How to do SEO for AI Search Engines

This is a subject that is now impossible to ignore. I’ve talked about AI in my newsletter (go subscribe, obvs) a couple of times, but I wanted to put my thoughts in one place and cover a few important aspects of it, and give some practical advice.

This is an article for both marketers who implement, but also business leaders who strategise - those who are interested in how this technology could impact their businesses digital performance.

The thing with technology is that people like to use it when it works. Sounds obvious, but non-techy people don’t generally care how it works, they just want a good experience, and will more than likely keep using something that gives them that.

I think this is happening more and more with AI.

As more people are are using ChatGPT and similar tools as part of their day to day lives, they begin to use it in a way that makes it the obvious choice for a lot of searches, especially when it’s for the types of searches that lead to planning, research and longer decision processes.

The thing with AI though, is that it hallucinates. It literally makes things up. And a lot of people do not know this, and probably don’t care, as long as they get their results.

The amount of people using ChatGPT for medical advice - some studies show up to 25% of under 30s - is pretty shocking when you combine it with the fact that it hallucinates. However, people will believe what’s in front of them, because it’s easier to take it as it is - the tool works.

So this is of course a problem that AI companies are trying to fix, and as they do, the tools become more trustworthy as search engines, and more people use them overall.

Personally, I think AI tools will eventually become the default way to search.

Can We Optimise for AI Search Engines?

People are using the term Generative Engine Optimisation (GEO) to refer to the new practice of optimising for AI search engines. And I’ll use it in this article too, but I want to be clear that currently, all of the practical elements of optimising for GEO are things good SEOs have been implementing for bloody ages.

It’s important to keep that in mind as with any new terminology there are snake oil salesmen trying to take advantage of people’s lack of knowledge on the subject. They are using scare tactics to sell something that any good SEOs are already doing.

How Does AI Work?

AI, at its core, is a prediction machine.

It's trained on vast amounts of data and learns to recognise patterns. When you give it a prompt, it predicts the most likely sequence of words to come next, based on the patterns it has learned.

While this might sound simple, the scale and complexity of these models allow them to generate remarkably coherent and creative text, making them incredibly powerful tools.

We call them LLMs - Large Language Models - because they are trained on huge amounts of written language.

But it is important to start by remembering this: the AI tools we are using are primarily prediction machines, and this can of course massively impact quality and accuracy.

And these prediction machines can be trained on the text we can provide them. It might feel like magic, but as ever, it all boils down to text.

How does AI search work?

A search engine like Google has a deterministic index.

They look for websites, decide what they might be useful for, then maintain and rank an absolutely massive database of them using very complex ranking methods.

This type of search is likely still going to be suited to when you know exactly what you want.

If you want “black merino wool t shirts for men” there’s not really much ambiguity with your search term. Google does a good job of this already, and it’s pointless using an AI engine which may give more inconsistent results.

AI’s search functionality is probabilistic.

It responds to queries by predicting the next most likely word or phrase based on patterns it has learned, which can lead to more varied answers compared to your typical ‘list’ view.

This is much better suited to something you’re not quite sure you want.

The ambiguity of “which black t shirts for men would suit a 2 day camping trip” means AI search engines do a good job of helping you explore the subject and learn more.

For brands, this is important stuff to remember.

If you sell black merino wool t shirts, you should have a well optimised product page, well optimised category pages, and plenty of informational content around these that show how amazing your particular brand is whilst adding value for customers who want to learn more about the type of product. Why are you experts on merino wool? How does the AI search engine know that? More on this later.

I like this analogy a lot: brain vs. bookshelf.

A database is like a bookshelf: you store information in books, and when you need something you look it up exactly as it was written.

An LLM is like a human brain: It doesn’t remember exact text but rather concepts and relationships, making it good at generalising but sometimes prone to errors.

What is Grounding & RAG?

The biggest weakness of a standard LLM is that its knowledge is static - it only knows what it was trained on, and that data can become outdated. You may have seen warnings whilst using the tools such as “this doesn’t include results after 2023” or something like that. It can also hallucinate, as mentioned previously.

