't want to fall victim to nibblers. Nibblers are people who bid a low amount that raises the price but is unlikely to win the auction. I.e. someone bids 30€ on an item that you would bid 50€ on. This raises the price to 30€ because of your automatic bids. The nibbler then bids 35€ to see if that was enough and it still wasn't, losing you 5€ from early bidding. If the nibbler thought he could get the item for 30€, you would only have to pay 30.50€ to beat him. The other reason is that you don't want to lock up your money since a long auction timer means you can't start bidding on the next auction in case you lose.
This was my thought as well, sniping bots have been around for as long as ebay has. Perhaps though, the sniping bots don't cause as much load on ebay's infrastructure?
I think AI is going to level the playing field with all these bots that have been used for things like this (including scalpers for those low supply/high demand items), and retailers will (hopefully) have no choice but to address the issue once everyone starts to use/abuse them.
I can only hope.
Scraping and buy for me bots cut out eBay. Sniping bots don't.
sniping bots keep people on ebay.com
Banning AI agents is the new "banning mobile browsers." Companies tried that too in the early smartphone era - remember when sites blocked mobile user agents to force desktop views?
The businesses that win will be the ones that build AI-agent-friendly interfaces, not the ones that try to ban them. eBay is protecting their ad revenue and impulse-buy funnel in the short term, but they're ceding the future to whoever figures out how to make agent-compatible commerce work.
Every product and platform will eventually have an "agent API" alongside their human UI. The only question is who builds it first.
> remember when sites blocked mobile user agents to force desktop views?
You mean today, when web apps refuse to render to mobile browser agents, forcing redirect to mobile apps, but work fine when toggled to a desktop agent?
TikTok, Instagram, etc.
Reddit.
I think more likely we will unlock browser agents and no company will develop an agent api. They will have a user facing website used by agents or humans the same way.
This also completely sidesteps any actions Ebay decides. It will have all credentials for them and mfa. To ebay this will look identical to the user with no real way to stop it.
Except speed and "translate the following to klingon", etc
If they don't currently see a way to make agent commerce work, the smart move is to stay alive long enough for someone else to figure it out then buy them or simply copy it. And if it never works out this will have been far cheaper in the medium run.
I do not endorse this approach but it is well established in the tech industry.
Tried selling on eBay as a regular Joe lately? Item sold for roughly $190 and I lost $45 in fees - I didn't even have a premium ad or pay for any of the boosting.
No wonder Facebook marketplace has destroyed them
The problem is with items that have a national market but not a local one. For example - there may be very few local buyers who will pay a decent price for a vintage slide rule, but many on eBay. My general strategy is to list on FBM first for the eBay price that I hope to get, and then accept offers down to 75% of the price. If I don't get any bites after about a month I switch to eBay.
This. I was selling an obscure book once. I doubt there is anyone local that would be interested in it. It was sold on eBay within a week.
Same for a half functioning Xbox. No "normal" person would want that. But apparently, on eBay, something like a dozen people took serious interest in it, and it was sold in a few days in "parts only" condition. For sure I didn't like how much the transaction fee I paid, but at least I got rid of it for a decent amount of money.
At least in the UK, I don't lose any of the selling price to fees, 0%. The buyer has additional buying fees on their side and postage is included in the final price.
Even if they use Paypal?
As someone who hasn't sold on eBay in a looooong time but was thinking about it for some stuff I haven't been able to sell on Marketplace, their pages and pages of fee structuring were intimidating. What was the breakdown of that $45, if you don't mind sharing?
Sure,
I listed the item as $185.00 + $10.00 shipping.
Order total = $195.00
- Transaction fees = $32.44
- Postage label = $14.65
Postage I can understand.
2 replies →
Yes sold a fair bit and never had issues with fee deductions. Think it’s mostly deducted seller side
Who cares, it's my browser, it is for me to decide what I run, not for eBay. LLM, AdBlock or whatever else I want I will run it.
