AI Search Optimisation: The Complete UK Guide for 2026

Some of your customers have already changed how they find a business like yours, and you probably haven't noticed. They ask ChatGPT for a recommendation. They read the AI summary at the top of Google and never touch the blue links underneath. They ask Copilot inside Word, or Gemini on the walk home. Every one of those moments ends with an AI system deciding which businesses to name, and most of the time it names almost nobody.

I've written this for owners and marketing managers, not for search specialists. If a term needs explaining, I explain it, and I don't make a claim without a source you can check. Read to the end and you'll know what the work actually involves, what the evidence says pays off, what is being oversold, and what to do first.

What is AI search optimisation?

AI search optimisation is the work of making your business easy for AI tools to find, understand and recommend, across Google's AI Overviews, ChatGPT, Perplexity, Microsoft Copilot and Google Gemini. The aim is blunt and practical: when someone asks an AI a question your business can answer, your business is the one it names, with the details right.

You'll hear it called AI SEO, or answer engine optimisation (AEO), or generative engine optimisation (GEO). The acronyms change; the job doesn't, and it leans heavily on ordinary search engine optimisation. If the alphabet soup grates on you, our short explainer on the difference between SEO, AEO and GEO untangles it. Here, I'll just say AI search optimisation and get on with it.

The thing that sets it apart from the search marketing you already know is the shape of the output. A classic Google results page hands you ten links and lets you pick. An AI answer picks for you. It might name two or three businesses. It might name none. There is no page two to scroll to. Being the fourth-best plumber in town used to be worth a trickle of clicks; in an AI answer it is worth nothing at all. So the work bends away from chasing rankings and towards a harder target: being the business the machine can describe with confidence and quote without hesitating.

Let me be clear about one thing up front, because it gets lost in the hype. Traditional search optimisation still matters, and everything below builds on it rather than throwing it out. Google has said in plain terms that ordinary, useful, crawlable content is exactly what feeds its AI features. This is an extension of doing the basics well, not an escape hatch from them.

How big is the AI search shift in the UK?

AI search has gone mainstream in the UK, and the gap between how people search and how businesses publish has become a chasm. Roughly one in three UK Google searches triggers an AI Overview (SearchScore, 2026), yet fewer than 5% of UK SMEs have taken any action on AI visibility (SearchScore, 2026). British customers changed their habits years faster than British businesses changed their websites.

The customer half of that is well documented. Around 47% of UK adults have used AI-powered search in 2026 (SearchScore, 2026), whether that means an AI Overview sitting above the Google results or a question typed straight into ChatGPT. The business half has barely moved: audits of more than 31,000 UK websites found that over seven in ten score under 50 out of 100 for AI visibility (SearchScore and Rank4AI, 2026).

All of this cuts two ways, and the two pull hard in opposite directions.

Start with the damage. When an AI Overview shows up on a results page, the top-ranking site takes on average 58% fewer clicks than it would without one (Ahrefs, 2026). And it isn't only first place that bleeds: in the same study, position two lost roughly half its clicks, and even position ten lost nearly a fifth. If your Google traffic has quietly slid since 2025, this is the usual suspect. We've written a separate piece on why AI Overviews are cutting UK website traffic and exactly what to check in your own analytics.

Now the upside, because there genuinely is one. The visitors who do come through from AI tools are worth far more than their numbers suggest. In Seer Interactive's case data, people referred by ChatGPT converted at around 16%, against roughly 1.8% for Google organic visitors (Seer Interactive, 2026). It isn't magic. By the time someone clicks a link in an AI answer, the AI has already vouched for you, so they land on your page half-sold.

And the door is wide open. An analysis of over 350,000 business locations found that ChatGPT recommends only 1.2% of local businesses in any given category (SOCi Local Visibility Index, 2026). Read one way, that's a gut-punch: if you're invisible, you're in the 98.8%. Read the other way, it's the opportunity of the decade, because you can act while fewer than 5% of your competitors have bothered.

How AI engines decide what to recommend

An AI builds its answer in three moves. It retrieves candidate pages from search indexes and training data. It selects the sources that look trustworthy, consistent and easy to quote. Then it names the businesses those sources describe. So to be recommended you need three things: clear pages, facts that agree with each other across the web, and mentions on sites the AI already trusts. Everything else is detail.

