How To: Optimise Your Website for ChatGPT, Gemini & AI Search

TL;DR:
There isn’t a separate magic formula for getting your website cited or recommended by ChatGPT, Gemini, Perplexity or other AI search tools. Start with the same foundations that make a site useful to people and understandable to search engines: make your important content crawlable, answer questions clearly, demonstrate genuine experience, build topical depth, use sensible internal linking and make it obvious who you are and why you’re credible. I’ve seen this work on my own site, including leads and links generated through ChatGPT, without creating content specifically to manipulate AI results.

Can you optimise a website for ChatGPT and Gemini?

Yes, but I don’t think optimising a website for ChatGPT, Gemini or other AI search experiences requires abandoning everything we already know about good SEO.

The way people discover information is changing. Instead of only typing a few keywords into Google, people are asking ChatGPT, Gemini, Perplexity and other AI tools detailed questions, asking for comparisons and requesting recommendations.

That creates a new visibility opportunity. You want your website and content to be something these systems can find, understand, trust and, where appropriate, cite or recommend.

But I haven’t seen convincing evidence that this requires an entirely separate type of website optimisation.

The fundamentals remain remarkably familiar: technically accessible pages, clear and useful content, strong topical relevance, sensible site structure, genuine expertise and evidence that supports what you’re saying.

The difference is that we’re now thinking beyond whether a page can simply rank for a keyword. We’re also thinking about whether its information is clear and useful enough to contribute to an answer generated from multiple sources.

I’ve seen this happen with my own website without deliberately creating pages “for AI”, which is what originally got me interested in the subject.

Case Study:  Getting Leads From ChatGPT

I track the leads I receive so I can understand where people are finding me.

In October 2025, an enquiry landed in my inbox where the person specifically mentioned that they’d found me through ChatGPT. This was the first time I’d seen such a clear example of ChatGPT directly contributing to a lead, so I asked whether they could share the conversation that led them to my website.

Fortunately, they could, and I was able to see exactly what ChatGPT had recommended.

https://chatgpt.com/share/68ee307c-3fc4-800d-8c9b-d5119013b8eb

What I found particularly interesting was that ChatGPT recommended my current website first, but also included my old website as another recommendation.

I’d recently migrated the old website to this one following a rebrand, so both domains appearing in the same response was an interesting insight into how historical information about a business can persist within AI search and recommendation systems.

I wouldn’t try to reverse-engineer exactly why ChatGPT selected me from this one example. What matters more is that the website was discoverable and relevant enough to be included in the recommendations, and that recommendation resulted in a genuine enquiry.

I hadn’t created a special page for ChatGPT, added AI-specific copy or deliberately optimised the site to appear in that response.

The pages ChatGPT found were the same service pages I’d already built around clear descriptions of what I do, relevant supporting content, genuine experience and the usual SEO fundamentals.

For me, that’s a much more useful starting point for AI optimisation than trying to find a shortcut specifically for ChatGPT.

Case Study:  Getting Links From ChatGPT

The second example appeared when I was carrying out one of my regular website performance audits.

I noticed traffic associated with ChatGPT and was able to trace it back to a shared conversation.

The person wasn’t searching for me or my services. They were trying to understand the srsltid parameter appearing in URLs, and ChatGPT used one of my articles as a source when answering the question.

https://chatgpt.com/share/68dce613-58a8-800c-93da-2f1692d6374b

The article it referenced was:

Solved: What Is The srsltid URL Parameter?

What’s useful about this example is why that particular article was suitable as a source.

It deals with a very specific problem, answers the question directly and is based on something I’d investigated myself. It doesn’t try to cover every possible aspect of URL parameters or ecommerce tracking. It satisfies one particular search intent properly.

And that’s relevant when thinking about how to optimise content for ChatGPT, Gemini and other AI search experiences.

You can’t guarantee that an AI system will cite a particular page. But you can make the information on that page clear, specific, useful and easy to understand, increasing its usefulness as a potential source.

This is also why I don’t think AI optimisation means creating hundreds of pages specifically for every conceivable AI prompt.

