Where does ChatGPT source its information from?

Ahrefs just put out a handy breakdown on where ChatGPT actually pulls its answers from. Check the full thing here Ahrefs on ChatGPT sources.

Four spots total. First is the training data the model builds on before release. That covers public web stuff, partnerships, and user inputs. Your site could sneak in there if OpenAI's crawler hits it, but it's slow and indirect. Months later at best.

They block a ton of junk like paywalls and spam. Blocking GPTBot in robots.txt keeps you out of future training. But it won't stop your pages from showing up in ChatGPT searches.

Second is live web search. This kicks in for anything recent like prices or news. ChatGPT rewrites your question into targeted searches and fans them out. One in five of those messages ends up as a search.

Third is licensed publisher content. OpenAI pays for summaries on news and reviews. Only shows if those pubs get picked up.

Fourth is your own context. Prompts, uploads, or connected apps.

For brands chasing ChatGPT traffic, the play is ranking for those search triggers. Your pages need to be solid for Google first. Opting out of training might limit long-term knowledge but won't kill short-term visibility.

Knowledge cutoff sits around April or May 2026 depending on the plan. After that, everything runs through search.

SEOGami

Great breakdown! i was not aware that the purpose of the training data and the live web search is so distinct from each other however i feel like creating good content is still crucial for gaining recognition through AI powered search also what is the potential for niche websites to be referenced by ChatGPT?

1:3 ratio is pretty good.

I mean 3 areas we are able to influence.

OP you post the craziest threads on SEO. Mad props to you.

Something important happening in SEO, chances are there's going to be an @SEOGami thread soon with the proper breakdown. And the source always linked.

The live search part is the bit I'm still trying to get my head around. If ChatGPT rewrites a question into a few different searches, does it mostly pull from pages that already rank well on Google, or does it surface different ones? It'd be interesting to know whether a smaller site can actually get picked up there without ranking first.

I think the key takeaway is that brands should focus on creating useful trustworthy content rather than chasing AI mentions alone being visible in traditional search can also help with discovery through AI search features it would be interesting to see how these sources change over time

For years, we've been working on getting our websites to rank well on Google, but now we also have to think about how to get AI tools to find our websites, and not just that, but actually recommend them too.

This is a wild thread. The breakdown makes a lot of sense. Though I know that ChatGPT and other Agents get both their training data and search results from local and traditional contents. So, the more answers your content is able to provide, the more likelihood it will be cited by AI

SEOGami said:

Fourth is your own context. Prompts, uploads, or connected apps.

Click to expand...

This one is particularly nonsensical and has no real use. One time I was doing a research on a topic and chatGPT sneaked in the name of a brand I mentioned to it in the past on different chat. I was surprised and didn't anticipate that, so I went to chatGPT on a different device without login, asked the same question and got a result that didn't include that brand.

Went back to my original chat to ask why it included the brand, the response was that I had mentioned it in the past so it felt it's relevant.

This goes to show that in some reasoning aspect, these AI tools are not as smart as they should and could quite easily mislead a user.

Really interesting breakdown. It shows why traditional SEO still matters, but optimizing for how AI finds and selects sources is becoming just as important

I find the opportunity for smaller niche websites particularly interesting. Can a site with lower domain authority still get cited by ChatGPT if it offers highly relevant, well-researched content? Perhaps topical relevance will matter more than domain authority alone.

mostly training data plus live search when it browses lol, i wouldn't trust it for anything that needs a fresh source tbh

ChatGPT pulls its information from four primary sources: pre release training data public web scrapes and partnerships up to its knowledge cutoff, live web search for recent or real time queries, licensed publisher content paid news and media deals, and your own user context
prompts, uploaded files, and connected apps.

this interesting because getting picked up search much more achievable than becoming part of the training data. focus publishing helpful content and getting the brand mentioned on other relevant sites too. the more useful information available about a brand the better chance it has showing up ai answers.

Most AI answers are based on information gathered from the web.

Welcome to the AIEO era!

On organic search visibility first, especially relevant rankings, quality content, and brand mentions across trusted sources, getting cited in ai answers is becoming another visibility channel, so building authority beyond your own website matters too

that is interesting beiing visible in chatgpt search could become a big traffic opportunity.

This is pretty interesting. I think the biggest takeaway for website owners is that you still need to focus on traditional SEO. If you pages rank well and provide useful information, there's a batter chance they'll also get picked up when ChatGpt searches the web.