It's worth a minute under the bonnet, because once you see how the machine works, the order of the checklist later on stops looking arbitrary and starts looking obvious.

Retrieval: the AI goes looking

Ask a modern AI tool a question and, most of the time, it does not answer from memory alone. ChatGPT with search, Perplexity, Copilot and Google's AI features all fire off live searches and pull in current web pages. Two consequences fall straight out of that. Your content can move an answer within weeks rather than years, which is fast by search standards. And if the AI crawlers can't reach your site in the first place, you can be left out of the running entirely, however good your actual business is.

Selection: the AI picks its sources

Out of everything it retrieved, the system reaches for what it can use without second-guessing: pages that answer the question near the top, state facts flatly, line up with other sources, and sit on a site with a bit of earned authority behind it. Ambiguity gets punished. If your website says one trading name, your Google listing says another, and some directory says a third, you've handed the machine a fact-checking headache, and the path of least resistance is to leave you out of the answer altogether.

Citation: the AI names businesses

This is the bit most owners miss. When an AI actually recommends businesses, it tends to lean on third-party pages, comparison articles, directories, review platforms, local press, rather than on the businesses' own websites. And the numbers back that up: SOCi found only a 45% overlap between businesses that rank well in Google's local map results and businesses that AI tools recommend (SOCi Local Visibility Index, 2026). Ranking well and being recommended are two different achievements. Your presence in the sources the AI quotes is a separate asset, and you have to build it separately.

Picture it. Two electricians in Sheffield. The first has a gorgeous website, but her listings don't match and nobody else on the web mentions her. The second has a plainer site that says exactly what she does and where, the same details everywhere you look, 80 detailed Google reviews, and a name-check in a local news piece about reliable tradespeople. Ask an AI for an electrician in Sheffield and the second one has handed the machine a pile of corroborated evidence while the first has handed it a shrug. That, in a sentence, is the whole game.

The complete AI search optimisation checklist

Seven jobs, and the order matters. Make your business identity unambiguous, publish answer-shaped content, add sensible schema markup, check that AI crawlers can reach your site, earn citations in the sources AI trusts, strengthen your reviews and directory listings, then monitor what the AI actually says about you. Foundations first, promotion second, measurement running underneath the whole time.

1. Make your business identity unambiguous

An AI stitches together a picture of your business from every mention of it online, and your job is to make sure those mentions tell the same story. Use the exact same business name, address, phone number and one-line description on your website, your Google Business Profile, Companies House, your social profiles and every directory you're listed in. Write an about page that says, in plain sentences, what you do, where you work, who you serve and since when. Your homepage should say what the business actually is inside the first paragraph, not three screens down past the slogans. Here's the test: if a stranger couldn't describe your business accurately after 30 seconds on your site, neither can a machine.

2. Publish answer-shaped content and FAQs

AI engines quote content that answers the question, and they do it fast. So shape your key pages that way: a heading that poses a real customer question, then the first 40 to 60 words underneath answering it in full, with the depth and the evidence following on. One topic per page. Be specific exactly where your competitors go vague: prices or price ranges, areas covered, timescales, what's included and what isn't. Then build an FAQ page out of the questions customers genuinely ask you on the phone and by email, each one answered honestly in a paragraph. You'll have noticed this article does the same thing, opening every section with a straight answer. That isn't an accident. It's the single highest-value writing habit there is for AI search.

3. Add schema markup, with honest expectations

Schema markup is code that labels the facts on your pages, your business type, opening hours, services, FAQs, so a machine can read them without guessing. Add it. It's cheap, it powers ordinary Google features, and it strips out ambiguity. But be honest with yourself about what it won't do. When Ahrefs tracked 1,885 pages that added schema, citations in AI answers barely moved, and in Google's AI Overviews they actually dipped slightly (Ahrefs, 2026). Pages the AI cites do tend to use schema, yes, but that's because well-run sites do a dozen things right at once, and schema is just one of them. Treat it as hygiene, like brushing your teeth, not as a lever that lifts you up the answer.

