Comment by tikotus
6 hours ago
I've had two people reach out to me asking about one of my services. They both said ChatGPT recommended it to them.
My service does kind of exist. It's a small tool I created for a client while retaining full rights to the tool. So I created (vibe coded) a site around it, making it look like an established service. Even ran google ads for it for a while.
The service still doesn't show up on google with relevant search terms. There hasn't been another client. I forgot about the service. And then ChatGPT started recommending it to people.
I wonder what I did to achieve this. Did vibe coding the business page inject it into ChatGPT's training data?
> Did vibe coding the business page inject it into ChatGPT's training data?
No, at least not directly. Inference does not train models. It is possible that OpenAI may separately collect the chat data, clean it, and feed it back into the model for future iterations. Or they could have extracted URLs for future indexing.
More likely though, I suspect, is your site just managed to be indexed naturally, and LLMs are very efficient at matching obscure data to relevant queries.
Interesting. Maybe someone could run bot farms that ask variants of the same question and subtly nudge the model by replying reasons why the model's recommended service A is inferior to service B. Or other forms of adversarial question answers sessions.
It's quite possible that SEO-wise the site does not make the cut into top x Google results but still is findable and considered by ChatGPT when it does its searches.
Especially in a longer ChatGPT conversation or via deep-research or more agentic modes (e.g. "Pro").
ChatGPT spends quite some time and diligence on searching.
Great for content that is not hyper search engine optimized but still (or even more) relevant. It bubbles up.
I think the chatgpt backend basically includes indexed web like Google, or any other search engine.
Could Google be actively trying skip generated-looking sites/content?