This is where grounding comes in.

Grounding is the process of connecting an AI model's output to a verifiable, external source of information. The most common technique for this is called Retrieval-Augmented Generation (RAG).

I’m not going to go in depth with this, as I think the basics is all you need to understand to make strategic decisions, but here’s how RAG works in very simple terms:

1. You ask a question - your search query.

2. The AI acts like an assistant - instead of guessing from its old, stored memory, it first does a rapid, live search for the most current facts on that specific topic.

3. It gives itself a briefing - taking that fresh, factual information and feeding it to itself.

4. You get a trustworthy answer - the final response you receive is now based on those real-time, verifiable facts, not on outdated data, making it far more accurate and reliable.

Of course, it still isn’t perfect.

For your business though, this is critical knowledge.

We’re doing the same as we were doing with SEO: for GEO we want your website to be the authoritative document that the RAG system retrieves.

If your content is clear, accurate, and well-structured, you have a much higher chance of being the source material for a grounded AI answer, complete with a valuable citation link back to your site.

Will we lose traffic?

Yes. The rise of ‘zero-click searches’, where a user's query is answered directly on the search results page, was already a significant trend and has been a big part of SEO for a long time.

But now, rather than just a small portion of the top of Google results giving you instant answers, the entire result is doing so.

However, it's not a simple story of traffic annihilation. The impact will likely vary hugely based on the query:

• Informational queries ("what is..."): these are most likely to be fully answered by AI, leading to the biggest potential traffic loss.

• Transactional queries ("buy..."): these are less affected, as users still need to click to a retailer to make a purchase.

• Complex research queries: for these, AI Overviews often act as a starting point. Users who need deeper, more nuanced information are the ones most likely to click on the source citations within the AI answer.

Can we measure GEO?

Currently, this is a difficult task.

On a simple level, you can head to Google Analytics, open the traffic report, add “Session source / medium” and see if there is any traffic being generated by ChatGPT. The most common referrers you’ll see are:

• chat.openai.com

• chatgpt.com

• gemini.google.com

• perplexity.ai

• copilot.microsoft.com

But when it comes to measuring performance in an actionable way, this is certainly harder than traditional SEO.

Traffic from all types of search engines will still remain very important, however there are a few things that will stand out for GEO.

The death of keywords

The internet is becoming incredibly "noisy" in that it is being flooded with generic content generated by LLMs. For good businesses, this should force a fundamental change in strategy.

The traditional Search Engine Results Page (SERP) - the familiar list of ten blue links - is becoming obsolete for many queries. Instead of sifting through websites, users now get a single, synthesised answer directly from the AI.

This means the old game of optimising for specific keywords to rank on a results page is disappearing - slowly - but it is going away. The goal is no longer to be on the list, but to be a trusted source within the AI's answer.

But let me be clear: as of writing, keywords are still very important!

Entities

When keywords are no longer the primary focus, what will be? The answer is entities.

An entity is a single, well-defined concept: your brand, your products, your CEO, or your physical locations.

The objective is to graduate from being a collection of web pages to being seen by AI as a distinct, reliable entity with specific expertise.

We can associate your expertise with the subject that is being researched, and look to become the consistent expert within a certain entity.

Citations

When your brand is established as a trusted entity, the new measure of success is the citation.

A citation is a direct link or mention of your brand within an AI-generated answer. You’ve probably seen these.

It is an endorsement that signals trust and authority, and which will hopefully drive qualified traffic to your site.

Earning these valuable citations requires you to provide something that AI cannot create itself: unique value. The two most powerful, defensible assets you have are:

• First-party data: publish original research, surveys, or analysis of your own data. This creates a source of truth that an AI can cite.

• First-hand experience: create content based on genuine, real-world experience (E-E-A-T, mentioned below). Your unique insights and opinions are something a generic AI cannot replicate.

GEO Tools: Which Ones are Worth Using?

The shift from measuring individual keywords to measuring something slightly less tangible, in combination with it being a very new area, means that we aren’t exactly at a point where we have tons of tools to choose from.