4. Check AI crawlers can reach your site

The AI tools send their own crawlers, with names like GPTBot (OpenAI), ClaudeBot (Anthropic), PerplexityBot and Google-Extended. Plenty of websites block them and the owner never finds out, because the block is buried in a robots.txt file a previous developer wrote, or a firewall, or a hosting or CDN provider that turns AI bots away by default. When the crawlers can't read your site, the AI shrugs and falls back on whatever third parties happen to say about you, which is often thin, out of date, or plain wrong. Checking is a five-minute job: open yourdomain.co.uk/robots.txt and look for those bot names. Our guide to finding out whether your website is blocking AI crawlers walks you through it line by line.

5. Earn citations in sources AI already trusts

This is the slowest job on the list, and for competitive queries it's also the most powerful. AI recommendations lean hard on third-party pages: "best of" roundups, comparison articles, trade association listings, local press, the established industry sites. So ask an AI tool about your sector and your town, write down every page it cites, and then work out, one by one, how to get yourself onto them. Sometimes that means pitching a local journalist. Sometimes it means joining your trade body and actually finishing its directory profile. Sometimes it's just emailing the person who wrote a roundup that left you out. We go through the tactics properly in our guide to getting your business recommended by ChatGPT in the UK.

6. Strengthen reviews and directory listings

For local and service businesses, review platforms are among the first places an AI engine looks. Keep your Google Business Profile complete and current, and gather reviews at a steady drip rather than in one panicked burst. A detailed review that names the service, the location and the result gives an AI something it can actually quote; a bare five stars gives it almost nothing. After that, cover the directories that carry weight in your trade, Trustpilot, Checkatrade or TrustATrader if you're in the trades, and whatever the equivalents are for you. Remember that ChatGPT recommends only 1.2% of local business locations (SOCi Local Visibility Index, 2026), so doing this thoroughly puts you ahead of very nearly everyone.

7. Monitor what AI says about you

You can't manage what you never look at. Once a month, put the same fixed questions to ChatGPT, Gemini, Copilot and Perplexity: who do they recommend for your service in your area, what do they say about your business by name, and are your details right. Write the answers down. Mistakes are everywhere, wrong opening hours, an address you moved out of three years ago, a competitor's work described as yours, and they nearly always trace back to one stale source you can go and correct. It costs nothing but a bit of time, and it's the feedback loop that tells you whether everything above is working.

What not to waste money on

Two things deserve a hard side-eye in 2026: llms.txt files and schema-only packages. Google has confirmed that llms.txt does not affect rankings or AI Overviews (Google Search Central, June 2026). And when Ahrefs tracked 1,885 pages that added schema markup, AI citations barely moved (Ahrefs, 2026). Neither is worth paying for on its own.

The llms.txt one is worth spelling out, because it's being sold hard right now. The pitch is a special file that summarises your site for AI systems. Then, on 15 June 2026, Google updated its official guidance to say you do not need such files to show up in Google Search, generative AI features included, because Google simply doesn't use them (Google Search Central, June 2026, via Search Engine Land). No major AI platform has said the file sways its recommendations either. Making one costs little and harms nothing, so fine, make one if you like. But if an agency puts llms.txt at the top of its deliverables, that tells you plenty about what's underneath it.

Schema-only packages fall to the same logic, for the reason I gave in step three: the best evidence we have says schema on its own does not buy you citations (Ahrefs, 2026). Add it because you're doing your site properly, not because someone sold it to you as a miracle at a premium.

And more broadly, be sceptical of anyone who guarantees you a spot in AI answers. Those answers shift by user, by phrasing, by location, by model version, and the underlying systems change month to month. Nobody controls what an AI will say. We don't, and neither does anyone charging you to promise otherwise. The honest version of this work is about improving your odds and correcting the record, and the honest providers show you evidence rather than promises.

How to measure progress

Track progress three ways. Put a fixed set of questions about your market to the main AI assistants every month and record what comes back. Watch AI referral traffic in your analytics. And keep an eye on branded searches and direct enquiries. Above all, expect it to move slowly: AI answers shift over months, not days, so read the trend, not any single check.

That monthly question set is your real scoreboard. Keep the wording identical every time, run each question more than once because the answers wobble, and note three things: whether you got mentioned, what was said about you, and which sources the tool leaned on. The citations list usually starts moving before the mentions do, so watch it as your early warning.