The better objective is to create content that genuinely deserves to be used when somebody asks a relevant question.

What actually helps your website appear in AI answers?

There isn’t a single optimisation that makes a website appear in ChatGPT, Gemini, Perplexity or other AI-generated answers.

Different platforms discover, retrieve and select information in different ways, and you can’t guarantee that a particular page will be cited or recommended.

What you can do is make your website a better potential source.

From my own experience, I’d focus on the following.

Make your important content accessible

If a page can’t be properly crawled or accessed, you’re immediately limiting its potential visibility. The same technical foundations that matter for SEO still matter here: accessible HTML, sensible internal linking, indexable pages and a site structure that makes important content easy to discover.

Answer the actual question clearly

Don’t make somebody dig through 1,000 words before reaching the answer.

If you’re explaining how something works, define it. If you’re solving a problem, explain the solution. If you’re comparing options, make the differences clear.

My srsltid article is a good example. It answers a very specific question directly, which also makes the information useful when an AI system needs a source for that question.

Optimise the content, not just the website

AI visibility isn’t only a technical website problem. The content itself needs to be useful enough to contribute to an answer.

Keep each page focused on a clear subject and intent, answer important questions directly, use descriptive headings, explain concepts in plain language and support claims with examples or evidence where possible.

That doesn’t mean writing specifically for ChatGPT or Gemini. It means making the information easier for both people and machines to understand and use.

Add something that isn’t available everywhere else

This is where first-hand experience becomes particularly valuable.

Use your own examples, tests, screenshots, case studies, observations and professional experience where they’re relevant.

There isn’t much value in publishing another generic summary assembled from the same information already available on hundreds of other websites.

Build depth around the subjects you actually know

One good article can be useful, but it shouldn’t necessarily exist in isolation.

I’ve been deliberately building SEO content hubs that connect useful supporting articles with the service and skills pages they relate to.

That gives users somewhere useful to go next and creates a much clearer body of information around the subjects I work with.

Make it obvious who is behind the content

Useful information is stronger when its source and context are clear.

Author information, genuine experience, case studies, testimonials, business information and supporting pages all help establish who created the content and why they might know what they’re talking about.

Use structured data where it genuinely describes the page

Schema can help machines understand information about a page, person, business, product or other entity, but I wouldn’t treat it as an AI visibility hack.

Use appropriate structured data because it accurately describes what’s already there, not because somebody has promised that adding a particular schema type will make ChatGPT recommend you.

Keep the site internally connected

Relevant internal links provide context and help connect related information.

For example, this article discusses the broader question of AI visibility. My separate guide to LLMs.txt deals specifically with that proposed standard, while the content hubs article goes much deeper into building interconnected topical coverage.

Those are distinct subjects, so they don’t all need to be crammed into this page.

The same principle applies when thinking about different AI platforms.

I wouldn’t currently create separate optimisation strategies for ChatGPT, Gemini, Perplexity and Claude unless you have evidence that a particular platform requires something genuinely different. Their underlying systems and methods of retrieving information differ, but the practical website objective remains broadly the same: publish accessible, relevant and credible information that is useful enough to be surfaced as part of an answer.

Ultimately, I don’t think the objective should be to make your content look as though it was written for an AI system.

The objective is to publish information that is accessible, clear, relevant, credible and genuinely worth using as a source.

What’s the difference between AI SEO, AEO and GEO?

You’ll see several different terms used to describe optimising content for AI-led search, particularly AI SEO, AEO and GEO.

They overlap considerably, but they’re not quite the same thing.

AI SEO is probably the broadest and most understandable term. It describes applying SEO principles with AI-powered search experiences in mind, including traditional search engines incorporating AI-generated results and standalone AI tools.

AEO, or Answer Engine Optimisation, is more specifically about making content useful for systems that provide direct answers to questions. That can include featured snippets, AI Overviews, voice search-style results and other answer-led experiences.

GEO, or Generative Engine Optimisation, is generally used to describe improving visibility within generative AI responses, including systems such as ChatGPT, Gemini and Perplexity.

There are theoretical differences between the three, but I wouldn’t get too hung up on the terminology.