Of course you can run anything you want on your computer. But it is also ebay's right to decide whether or not its computer will allow requests from your computer.
> But it is also ebay's right to decide whether or not its computer will allow requests from your computer.
That is dangerous thinking right there: Ebay does not have rights.
Of course ebay may do it anyway, and it may take time for justice to correct things, but it is not Right, nor their right, to violate law even to protect themselves.
7 replies →
One pattern I keep noticing is that when the future gets harder to predict, the first visible response is not innovation but a tightening of legal and risk frameworks. Platforms start hardening contracts and banning edge behaviors because internal models can no longer reliably track downstream consequences. A subtle form of constraint collapse where rules substitute for orientation.
Feedback still flows through metrics and policies, but it no longer carries enough of a cue to guide real learning, so it gets inverted into compliance and arbitration instead. Risk management becomes the substitute for understanding, and when context collapses, meaning drifts.
I'm not at a point that I trust an AI agent to buy something for me on a place like eBay...
ex: "...parts only", "foo for bar", ...
How likely am I to get the wrong product entirely or something that I can't actually use.
Who cares? eBay has free returns with postage paid by the seller for any item, so it really doesn't matter if you get the wrong thing etc.
Of course this is why they ban it because the odds of you getting something wrong is too high and sellers + eBay would lose out.
You don't have to obey user agreements.
Is so far as you don't have to use the site, that's true, but they are legally enforceable, and you could absolutely be sued for breaking them if you upset eBay enough.
Then most websites can sue OpenAI and Anthropic, but in reality, they can't.
1 reply →
I'm sorry Dave, I'm afraid I can't do that.
You do if you don't want to get banned.
I’ve been working on a small experimental gateway that sits between agents and customer-facing execution paths and forces decisions through policy + approval before anything irreversible happens.
v2 I just shipped adds:
• risk scoring on drafts • policy path traces • approval chain previews • highlighted spans showing what triggered the block • admin review endpoints
The motivation is exactly what people are pointing at here: once agents can transact, marketplaces end up banning them unless there’s a way to pause, inspect, and assign responsibility at execution time.
Curious what failure modes you’d want intercepted first if eBay or Amazon ever exposed agent purchase APIs.
https://authority.bhaviavelayudhan.com/v2/console
Speaking of which. Did anyone patent this 0-click buy method yet?
Meanwhile Google announces UCP to go in completely the opposite direction (or make marketplaces like eBay do so)
Anecdotally, I've noticed an uptick in my eBay feeds recently of items being immediately relisted after having supposedly sold. This has always happened occasionally, but within the past couple weeks, I've noticed it happening for like 25% of my watched items. I'm wondering if bots are buying things for which the owners are then canceling when it comes time to pay.
I've also had to return a few items for which eBay's AI-generated description was wrong in ways that couldn't be verified in the product's images. I can only imagine the increase in canceled/returned orders from all the different AI features and bots.
What is the use case for LLM agent shoppers? I can't imagine delegating the purchase of a used item to an AI (I'd be okay with AI identifying the best deals for me to review). This must be something for people who are doing something at scale like flipping items on Ebay or drop shipping.
I imagine this type of automation existed before LLM agents came along - what do they add? Is it just the ability to evaluate the product description? Item quality is already listed as a categorical variable.
"Hey, ChatGPT/Grok/GeneriBot4000, please watch for a great deal on a 1982 stratocaster guitar - must be in good or better condition, $600 or less, and if you see it, go ahead and buy it without confirmation"
Ongoing tasks, arbitrage for mispriced postings in ways that aren't currently exploited that LLMs make feasible - by banning auto-buy, maybe they're attempting to delineate between human seeming behavior and automation, and giving AI permission to buy looks too much like a real person?
Seems pretty petty to me.
I have decent tech company salary but I don't even buy $10 books without checking everything. This week I almost bought a wrong book (manually) because how similar the title is. Automating stuff with AI is interesting, but I don't want the hassle of getting surprised and handling returns, if the item can be returned at all, especially on eBay.