Inside Google Analytics, traffic from AI tools turns up as referrals from domains like chatgpt.com, perplexity.ai, copilot.microsoft.com and gemini.google.com. The volumes will look tiny next to Google organic, and that's fine, because you judge these on quality, not quantity. In Seer Interactive's case data these visitors converted at roughly nine times the rate of organic ones (Seer Interactive, 2026), which means twenty AI referrals can quietly outperform a few hundred ordinary clicks.

Last, keep half an eye on searches for your own business name in Google Search Console, and just ask new customers how they found you. People act on an AI recommendation by googling your name far more often than by clicking the link, so a chunk of your AI visibility will never show up as a clean referral. It arrives instead as branded search and "someone mentioned you" enquiries, and it counts every bit as much.

Should you do it yourself or hire help?

Most UK small businesses can handle the foundations on their own: a clear website, contact details that match wherever they appear, honest FAQ pages and a steady flow of reviews. Where paying for help starts to make sense is when you want a full audit of what the AI currently says about you, the technical fixes done properly, or someone grinding away at the citation work while you get on with running the place.

A sensible DIY route looks like this. Fix your listings and your about page this week. Check robots.txt this month. Rewrite your top five pages in answer-shaped form this quarter. And start the monthly monitoring habit today, not "soon". None of that needs an agency, and doing it will already put you ahead of the large majority of UK businesses that have done nothing at all (SearchScore, 2026).

Where help genuinely earns its fee is in the parts that are technical, tedious, or built on relationships: auditing every answer the major AI tools give about you and your rivals, diagnosing crawler and markup faults, and the patient slog of getting you into the roundups, directories and press the engines cite. That's exactly how we've built our own audit and implementation service: a fixed-price audit that shows you the exact answers AI gives about your business today, then flat-monthly implementation on the website you already have, if you'd rather the fixing was done for you. No platform migration, no long contract, no guaranteed placements, because, as I keep saying, nobody can honestly guarantee those.

Whoever you end up hiring, put three questions to them. What exactly is going to change on my website? Where, specifically, will my business be mentioned? And how will we know whether it worked? A vague answer to any of those is the single most reliable warning sign in this whole market.

Questions people ask

These are the questions UK business owners bring us most often about AI search, with the short answers here and the longer versions scattered through the guide above. If yours isn't on the list, ask us directly and you'll get a straight answer, not a sales pitch.

Is AI search optimisation different from normal SEO?

They overlap heavily. Most of what helps you rank in Google, such as clear content, sound structure, trusted mentions and reviews, also helps AI engines recommend you. The difference is emphasis: AI search rewards direct answers, consistent business facts and third-party citations more than traditional keyword targeting. Treat it as an extension of your existing search work, not a replacement for it.

How long does AI search optimisation take to work?

Simple fixes, such as unblocking AI crawlers or correcting wrong details, can influence answers within weeks, because tools like ChatGPT search and AI Overviews read live web pages. Deeper work, such as earning citations in trusted sources, usually takes three to six months to change what AI engines say. Nobody can promise a specific timescale, and you should doubt anyone who does.

Can anyone guarantee my business will be recommended by ChatGPT?

No. AI answers vary by user, wording, location and model version, and the systems change frequently, so guaranteed placement in ChatGPT or AI Overviews is not something anyone controls. What you can do is stack the odds in your favour: unambiguous business information, answer-shaped content, open crawler access, strong reviews and citations in the sources the AI already draws on.

How much does AI search optimisation cost in the UK?

Doing the foundations yourself costs time rather than money. Paid help ranges from one-off fixed-price audits to monthly retainers, and prices vary widely across the UK. Whatever you pay, ask exactly what will change on your website and where your business will be mentioned. Be wary of anyone selling llms.txt files or schema markup alone, as the evidence does not support either as a standalone fix.

Sources

• SearchScore: UK AI Search Visibility Report, Q2 2026
• SearchScore: AI search statistics UK 2026
• Rank4AI: AI Search Statistics UK 2026, original data
• Ahrefs: AI Overviews Reduce Clicks by 58% (February 2026)
• Ahrefs: We Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved (May 2026)
• SOCi: Local Visibility Index 2026
• Seer Interactive: How Traffic from ChatGPT Converts, case study
• Search Engine Land: Google says llms.txt files won't harm or help your search rankings (June 2026)