From a website owner’s perspective, the objective is largely the same: make your content discoverable, understandable, useful and credible enough to be surfaced when somebody asks a relevant question.

That’s why I don’t find the argument about whether GEO is “just SEO” particularly useful.

People aren’t only typing keywords into Google anymore. They’re asking longer questions, requesting comparisons, looking for recommendations and asking AI tools to summarise subjects for them.

So yes, the search landscape has changed enough that discussing AI optimisation makes sense.

But that doesn’t mean the underlying work has suddenly become completely different.

An example from my own website

My Google Analytics consultancy page is a useful example of how my approach has evolved.

The main part of the page is still built around its primary purpose: explaining the service, how I work, what I can help with and whether I’m the right person for someone who needs GA4 support.

It has the usual optimisation foundations:

• An optimised title tag, H1 and meta description
  
• Clear service-focused content
  
• Genuine testimonials and reviews
  
• Relevant structured data
  
• Strong site speed and Core Web Vitals
  
• Internal links from relevant pages
  
• Clear information about what I do and how I can help

But I’ve also become much more deliberate about covering the related questions and fan-out queries that naturally sit around the primary intent.

At the bottom of the page I now have two supporting areas.

The first is a set of FAQs based on genuinely relevant questions around things such as fixing an existing GA4 setup, implementation, access requirements, costs, timescales and ongoing support.

I’ve put these into accordions so they’re available when somebody needs them without overwhelming the main service content.

The second is a Google Analytics guides and insights hub, connecting the service page with relevant articles covering GA4 setup, debugging, data quality and tracking examples.

I’m not trying to turn the service page into a 10,000-word answer to every possible question someone could ask about Google Analytics.

The main page satisfies the commercial service intent.

The FAQs answer closely related questions that belong on the same page.

The supporting guides handle subjects that deserve much more depth.

For me, that’s a much better way to think about optimising for both traditional search and AI-led discovery.

Cover the useful fan-out questions, but don’t force every fan-out query onto one page.

Some belong as supporting sections. Some belong as FAQs. And some have enough depth or a sufficiently different intent to deserve their own page.

That gives users useful answers while also building a much clearer body of interconnected information around the subject.

Build topical depth around your subject

One page doesn’t need to answer every question somebody could possibly ask about a subject.

In fact, trying to force everything onto one page can make it less useful by mixing different search intents together.

A better approach is to build a connected group of pages around the subject.

I use SEO content hubs to connect my main service and skills pages with relevant supporting guides and articles.

For example, my Google Analytics consultancy page handles the commercial intent around getting help with GA4. Closely related questions can be answered on the page itself, while subjects that need more explanation are covered by separate guides and linked back into the wider topic.

This creates a clearer relationship between:

• The main service or subject page
  
• Closely related fan-out questions
  
• More detailed supporting guides
  
• Relevant case studies and practical examples

I think this becomes particularly useful when considering AI search.

Someone might begin with a broad question and then ask increasingly specific follow-up questions. Those questions don’t necessarily need individual pages, but your website should ideally contain useful information that can satisfy the different directions the subject naturally expands into.

The aim isn’t to manufacture a page for every possible ChatGPT prompt.

It’s to develop genuine topical depth, while putting each piece of information in the place that makes most sense for the user.

That’s also why internal linking matters. Related pages shouldn’t exist as isolated articles. Connecting them helps establish how the different parts of the subject relate to each other and gives users a logical route into more detailed information.

I’ve gone into the practical side of this separately in my guide to building SEO content hubs, rather than duplicating all of it here.

What should you not do?

This is where AI optimisation can quickly turn into chasing shortcuts.

You don’t need to rebuild your website around whatever happens to be the latest theory about how ChatGPT, Gemini or other AI systems select their sources.

In particular:

• Don’t create a page for every possible AI prompt or fan-out query.
  Cover closely related questions where they genuinely improve the page, and create separate supporting content where the intent or subject deserves more depth.
  
• Don’t add FAQs simply because you think AI prefers question-and-answer content.
  Add them when they answer useful questions that belong with the primary page intent.
  