2 replies →
I genuinely wonder, would you do that, really? Sure 600$ is not the end of the world for certain countries, but neither it is a sum I'm willing to just lose on random. What if the electronic parrot buys from an obvious counterfeit vendor or obvious scammer? Or what if it buys you a stratocaster but different? Or a random 1982 guitar? What if it ignores 600$? Or what if it buys 600$ item with 300$ shipping and 500$ customs from god knows where?
I've seen enough by now and I know that some people will just unleash LLMs on anything without almost no oversight. We can already see people use agentic IDEs with "do all the shit" flag, they would probably easily add finances to the list of automation.
But, honestly, would you?
Yeah I guess that makes sense for some people. I'm just not in a financial position where I'd let an AI buy a $600 used guitar without me taking a look at it first.
3 replies →
Yeah literally price mistakes being picked up right away. But also seems like a good way to get scammed.
2 replies →
"Hey ChatGPT, I need more glass cleaner"
*OpenAI issues a micro auction to glass cleaner companies and distributors to see who will bid the highest combined commision*
"Sure thing! I ordered some Glass Clean Plus from Target for you!"
[Recycling a joke from many months ago]
My mistake, you're completely correct, perhaps even more-correct than the wonderful flavor of Mococoa drink, with all-natural cocoa beans from the upper slopes of Mount Nicaragua. No artificial sweeteners!
(https://www.youtube.com/watch?v=MzKSQrhX7BM&t=0m13s)
2 replies →
Drop shippers who arbitrage between major and minor ecommerce platforms need to maintain their listings, re-price things, etc. They don't care if the AI gets it wrong sometimes as long as they more than make back the cost of deploying it.
So now imagine ten thousand of these jerks telling their AI of choice "hey go scrape everything you can and re-list it for 10% more". That's a lot of load on the platforms at both ends for listings that are unlikely to generate many sales.
But that also seems like a very inefficient way to accomplish this automation task from the drop shippers side too. What do you gain from the LLM that non LLM automation couldn't do more cost effectively?
2 replies →
How do ticket scalpers make money? It's an automation war. You can run arbitrage strategies at scale if you can scrape markets with bots that understand unstructured data. Even if trades go wrong sometimes it can be profitable on average.
"Hey ChatGPT I want to build my own personal cloud storage computer, buy all the hardware for me then walk me through building and configuring it. My budget is $600, try to get the best deals and make sure that all the parts are compatible. I'm fine with used parts as long as they're a good deal and are in working order."
You would really do this? You'd not even want to at least briefly review the cart before making a $600 purchase of used computer hardware?
2 replies →
You will end up with expensive scam parts.
"Buy good quality for less than market value and don't get scammed" is so far beyond AGI complete that I cannot fathom how people think it will be reliably doable by LLMs
I think it's telling though how everyone's example of how an AI agent can help you shop is "I want it to take advantage of someone else's mistake or knowledge asymmetry to make me free money".
Are you all just embarrassed used car salespeople?
Does it need a known and enumerated use case to be allowed? I don't like that implication.
An AI that shops for a blind user, for one free example of the untold and unexplored uses of new technology.
Can’t charge for something if you’re giving it away for free.
Data’s the only moat left. Companies like stack overflow need to build revenue streams from AI or they will cease to exist.
By banning bots and then licensing some kind of access, eBay can protect itself from merely being a listing point that no human actually visits. Tailwind and their adverts via docs model, eBay and its promoted listings model, we’re going to see businesses adapt or die on this.
Note that Stack Overflow has already ceased to exist. https://xcancel.com/marcgravell/status/1922922817143660783
LLM-initiated purchases probably rack up chargebacks, support calls, etc for mistakes the LLM makes. I'm not surprised they want to limit it.