I’ve tried a lot of them, and will be doing a longer article with more depth on the features, but there are a couple that stand out so far.

Waikay

In particular I like their facts feature which stands out as something that will be useful in the long run to determine which model is hallucinating, and lets you take action on potential out of date brand information.

As with most these tools, a lot of the use is in tracking your brand. This is a shift in mindset that we will have to get used to as an industry, as mentioned previously in the point on entities.

Waikay does a great job of it, and basically the longer you leave the reports running the more insights you get, such as:

• Topic reports - create an action plan on how you can improve upon topics you want to show for.

• Competitor reports - who is doing better than you for topics, and which LLM is showing more or less of you.

• Link reports - which links are being cited regularly in results that mention you.

Check out Waikay: https://waikay.io

Sistrix

Sistrix have always been a great ‘benchmarking’ tool for me, and it seems they are taking a similar approach with their AI offering.

Basically you can enter a brand name and see a bunch of data. The interesting prompts section stands out as one of the better features, allowing you to see which prompts mention you and what position you are cited in - likely something that would be good to integrate into a content strategy.

They also have a visibility index too, but I won’t try and explain how that works just yet!

Sistrix AI is currently in beta, but check out their main site: https://www.sistrix.com

What is Model Context Protocol (MCP) and How Could It Change the Internet?

I’m not going to go into depth on this because I think it’s too early to be practical with it (for the majority of businesses), and quite frankly I’m still figuring it out myself. But the Model Context Protocol technology has the power to potentially change the way we interact with the internet.

MCP acts as a universal standard to provide a structured way for AI models to use data and functions whilst accessing websites.

Think of the AI agent browsing the web for you and looking at other websites; it will be able to find important data and take action on your behalf like a real assistant. MCP is what lets website owners present the data and potential actions to the assistant.

This could (and probably will, due to the hype) cause a shift for businesses, moving focus from building websites to creating MCP servers.

A company's key digital asset becomes its ability to offer its unique data and functions as a secure tool. Consequently, the goal of SEO & GEO will evolve from ranking webpages to becoming the most trusted data source within the AI ecosystem.

But I also think it’s important to consider brand again here - people will always want to visually browse a brand that’s worth browsing.

Practical GEO: What Can We Optimise for Right Now?

As I mentioned before, this is stuff we should have already been doing with ‘regular’ SEO. I have a vast checklist which I run through in SEO audits, and this is a tiny portion of that.

But as so many are asking the question, let’s address some of the AI search enginge specific practical things we can be doing.

1. Write Interesting Content

Firstly a note on content.

I wrote about this before in a post titled ‘Write for People’. I’m sure there will be a lot of people who disagree with me, but I believe the writing AI produces just won’t be as interesting as your original thoughts, opinions and expertise.

That doesn’t mean you shouldn’t use AI tools to help you write. Just think before you shove a load of AI generated blogs on your website. Are they actually any good? Are they accurate? Are you proud to say they represent your brand?

Google's official stance is that it rewards high-quality content, regardless of how it's produced.

However, their systems are specifically designed to demote unhelpful, unoriginal content created solely to manipulate search rankings.

The most effective and sustainable approach is to use AI as a tool to assist human experts, not replace them. Use it to:

• Brainstorm ideas and outlines.

• Generate a rough first draft that a human expert then rewrites and adds E-E-A-T signals to.

• Summarise complex research or find new research.

• Check grammar and clarity.

Never just copy and paste from an AI. It lacks genuine experience, and your content will likely be generic and easily identified as low-value by both users and search engines.

2. Structure Content using Chunks

For a very long time now I have been recommending this, and it’s just something that is even more important now AI tools are taking charge.

Essentially, break your content down to make it easier to read in chunks. This makes sense for users reading on screens (and getting distracted!) and computers digesting the information.

• Use a logical heading structure: use H1, H2, H3 headers in the correct order to create a clear hierarchy. Each heading should accurately describe the content that follows.