• Don’t rewrite everything into an artificial “AI-friendly” format.
  Clear headings, focused sections and direct answers are useful because they help people understand the content too.
  
• Don’t treat schema as a shortcut to AI visibility.
  Use structured data where it accurately describes the content and entities already present on the page.
  
• Don’t manufacture expertise or citations.
  First-hand experience, real examples, genuine reviews and useful external references are far more defensible than creating signals purely to influence AI systems.
  
• Don’t assume you need an LLMs.txt file.
  I’ve written a separate guide to LLMs.txt. It’s an interesting development, but it isn’t a magic switch that makes your website appear in AI answers.

The problem I see with a lot of AI optimisation advice is that a genuine change in how people search gets reduced to another checklist of supposed hacks.

Create this file. Add this markup. Structure every paragraph like this. Create a page for every prompt. Mention particular entities. Get mentioned on particular websites.

That’s not much of a strategy.

I’d rather focus on making the website technically accessible, genuinely useful, well connected and demonstrably credible, then use the evidence we get from search and AI visibility to refine it.

The objective isn’t to manipulate an AI system into mentioning you.

It’s to make your website a source worth mentioning.

Why trust and first-hand experience matter for AI visibility

If an AI system is going to use your content as a source, cite your website or recommend your business, it makes sense that the credibility of that information matters.

That doesn’t mean you need to pretend to be a huge brand. It means making it clear who is behind the content, what experience they have and why the information is worth trusting.

For me, that means things such as:

• Clear author information
  
• Genuine first-hand experience
  
• Real examples and case studies
  
• Testimonials and reviews
  
• Clear business and contact information
  
• Consistent information about the services I provide
  
• Supporting content demonstrating experience across the subject

This is another reason I think non-commodity content is increasingly important.

Anyone can publish a generic explanation of a subject, and generative AI itself can produce one in seconds.

What’s much harder to replicate is your experience of actually doing the work.

The two ChatGPT examples earlier in this article are a good illustration. I’m not theorising about whether AI tools might find my website. I’ve been able to see instances where ChatGPT has referenced my content and where somebody subsequently contacted me after I was recommended.

The same principle applies across different types of websites.

For service businesses, useful trust signals can include testimonials, author information, case studies, client examples and clear service descriptions.

For ecommerce websites, that might include detailed product information, genuine reviews, delivery and returns information, original imagery and useful buying information.

For local businesses, it can include consistent business information, reviews, opening hours, service areas, location information and appropriate LocalBusiness structured data.

None of those things guarantees inclusion in an AI-generated answer.

But if you’re trying to make your website something that people, search engines and AI systems can understand and potentially rely upon, making your experience and credibility visible is a sensible part of the job.

Same foundations, new search landscape

The way people find information is changing.

They’re still using Google, but they’re also asking ChatGPT, Gemini, Perplexity and other AI tools to answer questions, compare options, research subjects and recommend businesses.

That creates another place where your website and content can potentially be discovered.

But I don’t think it requires throwing away everything we’ve learned about good SEO and starting again.

The websites best placed to benefit are still likely to be those that make their information easy to discover, understand and trust.

That means:

• Build technically accessible pages
  
• Answer the questions that genuinely matter
  
• Demonstrate real experience rather than producing generic content
  
• Cover relevant fan-out queries without creating pointless pages for every possible prompt
  
• Build topical depth through useful supporting content
  
• Connect related pages logically
  
• Make it clear who you are and why somebody should trust what you’re saying

And never mind the bollocks like FAQ padding for the sake of it.

The two examples in this article happened without me deliberately trying to optimise my website for ChatGPT. The content existed because it was useful for the people I originally wrote it for.

That’s still the approach I’d take now, albeit with a better understanding of how search behaviour is evolving and how the questions surrounding a topic can be covered across a website.

I don’t see AI optimisation as a replacement for SEO.

I see it as another layer of visibility built on many of the same foundations.

Search is changing. The language is changing. The way people discover businesses is changing.

But websites still need to be useful, accessible, trustworthy and clear.

That has always been the job. AI just makes the weak spots harder to ignore.

---