I might be out of the loop, but are agents actually out there buying stuff from "unwilling" vendors at any significant scale? I thought that was still mostly limited to opt-in partnerships with retailers. Still, eBay might be anticipating the issues you mentioned and trying to get ahead of them.
Not commonly known (I work in this space), but yes.
Agents are being used to automate things like non-cash account balance arbitrage, stacking and abusing marketing promotions, triangulated purchasing schemes, and purchase-refund arbitrage schemes at an increasingly large scale.
They should treat it like freight forwarders. They're allowed, but when you use one you don't get the return policy.
More likely, they want to be the exclusive provider of LLMs that can purchase off of eBay, or at least charge for API access.
They may have an inkling that the big LLM companies will want to pay for future/past data... I imagine either Google or OpenAI has something predictive and shopping-related in the books.
This; "certified / authorized by eBay" and then agents have to pay access to the catalogue
Right -- this seems more of a protective measure than something they will proactively enforce.
If you have a well-behaved agent that uses a browser to buy on eBay, I doubt that will cause issues. But if it leads to issues, they can point to that clause instead of having to help repair the issues caused by someone else's software.
And we enter the predicted cycle of "new thing that's going to Connect Everything"
Just like faxes, the internet and the world wide web, sure, this new thing Could connect everything, but that's not nearly as much of a technical problem as a sociopolitical one. Same as it ever was.
How can they tell its AI buying if the agent uses the right user agent and works through a real browser?
It may just be to stop third parties from creating a whole business out of "shop for me" AI bots. Individual users getting away with it might not be a problem, but with it being against ToS, it'd be a lot more shaky to build a business out of it.
In fact, it may just be that eBay wants to be the business selling AI "buy for me" agents.
Oh so just “human review “ it
This is a case where it may be that people are outsourcing shitty user experiences to an AI.
I’m not a huge ebayer but I’m usually watching one or two auctions at a time. The problem is that you can’t disallow marketing notifications. So, if I want to be usefully alerted for a new item in a search, or that I’ve been outbid, or the imminent end of an auction, I’ll also be getting notifications and emails about all kinds of shit I don’t care about. $5 off coupons (that only apply to 8 items that I don’t want). “You might like this!” notifications (spoiler: I never do). Group buying times (who cares?).
So I either disable ALL notifications (and have an LLM write a script that crawls searches manually and much more appropriately notifies me on its own), or I enable notifications and get a bunch of trash spam.
As it relates to specifically to buying, we’ve known for a long time that we’re all up against some kind of bot that’s timed the exact last moment and amount to outbid us. It’s no fun.
I’ve been an eBay user since 1998 and it’s been on a very slow roll of enshittification since then.
Make your experience better for humans and maybe we’d be less inclined to outsource negative experiences to AI.
Only until Agent Commerce Protocol is more standardized: https://www.agenticcommerce.dev
I don't see why this thing has much reason to be focused on agents or AI - a standardized API for ecommerce is useful regardless of that usecase.
I loved early eBay but gave up on it once became clear how rife it was with bid snipers, fraudsters and stolen goods.
example of focused leadership - a commenter already noted how wondering listings drive revenue
if it was some Bozo executive as we see at most tech companies - they would be advocating to implement the Open Agentic Commerce whatever being pushed by google while not noticing its killing their own company
Somehow, there's a relevant XKCD: https://xkcd.com/576/
That's the first thing I thought of. I remembered the title text line about a Bobcat, but I had forgotten everything cost only $1, with free shipping.
This does make me wonder. I see on HN (and hello if you see my comment) people who use screen scrapers or screen readers to read and use the web. I would be REALLY interested to know how many of these users use any of these newer AI browsers like Comet, I forget what the one from ChatGPT is called, but I know as a regular user I can make Comet do automated things like price comparisons across tabs and websites. I could totally see the immense value in someone who relies on a screen reader to access the web having access to an AI powered browser, but I don't know that any of them are designed with these users in mind necessarily.
