How to Optimise for AI Search: A Practical UK Guide

To optimise for AI search, start by making sure AI systems can actually reach and read your pages. Then write clear, answer-first passages that are worth quoting, build proof of your expertise on other people’s websites as well as your own, and measure what that visibility brings in as enquiries and revenue. None of it is magic, and most of it is good SEO done properly. This guide takes each step in the order we’d tackle it for a UK business, with the evidence behind it and the bits we think are oversold.

Not sure whether ChatGPT, Gemini or Google’s AI Overviews mention your business at all? Castle’s AI SEO team can show you where you stand and what to fix first.

Explore our AI SEO services

TL;DR

• AI search optimisation builds on SEO. If your pages aren’t crawlable, indexable and readable in plain HTML, nothing else on this list matters.
• Lead each section with the direct answer, keep passages self-contained, and back every claim with a named source.
• AI assistants judge you partly by what other sites say about you, so reviews, press coverage and expert profiles count.
• Ecommerce brands should start with complete product data and honest buying guides. Lead-generation and local businesses should start with service and location pages that answer buyers’ questions.
• Track visibility, then traffic quality, then enquiries and revenue. Mentions on their own pay nobody’s wages.

Table of Contents

• What optimising for AI search actually means
• How AI systems choose what to cite
• Step 1: Make your site readable to AI crawlers
• Step 2: Write content AI can lift and trust
• Step 3: Build authority beyond your own site
• What to prioritise for ecommerce and lead generation
• How to measure AI search results
• What the evidence supports, and what it doesn’t
• Our perspective
• Get help optimising for AI search
• FAQ
• Sources

What optimising for AI search actually means

AI search means tools such as ChatGPT, Perplexity, Gemini, Copilot, and Google’s AI Overviews and AI Mode. They answer the question directly and usually cite a handful of sources. Optimising for them means improving your content and your wider reputation so that your business gets retrieved, cited and recommended inside those answers.

The work goes by a lot of names: GEO (generative engine optimisation), AEO (answer engine optimisation), LLMO, AI SEO. Semrush’s guide puts it simply: AI search optimisation is also known as GEO or SEO for AI search. The labels matter less than the job, which is the same under all of them.

Google’s own position is worth knowing, because it cuts through a lot of noise. Its Search Central team says that optimising for generative AI search is optimising for the search experience, and that SEO best practices continue to be relevant. The same page says you don’t need new machine-readable files, AI text files or special schema markup to appear in Google’s AI features. So if someone is selling you an llms.txt file as the key to AI Overviews, Google has already told you otherwise.

If you want the vocabulary in more depth, our AEO, GEO and AI SEO glossary and our explainer on what AI SEO is cover it. For how the disciplines compare, read SEO vs GEO.

How AI systems choose what to cite

Understanding the pipeline shows you where you can influence the outcome. The AI Search Handbook’s breakdown of retrieval describes the usual sequence. The system turns a question into one or more searches, retrieves candidate documents or passages, splits them into chunks, reranks the candidates, writes an answer from the evidence it picked and attaches source links. It also notes that a page can shape an answer without getting the visible citation, which is mildly infuriating but worth knowing when you read your reports.

Google confirms that its generative features use retrieval-augmented generation and query fan-out: the model fires off a set of related queries at the same time to gather more information. In practice you aren’t competing for one keyword. You’re competing across a cluster of related questions, and each page needs to answer its corner of that cluster well.

ChatGPT behaves differently from Google, and the numbers show by how much. A University of Toronto study, What Drives AI Citations?, found that ChatGPT splits a question into about four keyword-style queries and adds the current year to 23% of them. It also found that 97% of the pages ChatGPT retrieved were absent from Google’s top 50 results for the same queries. Ranking well in Google is a strong foundation, then, but it doesn’t guarantee a place in ChatGPT’s answers. Your content and your reputation need to work across several retrieval systems at once.

Step 1: Make your site readable to AI crawlers

No amount of clever writing helps if the systems can’t reach or parse your pages. This is the boring bit, and it’s where we find the most problems.

