This is a great piece. Hard agree on seeing the difference in how you optimize for one vs the other. Some of my clients are showing up and getting traffic from GPT and im noticing a few patterns across them.
1. Contextual relevancy - Like you mentioned above with AI handling longer queries, Im finding that this is making it extra important that the content on the article is relevant to a longer, more specific query. Having the information more specific to pains and potential use cases the user could be searching for. Mainly because a normal Google search could be "best CRM software for enterprises", while the GPT search is "I'm leading the marketing team at [x company] and we need a better solution to pipeline management because we're struggling with [x], [y], and [z]. So im finding making sure those elements are included in the article are extra important.
2. Authoritative sources - When I was looking at some deep research activity from GPT I asked it for [competitor] alternative because that's a piece I wrote for my client and I wanted to see if it would surface it. I typed it as a longer search query, then it prompted me asking for deeper context (what integrations did I want, how big is my org, etc...) It did show up as the first result but the thought process it went through (15 min long) was fascinating. Not only did it check the client's website for context, but it kept checking G2, and Capterra too. It would even go to competitors' website to validate that they didn't have what I asked for.
All that to say, I'm definitely seeing a slight shift in how you optimize for an LLM vs Google (or other general search engine).
I think it's a bit of a narrow perspective, looking at winners in regular 'pure' search vs LLMs. I see AI as an extension of search, the same way that Google Maps is an extension, and Google News is an extension. Distinct verticals of SEO, with their own specific tactics and strategies, yet all built on the same foundation. AI is also its own distinct vertical with its own specific tactics and strategies, yet built on the same foundation. We don't need to rebrand SEO with another instrutable acronym that will rapidly be associated with snakeoil salesmen and scammers. We just need SEO to do what it's always done: evolve.
The 4 categories of websites you looked at all rely on classic 'blue link' Google rankings for their traffic. I think that's a narrow perspective and, in my view, doesn't form a valid comparison to what LLMs are trying to do.
I think LLMs are better understood as a new extension of search, like maps and news were in Google. The sites that get a lot of visits from Google Maps are not always aligned with 'winners' in regular ten blue links.
Same with news publishers that achieve success in Top Stories and Discover. These are additional, expanded surfaces of the search ecosystem.
LLMs are a new expanded surface. It'll have its own winners and losers, separate from classic search, as is the case with maps and news. But it's still a part of SEO - a broader SEO that encompasses LLMs, just as previously SEO expanded to encompass maps and news and images and videos and all the other fun stuff.
Ah, I see. Which categories would you say don't rely on classic "blue link" ranks? I'm not sure I agree with the characterization because queries related to handbags mostly return Google's new shopping experience, but it was mostly my goal to compare "classic SEO" results with LLMs and which brands are present more/less in which of those to see if SEO is the same as GEO/AEO/LLMO, etc.
I agree that LLMs can do much more than search and that they're more and more used as an add-on, but in some cases also as a substitute for Google. And I think it's important to highlight that they're not all the same and the rules are similar but converging.
Hey, thanks a lot can you elaborate on what this authority signals you believe are relevant for AI?
Authority signals: While implemented differently, both systems rely on signals that indicate trustworthiness and expertise.
I guess this is what you mean with branded searches/brand popularity, but is that it? Which others do you see?
I imagine what people are saying is also extremely relevant. If you have a brand highly associated to being cheap in online mentions/content by pure probability then it should have a higher likelihood to appear in prompts focused on that characteristic. This would mean mentions in media, especially those associated with what your brand does would increase in value, compared to SEO, very significantly .
Well, one authority signal is that a brand is mentioned on a (reputable) publisher. The data shows that publisher domains are weighted higher in training data. It seems that social media content is weighted lower, on the other side.
This is a great piece. Hard agree on seeing the difference in how you optimize for one vs the other. Some of my clients are showing up and getting traffic from GPT and im noticing a few patterns across them.