My question then becomes, does this policy violate the ADA for those users in particular? UIf it doesn't today, should it tomorrow? Especially if these AI browsers actually become viable for those users. Will there be a future where if you're protected by ADA you can be cleared to use a more automated browser? I would imagine a sane rule for such an exception would require you to fully identify yourself to the website in order to prevent abuse by bots pretending to need that type of access (the good old "trust me bro" problem). Or maybe they get to use it but it becomes more rate limited to the average user speed or whatever.
I'm sure this will definitely not be ignored and taken very seriously
Smells like an opportunity
They want to avoid the fate of tailwind.
https://news.ycombinator.com/item?id=46527950
Dumbest thing in the world, not wanting buyers who are ready to complete transaction the moment they find what they want. When my car broke down and cost too much to repair I described what I need - low mileage, big trunk, 40+ MPG, under certain price, close to dealership where my broken car sat in service. That I gave the query to Grok 4 Expert (because it does heavy web scraping), found a 2020 Ford Escape Hybrid and drove it away two hours later, because rationally speaking missing a lot of work in this economy is a bigger financial risk than missing $3000 on manual bargain hunting vs a good AI hunt. If any dealers blocked scaping on sales pages, too bad for them. Speaking of which, any good bargain / secondhand market AI agent friendly API?
No one wants AI to spend their money, checked or not. The few people who would want AI, want AI to save them money
not the User Agreement!
Impossible to enforce, they can read browser windows and pass captchas
Probably less about direct enforcement, more about after the fact. Ebay doesn't want to deal with charge backs for hallucinate purchases
Yeah, they're hedging against "AI purchases". eBay has already been dealing with automated/bots for years.
> Ebay doesn't want to deal with charge backs for hallucinate purchases
A charge back doesn’t mean buyer always wins. Imagine if credit card companies also pass a rule - “LLM or AI purchases are non-refundable”.
On a different note - once I tried to cancel an eBay order within a minute, both eBay and seller declined. It’s so fked up with them.
This. These kinds of "rules" are basically useless because they are not enforceable. It's exactly like having speed limits but no cops.
> Impossible to enforce
Maybe, but a policy's or law's validity or importance are not contingent on them being enforceable.
No, but when instituting bullshit policies or trying to regulate natural/normal behavior for selfish gain, it helps you if you can enforce the policy, otherwise people will just ignore it.
eBay is hyper aggressive about fingerprinting, they will catch things like it trivially. Browsers leak all sorts of information like what sockets are open on localhost, making yourself look like an actual person is very challenging to someone motivated to detect you.
LLMs don't need browser automation though. Multimodal models with vision input can operate a real computer with "real" user inputs over USB, where the computer itself returns a real, plausible browser fingerprint because it is a real browser being operated by something that behaves humanly.
3 replies →
If Amazon has not defeated Perplexity yet, eBay is not going to stop anyone.
Scalpers work around it. But a normal person looking for something? Fuck!
Man I hate this
...haven't bots been buying things off eBay since the 90s?
This ban isn't about 'fairness' or bot protection; it's about protecting the Impression Funnel.
Marketplaces like eBay are designed to monetize 'Wandering Attention'—sponsored listings, 'customers also bought', and sidebar ads.
An AI Agent represents 'Laser Focused Attention'. It executes a transaction with zero wandering. It effectively turns the marketplace into a commoditized backend database / dumb pipe.
From a Growth/Unit Economics perspective, an AI Agent is a nightmare customer. It has zero probability of impulse buying and generates zero ad revenue. They are banning them to save their business model, not their inventory.
We've been getting complaints about your account posting generated comments. Is that the case?
(Generated comments aren't allowed here - we only want commenters who write in their own voice. More explanation at https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que... and https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...)
