James Dooley: GEO versus SEO. Is it the same thing? I always ask different SEOs and marketers this question. Today I’m joined by Benjamin Tannenbaum from AISO, and he has a lot of data and evidence around artificial intelligence and large language models. So let’s jump straight in. Is GEO just SEO, and do you have evidence to back that up?

Benjamin Tannenbaum: That’s a big question and there’s a lot of debate around it. I’ll give a short answer and then we can go deeper.

Benjamin Tannenbaum: The first thing you need to know is whether people are asking the same things to AI as they do to Google. Because if they’re asking completely different questions, even if the techniques to rank are the same, you’re still targeting something different.

Benjamin Tannenbaum: We covered this in the first part of the series. AI queries are more informational, more specific, and sometimes the entire funnel happens within one conversation.

Benjamin Tannenbaum: Now once you know that, the next question is whether the techniques are the same. Could you just use your existing SEO tools and strategies and appear in AI answers?

Benjamin Tannenbaum: The short answer is that it’s about 80 percent true. The techniques are very similar. But you need to aim for the right target.

Benjamin Tannenbaum: That’s where query fan out comes in. When you send a question to AI, it often summarises the question and sends background web searches. Those background searches are the query fan outs. If you want visibility in AI answers, you need to rank for those fan outs.

Benjamin Tannenbaum: There’s more nuance. Being number one is much less important in AI. AI might take the top 30 to 50 results from those fan outs. You don’t need to be number one. If you’re in the top 10, that’s often enough.

Benjamin Tannenbaum: There’s also a bonus if you appear multiple times in the top 30. For example, if you’re not in the top 10 but you own positions 20 to 30, you’re still very likely to appear in the AI answer.

James Dooley: Let me stop you there. In traditional SEO, if you’re not top five, you’re not getting clicks. Now you’re saying someone ranking 12 to 17 can appear in AI overviews and possibly get cited more than position one. Plus the searches are longer and more conversational. That sounds very different to traditional SEO. Would you agree?

Benjamin Tannenbaum: I agree it’s different in terms of targets and behaviour. But once you decide to optimise for the fan outs, the way you rank for them still relies on traditional SEO techniques.

Benjamin Tannenbaum: You still need crawlable pages, authority, backlinks, good content, and trust signals. So GEO is SEO applied to slightly different targets.

James Dooley: So the nuances are different but the tools are similar. How do you personally find query fan outs on ChatGPT and Gemini? Are you using a tool or looking in the code?

Benjamin Tannenbaum: There are several ways. The most manual is opening a session, inspecting network requests, and finding the JSON that contains the fan outs.

Benjamin Tannenbaum: There are also Chrome extensions that expose fan outs. But we wanted to generate thousands at scale, so we built our own internal tool and added it to our app.

Benjamin Tannenbaum: With our tool, you can input many prompts and generate fan outs in bulk. You can also see which brands or sources dominate those fan outs.

James Dooley: For anyone listening, what is the tool called?

Benjamin Tannenbaum: It’s called AISO. The website is getaiso.com. There’s a free trial with a few credits so users can test multiple prompts and see the fan outs generated by ChatGPT and Gemini.

James Dooley: On a previous episode, someone said ChatGPT averages three fan outs while Gemini averages ten. Why does Gemini generate more?

Benjamin Tannenbaum: In our data it’s about two for ChatGPT and around ten for Gemini. The likely reason is cost. ChatGPT has to pay for each web search it runs. Gemini, being part of Google, likely has much cheaper internal access to search infrastructure. So it can afford more fan outs.

James Dooley: If someone finds ten fan outs from Gemini, how should they optimise? Should they create new pages, subheadings, or address contradictory queries like scam or complaints?

Benjamin Tannenbaum: It depends on trade offs. First, take the fan out query and check the top 30 Google results. You’ll typically see three categories: brand websites, media listicles, and social media like Reddit.

Benjamin Tannenbaum: If your own website is already there, that’s the lowest effort. Maybe you just refine the content to match the specific intent more closely.

Benjamin Tannenbaum: If media listicles dominate, that’s a digital PR opportunity. It’s higher cost but potentially higher reward, especially if one media site appears multiple times in the fan out.

Benjamin Tannenbaum: Social media like Reddit is high risk. If you approach it badly, the community may react negatively. AI systems can detect that. A poor Reddit strategy is worse than none.

James Dooley: Someone previously broke query fan out into six dimensions: entity, attribute, freshness, consensus, reputation, and contradictory. Have you seen similar patterns?

Benjamin Tannenbaum: Yes, similar ideas. One category we identified is time dependent fan out. AI often adds location and date automatically. If your content is updated and clearly current, you increase your chance of appearing.

Benjamin Tannenbaum: Another category involves repetition or weighting. The fan out may repeat certain attributes to emphasise them. If your brand has a strong differentiator, you need to make that extremely clear in your content so it stands out in weighted searches.

James Dooley: On consensus, if one brand ranks number one but appears once, and another brand ranks 21 to 23 but appears three times, which is more likely to be cited?

Benjamin Tannenbaum: Our anecdotal evidence suggests frequency matters more than position. It’s like a probability model. If you appear once, you have one chance out of 30. If you appear three times, you effectively increase your probability of being selected.

James Dooley: Last question. On chunking and passages, do LLMs open pages or just use meta titles and descriptions?

Benjamin Tannenbaum: There’s a difference between ChatGPT and Gemini. ChatGPT often gives the impression it read the full page, but in most cases it only uses the search result snippets.

Benjamin Tannenbaum: Gemini more often reads deeper into the page, but even then it may stop at the search result summary if it finds enough information. ChatGPT almost never fully opens the page in most cases.

James Dooley: That’s fascinating. So GEO versus SEO has overlap, but there are real nuances around fan out, consensus, and how AI pulls information.

James Dooley: For anyone watching, we covered how AI searches differ from Google in a previous episode. In the next episode, we’ll talk about personalisation in large language models and why the same question can produce different answers for different users. Make sure you check the links in the description.