• Indexing and snippets. Google says a page must be indexed and eligible to be shown in Google Search with a snippet to appear in AI Overviews or AI Mode. Check for stray noindex tags and nosnippet restrictions, especially on templates a developer touched last.
• Crawler access. Duda’s technical guidance says to confirm that robots.txt and meta-robots tags allow known AI crawlers such as GPTBot, Claude-Web and CCBot into the parts of the site you want seen. It also flags misconfigured canonicals, long redirect chains and content locked behind logins or paywalls.
• Content in the HTML. Prismic reports that Vercel and MERJ analysed over 500 million GPTBot fetches and found no evidence of JavaScript execution at all. If your product details or service copy only appear after scripts run, many AI crawlers see an empty page. Serve the main content in the HTML the server sends.
• Firewall settings. Salesforce UK’s AI-readiness checklist advises configuring your web application firewall (Cloudflare, for example) so it doesn’t block AI crawlers by default. Plenty of sites have done this without anyone deciding to.
• Speed. HubSpot cites SE Ranking data showing that pages with a First Contentful Paint under 0.4 seconds averaged 6.7 ChatGPT citations, against 2.1 for pages slower than 1.13 seconds. That’s a correlation rather than proof, but fast pages convert human visitors better too, so the work pays either way.

A technical SEO audit finds most of these issues in one pass, and it’s always the first thing we look at on an AI visibility project. Our 40-point AI visibility audit checklist lets you run the basics yourself.

Step 2: Write content AI can lift and trust

AI systems pull passages, not whole pages. Your job is to make each passage clear, complete and worth quoting on its own.

Lead with the answer

Put the core point at the top of each section, then the detail, then an example. Search Engine Journal recommends clear H2 and H3 headings, one idea per paragraph, numbered lists for processes and tables for comparisons, adding that the more predictable your structure, the easier it is for AI tools to extract and summarise. Headlines matter too. Prismic cites a Growth Memo and AirOps analysis of around 815,000 query-page pairs which found that pages whose headlines directly answer the question were cited 41% of the time, against 29% for loosely related headlines.

Make every passage stand on its own

A chunk lifted out of context should still make sense. Name the subject rather than leaning on “it” or “this”, keep the figure next to its source, and put definitions beside the term they define. Think of it like a pub quiz answer sheet: if the marker can’t tell which question you were answering, you don’t get the point.

Offer a point of view

Google asks for unique, non-commodity content that readers will find helpful and reliable, organised with clear headings and supported by high-quality images and video. It also warns that mass-producing pages can breach its scaled content abuse policy. A page that repeats the same ten tips as everyone else gives an AI system no reason to pick yours. Original data, first-hand experience and a clear opinion do.

Keep structured data honest

Schema should describe what’s visible on the page, no more. HubSpot’s six-step workflow runs like this:

1. Research and map entities: group buyer questions and map how your brand, products and topics connect.
2. Draft answer-first, so each section resolves its question in the opening lines.
3. Add structured data that matches the visible page.
4. QA before publishing: crawlability, server-side rendering of the main content, schema validation.
5. Publish and record a baseline of where the page stands in answer engines.
6. Set a refresh cadence to update statistics, examples and claims.

The last step is the one people skip. Stale facts lose out to fresh ones, and a “2024 guide” in late 2026 tells every reader, human or otherwise, that nobody’s minding the shop.

Step 3: Build authority beyond your own site

AI systems don’t only read your website. They form a view of your business from everything published about it, which is why two equally well-written sites can get very different treatment.

• Topical depth. CGO Media’s research paper on content authority in AI search argues that authority develops when an organisation creates a coherent body of accurate, useful and differentiated information on its subject. One good article is a start. Twenty that join up are a reputation.
• Named experts. WordStream recommends author bios with relevant credentials and experience, Author schema linked back to your organisation page and a consistent author archive, so AI tools can connect your articles to the same expert.
• Third-party confirmation. Reviews, press coverage, industry directories and talks give AI systems repeated evidence that your business is what it says it is. WordStream quotes Tamara Omerovic: “AI doesn’t just look for what you publish; it looks for what others confirm about you.”
• Consistency everywhere. Search Engine Land’s guide to your brand in generative search recommends identifying the key entities in your niche, leaning into summaries, guides and FAQs, and keeping your text, images and videos consistent across websites, directories and social platforms. It also tells SEO and paid search teams to stop working in silos, which we’d frame and hang on the office wall if we could.