I suspect they might lose money on returns too, which are probably more likely if an AI misunderstands what the buyer wants or misjudges quality or can’t detect fake listings etc
I don’t know if there are other ways eBay could lose money on returns. But my single data point: the very first thing I sold on eBay (a manual lever espresso machine) got returned because the buyer clearly didn’t know how to use it, and claimed it was defective. And because eBay has a money back guarantee, they just reached their hands into my back account and withdrew the earnings from the sale + the shipping costs for the delivery to the buyer + the shipping fees for the return. They even kept their listing fee and the sales tax. So… I don’t think eBay stands to lose money directly from returns. Maybe they risk pissing too many sellers off with an increased rate of this horrific experience?!
20 replies →
"Mint condition MacBook Pro M5, 64GB RAM, 2TB storage, midnight black, box only" and it sells for $1800 because someone didn't see "box only".
6 replies →
Another reason I could think of is Security. There is a bunch of cheating goes on there. As a seller I lost my laptop to a scammer. Seller paid be until I shipped & cancelled the transaction. Buyer asked me ship it to their son’s address. Since I didn’t use buyer’s address registered in their eBay account eBay/Paypal didn’t pay me either. AI accelerates these scams.
Or probably even wrose, it actually shifts the attention and the wandering. That phase will happen inside the LLM, where the LLM decides which link to suggest, i.e. whoever pays the LLM the most. And worse yet, that will apply not just for products, but for platforms, so if amazon pays chatgpt more than ebay does, there goes your sale.
An AI-Agent browsing eBay for a "widget" for a given individual will also likely not be browsing eBay's advertising listings (sponsored and promoted listings [1]) which would potentially equate to a loss in ad revenue for eBay. So there is likely a "protect the advertising moat" aspect to their "ban" as well.
[1] Given how hard eBay pushes sellers to purchase the sponsored and promoted listing tiers (at an additional fee of course) implies they make some nice revenue stream from these advertisements.
This assumes the LLM ecosystem stays centralized. Open source models running locally or on user-controlled infrastructure flip this - the agent works for you, not for whoever pays the model provider.
The race is already happening: open weights models are getting good enough that "your personal shopping agent" doesn't need to phone home to a company with ad incentives. The future probably looks more like ad blockers than ad platforms - agents that aggressively optimize for user preferences, not platform revenue.
Or if eBay blocks ChatGPT while Amazon doesn't.
1 reply →
Open source models can short circuit that hopefully.
Nilay Patel is calling this "the DoorDash problem" and has written an essay on it here: https://www.theverge.com/podcast/823909/the-doordash-problem...
Fewer companies get the chance to enshittify my experience? Sign me up!
1 reply →
I'd say it mostly has to do with limiting their own liability and reputational damage if an AI bot "hallucinates" and places hundreds of incorrect orders and sellers get hit with negative ratings and refund requests due to no fault of their own.
Stated more cynically, many platforms have an interest in attention hijacking. Done well, agents' 'laser focused attention' could help users avoid wasting time (wandering attention) and money (impulse buys). This is a good thing, even if it dings revenue of some existing platforms. If a company's business model is impulse buying and ad revenue (this isn't eBay IMO), then good riddance.
Great analysis of the real motivation here. But this feels like the record labels trying to ban MP3 players. You can protect the impression funnel today, but the trajectory is clear - consumers will increasingly delegate purchasing decisions to agents, and the platforms that adapt will capture that flow.
The marketplace that builds "agent-friendly" commerce (verified listings, structured data, transparent pricing, API access) becomes the default backend for AI shopping. The one that bans agents becomes a legacy system humans have to manually navigate when the agent can't help.
eBay's current business model may be a "nightmare customer" for AI agents, but that's a problem with the business model, not with the agents.
>The marketplace that builds "agent-friendly" commerce (verified listings, structured data, transparent pricing, API access) becomes the default backend for AI shopping.
I'd like to believe this, but claims like this have been made since the early days of internet commerce. After all, it's not hard to specify structured data about items and run queries against it. But it largely has not materialized outside of a few special suppliers.
