The Secret Engine Behind ChatGPT? It's Google

ChatGPT relies on Google more than many people realize

ChatGPT started actively searching about a year or so ago. And throughout those months, we were all witnessing its quiet reliance on Google develop right in front of our eyes.

And it is not just searching… It is also a weird reliance on Google’s indexation and cache.

>>>> JOIN US to discuss this LIVE on LinkedIn!

About 7 months ago, Aleyda Solis did a test to find ChatGPT citing Google’s search snippet, word-for-word.

A week ago, I did a test to see if ChatGPT reads schema and found that ChatGPT refuses to see updates on the page unless they are indexed by Google. Basically, the steps were as follows:

1. Added a fake company name to a page

2. Asked ChatGPT to find that company name on the page

   1. ChatGPT confidently responded that there was no company name mentioned on that page

3. Requested re-indexation of the page through Google’s Search Console and waited for it to be done

   1. Immediately after it was re-indexed by Google, I used the same prompt, and ChatGPT was able to see the new name right away.

Which version of the page ChatGPT is accessing is a mystery (Google’s cache?), but the reliance on Google’s indexed version is obvious at this point.

Note: When prompted to visit a page, ChatGPT uses a specific bot that is supposed to open a page (regardless of its Robots.txt directives). It is not a search bot. So why does it query Google’s index instead of visiting the page directly remains a mystery.

Another finding from last week was that ChatGPT also scrapes Google’s sponsored results, as it is sending referral traffic from UTMs used only in Google Ads:

And let’s not forget that OpenAI’s browser openly runs on Google search (well, because it is built on Google’s Open Source software:)):

So what does this all mean?

Given that ChatGPT and Gemini are direct competitors, this weird relationship is very interesting to me.

But for SEOs, this means:

• Traditional SEO (technical, content, backlinks) remains not only fundamental but also key to LLM/AI visibility.

• Google is not going anywhere. It seems they have the keys ChatGPT is struggling with.

• Google is better equipped to win this race, unfortunately, as it doesn’t have to rely on third-party indices or data.

• Bing is not good enough (OpenAI tried to use it, but there must have been a good reason why they ended up with Google).

So why don’t ChatGPT citations 100% overlap with Google’s rankings then?

This is one of the more popular arguments among those who try to declare SEO dead. Different studies suggest an average of 30% overlap between ChatGPT citations and top Google rankings (which is huge, in my opinion).

My thoughts:

• These studies don’t seem to take “fan-outs” into account. ChatGPT doesn’t search for prompts or obvious underlying queries. It fans out to what it needs more info on (and pulls the rest from the training data).

• Let’s not forget that OpenAI has lots of partnerships with publishers and may pull citations from those. I once went down a rabbit hole of tracking down unindexed pages to OpenAI partnerships and succeeded.

SEO for AI / GEO / AEO strategy is not all about optimizing for Google. I am not trying to make this point, though. But, so far, Google seems to be the fundamental part of it.

Let me know if you’d like me to talk to you about other parts of it.