Our guide to earning LLM citations goes deeper on the off-site side.

What to prioritise for ecommerce and lead generation

The principles are the same for every business, but the first moves differ by model. These priorities draw on Ansvisor’s AI search playbooks by industry and Salesforce UK’s checklist.

Ecommerce: start here

• Put specifications, sizes, materials, price, stock status, delivery and returns in the page’s readable text, not only in images or tabs that load later.
• Keep product schema and your Merchant Center feed consistent with what the page shows.
• Publish buying guides for “best for” questions, such as wide feet, small kitchens or tight budgets.
• Write detailed reviews and product content that cover fit, use case and durability.
• Where AI platforms offer merchant programmes, submit your products directly. Salesforce UK recommends connecting to LLM merchant forms to improve visibility in AI shopping results.

Lead generation and local services: start here

• Keep your Google Business Profile and Bing Places listing complete, with identical facts on both.
• Build one page per location covering services, hours, service area and how to book.
• Answer the urgent questions directly on the page: emergency availability, typical prices, accreditations.
• Encourage reviews that describe the specific job and how it turned out.
• Add named authors, case evidence and third-party mentions, because trust decides who gets recommended.

Format matters too. In a 30-day visibility experiment Prismic ran on its own category, the brand appeared in 90.9% of answers to recommendation prompts and 98.2% of scenario prompts, but only 34.5% of how-to prompts. Comparison listicles took 51.27% of all non-homepage citations. It’s one company testing one category, so treat it as a pointer rather than a law. The pointer is useful though: honest comparison and scenario content for the decisions your customers actually make tends to get pulled into answers.

How to measure AI search results

Record a baseline before you change anything, then track three layers.

1. Visibility. Google’s Generative AI performance report in Search Console shows how your content performs in its generative AI features. For ChatGPT, Perplexity and the rest, run a fixed set of the questions your buyers ask on a regular schedule and record whether you’re mentioned, cited or recommended.
2. Traffic quality. AI referrals are still small, but they can be valuable. HubSpot reports that Microsoft Clarity’s study of more than 1,200 sites found AI referrals below 1% of visits, yet with a 1.66% sign-up click-through against 0.15% for search. It also cites WebFX’s analysis of 2.3 billion sessions, which found generative AI visitors converted roughly 1.2 times higher than organic search visitors.
3. Business outcomes. Tie it all to enquiries, sales and revenue in your analytics and CRM. This is the layer that tells you whether the work pays for itself.

Here’s an illustrative example of why the third layer matters. Say a kitchen fitter gets 4,000 organic visits a month and 120 from AI assistants. On traffic alone the AI channel looks like a rounding error. But if organic visitors enquire at 1.5% and AI visitors at 4%, that’s 60 enquiries against roughly 5. Close a quarter of those 5 at an average job of £9,000 and the AI channel is worth over £10,000 a month. The numbers are made up; the method isn’t. Our article on what ChatGPT SEO involves and how to measure it walks through the tracking setup.

What the evidence supports, and what it doesn’t

This field is young, and plenty of confident advice rests on very little. Two recent reviews are worth reading. A critical survey of 45 GEO studies by Olivier Martinez of Sciences Po found that topical relevance and context position are the most reproducible levers, and frames GEO as a multi-stage pipeline where an intervention that helps at one stage might hurt at another. The London Research Centre for Marketing & AI’s review of generative search visibility, which it describes as AI-assisted and not externally peer reviewed, concludes that the practical opportunity is to make useful facts discoverable, understandable and correctly attributed, then measure what happens after the answer is generated.

Put those together and the message is clear. Relevance is the dependable lever. Everything else works as part of a system, and pulling one lever on its own rarely shows up in the results. Anyone promising that one schema type or one heading trick will get you into ChatGPT is selling a switch that doesn’t exist.