You can't actually search Amazon or eBay or Wayfair for things with specified dimensions or characteristics. You can, however, find lots of listings for things like "Gzsbaby 6 Piece Jumbo Dinosaur Toys for Kids 3-5 and Toddlers, Large Soft Dinosaur Toys for Lovers - Perfect Party Favors, Birthday Gifts "
Perhaps this time is different? But why is it different? What economic incentives will lead to good structured data and transparent pricing, rather than whatever the AI equivalent of glurge/slop listings is?
3 replies →
I don't think the 'wandering attention' dissapears, rather it is pushed to the LLM product. It's more of a competitive transfer from the incumbent product category to the new one, it's not that the new product category 'fixes' the 'problem'
If I am able to find what I’m looking for and purchase it via AI why must I be subjected to advertising & promotion of items in have zero interest in? Amazon has made finding what I want painful. I suspect eBay is just as painful (I see to work there).
—It's not X, it's Y?
Comment definitely reads like AI
Stop the ai witch hunt madness
This reply has hallmarks of AI slop. Green name, 2x its not X its Y, em dashes.
If you want to be a productive member try commenting what you put into your prompt instead of the slop that comes out.
Also classic ai drivel: This is about protecting the business model, not their inventory. My brother in AI, eBay doesn't have inventory. They're a platform.
I didn't catch that on first read, but I see why you'd say that. LLMs are ridiculous in the constant usage "it's not X it's Y" -- It's in almost every response from Opus 4.5. "It's not X it's Y" is ruined for regular writing.
I'm also skeptical of anything that claims to reliably detect AI writing. FWIW, I plugged the comment into Pangram Labs, which claims to be the most reliably and seems to have worked well before. It categorized the comment as 100% human written with medium confidence.
What does green name signify on HN? I've always wondered.
1 reply →
This slop so annoying to see, and tiring – but worth it here – to call out.
Dead Internet theory
100%.
Hallmarks of AI or AI slop? Or are you suggesting that AI is slop by definition?
Let's assume it's a bot. Is the point it's making unreasonable? Is it really unreasonable to refer to eBay's listings as inventory?
Yeah, it really says something about the state of HN these days that this account racked up (net) 261 votes in nine days writing absolute drivel.
Why not share your own thoughts rather than this LLM slop? Or maybe this is just a bot account. Either way, disappointing to see on HN.
The bots are learning to defend themselves on social media
Seriously, how is this the top comment?
That's a cynical take, so it will probably get upvoted, but what are you basing it on?
Ebay is a pretty eclectic marketplace and I can think of a number of possible reasons that have little to do with ads. For example, they may be worried about high error rates, and thus buyer and seller dissatisfaction. If I instruct an agent to buy X, eBay is almost never interchangeable with Amazon or Target.
They have no problem surfacing their listings on Google Shopping.
But, ads directly correspond to revenue stream, and a loss of ad "impressions" would result in a reduced revenue stream, so a "protect the advertising" response is not at all unusual to consider as a portion of their (eBay's) reasoning for this ban.
Given how hard they push sellers to purchase their "extra cost listing enhancements" (i.e., purchase to have your listings show in the "advertisement" spots) it appears that they may make a decent revenue stream from these advertising angles. An AI-agent could find listings without going through the advertising displays and as such cut into this revenue stream.
Is my primary user agent, my web browser, still allowed? /s
Maybe if you'd actually read the agreement
> In connection with using or accessing our Services you agree to comply with this User Agreement, our policies, our terms, and all applicable laws, rules, and regulations, and you will not...
> use any robot, spider, scraper, data mining tools, data gathering and extraction tools, or other automated means (including, without limitation buy-for-me agents, LLM-driven bots, or any end-to-end flow that *attempts to place orders without human review*) to access our Services for any purpose, except with the prior express permission of eBay;
This is so ironic, eBay generates AI descriptions for the things you are selling which is so stupid already.
Hasn't eBay's traffic been 80% bots since day one? I haven't participated in an auction in forever because even 20 years ago you were guaranteed to get sniped by a bot on anything except actual garbage.