1. Contextual relevancy - Like you mentioned above with AI handling longer queries, Im finding that this is making it extra important that the content on the article is relevant to a longer, more specific query. Having the information more specific to pains and potential use cases the user could be searching for. Mainly because a normal Google search could be "best CRM software for enterprises", while the GPT search is "I'm leading the marketing team at [x company] and we need a better solution to pipeline management because we're struggling with [x], [y], and [z]. So im finding making sure those elements are included in the article are extra important.
2. Authoritative sources - When I was looking at some deep research activity from GPT I asked it for [competitor] alternative because that's a piece I wrote for my client and I wanted to see if it would surface it. I typed it as a longer search query, then it prompted me asking for deeper context (what integrations did I want, how big is my org, etc...) It did show up as the first result but the thought process it went through (15 min long) was fascinating. Not only did it check the client's website for context, but it kept checking G2, and Capterra too. It would even go to competitors' website to validate that they didn't have what I asked for.
All that to say, I'm definitely seeing a slight shift in how you optimize for an LLM vs Google (or other general search engine).
thanks for sharing your experience. seeing the same thing(s) on my end!
I think it's a bit of a narrow perspective, looking at winners in regular 'pure' search vs LLMs. I see AI as an extension of search, the same way that Google Maps is an extension, and Google News is an extension. Distinct verticals of SEO, with their own specific tactics and strategies, yet all built on the same foundation. AI is also its own distinct vertical with its own specific tactics and strategies, yet built on the same foundation. We don't need to rebrand SEO with another instrutable acronym that will rapidly be associated with snakeoil salesmen and scammers. We just need SEO to do what it's always done: evolve.
Agree with your takeaway, Barry!
Tell me more about "looking at winners in regular 'pure' search vs LLMs". I didn't get that :).
The 4 categories of websites you looked at all rely on classic 'blue link' Google rankings for their traffic. I think that's a narrow perspective and, in my view, doesn't form a valid comparison to what LLMs are trying to do.
I think LLMs are better understood as a new extension of search, like maps and news were in Google. The sites that get a lot of visits from Google Maps are not always aligned with 'winners' in regular ten blue links.
Same with news publishers that achieve success in Top Stories and Discover. These are additional, expanded surfaces of the search ecosystem.
LLMs are a new expanded surface. It'll have its own winners and losers, separate from classic search, as is the case with maps and news. But it's still a part of SEO - a broader SEO that encompasses LLMs, just as previously SEO expanded to encompass maps and news and images and videos and all the other fun stuff.
Ah, I see. Which categories would you say don't rely on classic "blue link" ranks? I'm not sure I agree with the characterization because queries related to handbags mostly return Google's new shopping experience, but it was mostly my goal to compare "classic SEO" results with LLMs and which brands are present more/less in which of those to see if SEO is the same as GEO/AEO/LLMO, etc.
I agree that LLMs can do much more than search and that they're more and more used as an add-on, but in some cases also as a substitute for Google. And I think it's important to highlight that they're not all the same and the rules are similar but converging.
Hey, thanks a lot can you elaborate on what this authority signals you believe are relevant for AI?
Authority signals: While implemented differently, both systems rely on signals that indicate trustworthiness and expertise.
I guess this is what you mean with branded searches/brand popularity, but is that it? Which others do you see?
I imagine what people are saying is also extremely relevant. If you have a brand highly associated to being cheap in online mentions/content by pure probability then it should have a higher likelihood to appear in prompts focused on that characteristic. This would mean mentions in media, especially those associated with what your brand does would increase in value, compared to SEO, very significantly .
Well, one authority signal is that a brand is mentioned on a (reputable) publisher. The data shows that publisher domains are weighted higher in training data. It seems that social media content is weighted lower, on the other side.
Thanks for the writeup Kevin! I'd like to throw LLMO ("Elmo") into the hat for a name. ;)
We could call ourselves muppets ;-)
Great read! Thanks Kevin! 👌
Thank you, Dragan!