How ChatGPT Finds Information

追加: 2026-05-28

[00:00:00]You know, I've been thinking, um I just the idea that people are are verifying um the answers that they're getting from LLMs. Part of it might be that it's just a black box. Like, we have some at least some semblance of understanding of how Google decides to rank in their visibility as people and as marketers.

[00:00:17]So, let's get into that piece of piece of the conversation about how does AI go about answering a complex question like this.

[00:00:26]Um so, Mike, do you want to kick us off? Like, what's your perspective? How would you How would you explain this to my mom?

[00:00:33]Oh, as far as how it works? Well, basically, the way it works is that the large language model takes the prompt that the user puts in and then it extrapolates to a series of what we call synthetic queries background. And then it performs all these searches and then it pulls pieces of content and feeds that to the language model to then generate the response. So, um that's why we say that like organic search is still a very big component of it, even though it does require like a different subset of things to do in order to get the visibility in those channels.

[00:01:09]Yeah. And that process, like we call that rag, you know, retrieval augmented generation. And the R, the retrieval, is why SEO is such a found is so foundational to good GEO. Cuz to Mike's point, at some point in that process, it's probably going to happen to in two or three or four cycles, there will be some core some kind of retrieval to some source document or more to the point, some source chunk of text that then gets pulled back into that response. So, I think it's really important as we're talking about how this happens, why it happens, uh to be successful in this era, I think it is really important that publishers understand that rag model. It's you're not just trying to get discovered in 10 blue links. Now it's you've got to win three separate actions. You've got to get retrieved.

[00:02:03]You've got to get selected from everything that's retrieved cuz the LLM does make choices and rank the information that it retrieves and then you've got to get cited.

[00:02:14]So, instead of having sort of one big job, you know, we've got three smaller separate jobs and see three separate things to think about like being retrieved, being selected, being cited.