Our perspective

The arrival of AI answers hasn’t changed the core job: help the right customers find a business they can trust. What’s changed is the number of places that discovery happens. A buyer might ask ChatGPT for a shortlist, skim a Gemini summary, then check Google and your reviews before they ever fill in a form. You need to be credible at every one of those stops.

That’s why we run SEO, GEO and AEO through one in-house team rather than treating AI search as a bolt-on. Technical health, expert content, digital PR and paid media all feed the same outcome, and splitting them across suppliers tends to leave gaps nobody owns. We’d also push back on the panic. Most businesses don’t need a new strategy for AI search; they need to finish the SEO basics they started, publish things only they can say, and get other people saying good things about them.

Our reporting ties activity to enquiries and revenue, so you can see what AI visibility is contributing and where to put the next pound.

Get help optimising for AI search

If you’d rather have a specialist team handle the audit, content and measurement, Castle’s AI SEO service combines traditional SEO with GEO and AEO to put your business in front of customers in search results and AI-generated answers. We work with ecommerce brands and lead-generation businesses, from technical audits through content, digital PR and reporting.

You can see the full range on our services page, read our complete guide to AI SEO and GEO, or learn how to choose an AI SEO agency before you speak to anyone.

Tell us which questions you want to be the answer to, and we’ll tell you how far off you are.

Talk to our AI search team

FAQ

Is optimising for AI search different from SEO?

It builds on SEO rather than replacing it. Google says SEO best practices remain relevant to its generative AI features. The extra work is making passages easy to quote, strengthening your reputation across other sites and measuring AI visibility. Our SEO vs GEO article explains the differences in detail.

How long does it take to appear in AI answers?

There’s no fixed timeline. It depends on your starting authority, your sector and how often each AI system refreshes its sources. Technical fixes can take effect once pages are recrawled, while authority built through content and third-party mentions accumulates over months. Set a baseline first so you can see progress.

Do I need to allow AI crawlers on my site?

If you want a system to retrieve and cite your pages, its crawler needs access. Check robots.txt, your firewall and whether your main content renders in plain HTML, as Salesforce UK’s checklist advises. Decide platform by platform according to your commercial goals.

Do I need an llms.txt file or special schema?

Not for Google. Its guidance says you don’t need new machine-readable files, AI text files or special schema.org markup to appear in its generative AI features. Accurate standard schema that matches the visible page is still worth having.

Can I measure traffic from ChatGPT and other AI tools?

Partly. Search Console reports on Google’s generative features, and your analytics can segment referrals from AI platforms. Pair these with a fixed set of test questions to track mentions, then connect everything to enquiries and revenue in your CRM.

Should ecommerce and service businesses optimise differently?

The foundations are shared, but the priorities differ. Ecommerce brands should focus on complete product data, matching schema and feeds, and buying guides. Service businesses should focus on location and service pages, consistent business listings, named experts and reviews that describe real jobs.

Sources

• Optimizing for generative AI features in Google Search, Google Search Central
• How to optimize for AI search results, Semrush
• How AI search retrieves and cites sources, The AI Search Handbook
• What Drives AI Citations? Content, Retrieval and Model Knowledge in GEO, Chen and Koudas, University of Toronto (arXiv)
• Five ways to optimize your clients’ sites for AI search, Duda
• How to optimize content for AI search engines, Prismic
• AI-readiness SEO checklist, Salesforce UK
• How to optimize your website for AI search, HubSpot
• Future-proofing WordPress SEO for AI search features, Search Engine Journal
• Content Authority in AI Search, CGO Media
• Ways to build brand authority for AI engines, WordStream
• Your brand in the age of generative search, Search Engine Land
• AI search strategy by industry, Ansvisor
• Optimizing Visibility in Generative Engines: A Critical Survey of GEO (2023-2026), Olivier Martinez (alphaXiv)
• What does the evidence establish about visibility in generative search? London Research Centre for Marketing & AI

Recommended

• The AI visibility audit: a 40-point checklist for UK businesses
• LLM Citations: What They Are and How to Earn Them
• What Is ChatGPT SEO? What It Involves and How to Measure It