Decorative feature graphic to accompany the post

AI SEO: how do you optimize for AI search results?

The future of search is clearer than ever. Google is going to attempt to answer searcher questions directly on the page with genAI.

The old playbook for SEO success may no longer cut it in an AI-driven world.

You should have 2 reactions to this:

  1. Target topics that drive clicks
  2. Ensure you’re included in the generated text

AI SEO is a critical emerging field. You need to move quickly to make sure you’re still relevant.

When Google recommends someone, how do we make sure they recommend you?


Google will generate answers based on the top search results.

This will make SEO more of a winner-takes-all game as gains will aggregate further to the results currently at the top.

These are our expectations for AI answers in search:

  • Google will “read” the top results to generate a response
  • The top 1-6 results will be used
  • There will be bias towards using the highest results in the generated answers BUT you may get more text from lower-ranked results where that text is considered novel or useful
  • There will be strong citation of sources

To win here, you need to be in the top 1-6 results and optimize your content for genAI to pick it as the answer.

Expect basic LLMs and limited capabilities – for now

Perplexity is a leader in AI-powered search. Perplexity does something similar to Google, summarizing top results. For example:

Image of Perplexity AI's summary of top results for search for Apple foldable MacBook rumors including sources and answer.

Google can do this, but you need to consider the cost of AI responses at their scale.

Google operates at a much larger scale than Perplexity and I don’t see them charging for search any time soon, potentially ruining what’s arguably the greatest business model ever. Any solution for Google’s scale needs to be fast and cheap.

You get fast and cheap by using smaller Large Language Models (LLMs). You should see Gemini 1.5 Flash as critical here: announced at Google I/O recently, it’s fast and cheap.

A smaller model is less capable, so you get a more basic summarization. Here’s a 1 paragraph summary of our post on how to not get screwed by AI using Gemini 1.5 Pro:

> The AI revolution is rapidly transforming SEO, presenting both significant challenges and opportunities. Agencies must embrace advanced technologies and strategic approaches to thrive. Key considerations include avoiding superficial SEO tactics, understanding AI’s role in content creation, focusing on conversions over rankings, and adapting to evolving search dynamics. Ellipsis stands out by leveraging tools like FALCON AI to predict commercial value and optimize content, ensuring clients succeed in this competitive landscape.

And here’s the same with 1.5 Flash:

> To ensure you’re not left behind in the rapidly evolving landscape of SEO, it’s crucial to ask your agency the right questions about AI integration. Seek clarity on how they’re leveraging AI to create differentiated content, their approach to adapting strategies in response to algorithm updates, their methods for calculating search volume, and most importantly, how they’re delivering tangible value beyond just rankings. Embracing AI’s potential while understanding its limitations will be key to staying ahead in the game and ensuring your content remains indispensable amidst the changing digital landscape.

The Flash response is much simpler and the summary lacks nuance.

This limits the utility of the generated response: it’s useful, but not too useful. It also simplifies how we can optimize for inclusion as there’s less sophistication to deal with, at least for now.

Broadly, you will see the top ranking content summarized, and the way to be included in the summary is to have a top-ranking result. We can unpack that further.

Top ranking results will be included more… mostly

You’d expect inclusion in the generated text to be weighted towards the top-ranking results.

The top 1-6 results are likely to be used (with no text coming from anything outside of these), but within that you’d expect the top 1-2 results to make up most of the text.

This all makes sense. Google has already ranked the websites against the search topic and presumably thinks that the very top results are the best.

But, genAI opens up the ability to understand content much better and it’s possible other factors like information gain and a much more nuanced take on content quality will be used instead.

The system prompt for ChatGPT’s browser function, which can look at the internet, includes the following:

You should ALWAYS SELECT AT LEAST 3 and at most 10 pages. Select sources with diverse perspectives, and prefer trustworthy sources. Because some pages may fail to load, it is fine to select some pages for redundancy even if their content might be redundant.

The key points:

  • “select sources with diverse perspectives”
  • “prefer trustworthy sources”
  • “select some pages for redundancy”

The prompt lacks explanation for “diverse perspectives” or “trustworthy,” leaving it to the LLM to decide.

This makes me think about content quality, ensuring it stands out from the “average” article (and from what ChatGPT generates), and the Experience, Expertise, Authority, and Trust (E-E-A-T) push we’ve seen from Google.

It’s all about making your content better and ending the “good enough” content that plagues the internet.

The end of “good enough” content

Your content must be differentiated. If your content is the same as everyone else’s (or worse, undifferentiated from ChatGPT) then you will have a problem.

Information gain refers to the new information in a piece of content compared to competing content.

Google filed a patent for this in 2020 and we discussed the need to react in a genAI world in April 2023.

If your content covers the same topics as everyone else, it will have a low information gain score. If it covers different topics, it will have a higher score.

Content with “diverse perspectives” is more likely to be included in the genAI results. This requires saying something different.

As AI gets better, so does Google’s ability to understand what’s on the page. This creates opportunities. More useful content will be rewarded.

Internal Google documents have discussed the difficulty in assessing the usefulness of content for a couple of years now. When there’s so much on the SERP, it’s harder-than-ever to identify which specific bit is useful.

GenAI featured snippets side-step this problem by extracting the most useful text from the top 1-6 results. If the text is the same, the answer will come from the top result. If your content is unique, you have a chance of getting featured even with a lower ranking.

If you’re “David” ranking 5 then you might be featured more in the AI featured snippet than “Goliath” at number 1, as a result of offering more useful content.

If you’re aligned with E-E-A-T trends then you’re in the right ballpark already. Information gain becomes another metric in the mix.

How you assess information gain is trickier. We’ve built internal tools to sit alongside our FALCON AI for this. These assess the competing search results for a target keyword, read them, and then score for information gain on a 1-5 scale.

We then output a score for each individual search result, and an average score for the keyword. This lets us quickly ballpark the level of new information in the competing results, and we can tailor our work from there. Sadly, I’m not aware of any SaaS products doing this, so you’ll have to work with us to get the benefits 😉

Seize the AI SEO opportunity

The rise of AI-generated search results is not a distant prospect—it’s an imminent reality that SEOs need to prepare for now. As Google relies on AI to directly answer searcher questions, the old playbook for SEO success will no longer suffice.

To stay ahead in this new AI-driven landscape, SEOs must adapt by focusing on topics that drive clicks, optimizing content for AI-generated snippets, and creating differentiated content with unique value.

While top-ranking results may have an advantage, the use of AI also presents opportunities. By providing novel information and in-depth, useful content, even sites not in the coveted top positions can be featured prominently in the AI-generated results.

Move quickly. As AI capabilities evolve, so will SEO best practices. Staying informed on the latest developments and refining your approach will be essential.

But one thing is clear: the era of “good enough” content is over. In an AI-powered search landscape, only the most relevant, useful, and unique content will thrive.

So don’t wait. Start optimizing your content for AI search results today. Embrace this shift as an opportunity to showcase your expertise, provide unparalleled value to your audience, and come out on top in the age of AI SEO.

Picture of Alex Denning

Alex Denning

Alex Denning is the Founder and Managing Director of Ellipsis®, a world-class SEO Content agency. Alex is the inventor of FALCON AI®, the next-generation AI SEO platform that helps you predict how your content will perform – before you create it. Alex is an international speaker and publishes regularly on the future of SEO. @AlexDenning on Twitter