• Write concise paragraphs: keep paragraphs focused on a single idea. This makes it easier for the AI (and users!) to isolate a specific fact or statement.

• Use lists and bullet points: break down complex information into ordered or unordered lists.

Once again, this is stuff good SEOs have been doing for a very long time.

3. Think About Important Questions & Answers

• Lead with the answer: use the "inverted pyramid" style of journalism. Start with the most important information or a direct answer to a question, then provide supporting details.

• Use a factual, unambiguous tone: avoid overly florid or metaphorical language in informational content. State facts clearly and directly.

• Leverage structured data: implement FAQ schema for question-and-answer sections and How To schema for instructional content.

For the second point, it is important to note the ‘informational’ content aspect. If you’re writing creative copy, the last thing you want to do is make it more factual and dry.

4. Optimise With Citations in Mind

As mentioned previously, GEO is more about citations than positions. So we want to be proving that we can be the authority that gets cited.

• Publish original data and research: commission surveys, analyse data, and publish unique findings. This creates a source that others, including AI, will deem trustworthy.

• Cite your own sources: link out to authoritative studies, reports, and academic papers to show that your content is well-researched and trustworthy.

• Keep content updated: ensure your facts, figures, and dates are current. A "last updated" date is a strong signal of freshness.

5. Focus on Authority

An AI needs to be confident that you are a genuine expert on a topic. This is achieved by covering a subject with both breadth and depth, not just writing a single article.

Start by choosing the topics you want to be an expert on. Sounds obvious, but you actually have to be an expert on these for it to work.

• Use the pillar and cluster model: create a comprehensive ‘pillar’ page on a broad topic (e.g. guides), then create multiple ‘cluster’ pages (e.g. blogs) that delve into specific sub-topics within that guide.

• Fill content gaps: analyse what questions users are asking and what your competitors are covering. Ensure your content cluster addresses every conceivable aspect of the topic.

6. Create Content with Multiple Formats

AI answers are becoming ‘multi-modal’, incorporating text, images, videos, and tables. If your content is purely text-based, you are ineligible to appear in these richer answer formats.

So try out multiple types of content for your business. It just so happens users love multiple formats too.

7. E-E-A-T (again)

Demonstrating your Expertise, Experience, Authoritativeness, and Trustworthiness (E-E-A-T) is the most critical defence against being dismissed as low-quality content. I would go as far as to say this is the number one thing you can do for SEO and GEO combined right now, and it has been just as important for quite a few years now.

Things you can do include:

• Detailed Author Bios with qualifications, links to LinkedIn or similar.

• For ‘experience’, include original photos, videos, and first-person accounts. Instead of saying a product is good, show yourself using it.

• Build external validation - actively pursue digital PR to get your brand and experts mentioned on other authoritative sites. This is a powerful signal of ‘authoritativeness’.

8. Ensure Flawless Crawlability Through Technical SEO

The most brilliant content is useless if an AI crawler can't find, access, and read it correctly.

• Check your robots.txt: ensure you are not blocking key AI user agents like Google-Extended, GPTBot or ClaudeBot.

• Prioritise clean HTML: ensure your key content is rendered in the initial HTML load. Over-reliance on client-side JavaScript to load important text can make it invisible to many crawlers - crawlers currently can’t ‘see’ JavaScript rendered content.

• Use a logical internal linking structure: strong internal linking and information architecture helps crawlers understand the relationships between your pages and the overall structure of your site's expertise.

Make Great Stuff

I hope this article gave a good overview of what is still a new and raw subject, and I hope you learnt something if it’s new to you.

AI will continue to grow in popularity and power, and it’s difficult to imagine what the internet will look like in 10 years.

One thing I know for sure is that we should still focus on creating awesome stuff that people will actually want to interact with, that provides useful information and demonstrates our expertise. We’re at a time where we can use AI tools to improve experiences, or clog up the internet with low quality, easily replicated junk. I know which I’m doing.

—

And of course, if you need help with any of this I offer SEO Consultancy & Delivery as well as Digital Product Consultancy & Builds through my company Very Good.