Skip to content
SEOUpdated 18 min read

AI Search Optimisation: A 2026 UK Playbook for SMBs

A practical 2026 UK playbook for AI search optimisation: how SMBs win citations in Google AI Overviews, ChatGPT and Perplexity without the GEO acronym hype.

Aidan van VuurenHead of Digital, Peak Digital
Contents11 sections01What AI search optimisation actually means in 202602Six moves UK SMBs should make this quarter03What Google's official guidance actually says04Let the AI crawlers in before you optimise anything05How to write pages that AI search can quote06Schema markup: worth having, not a lever07Off-site signals: earn mentions, do not manufacture them08How to measure AI search performance without burning budget09What not to do (the snake oil checklist)10A 90-day AI search optimisation plan for a UK SMB11Frequently asked questions

AI search optimisation is now part of ordinary search work for UK businesses. Google AI Overviews, AI Mode, ChatGPT search, Perplexity and Claude all decide every day whether your business gets cited or quietly skipped, and the rules for getting picked up are simpler, and less mysterious, than most of what is being sold. This is the practical playbook for UK SMBs who want citations without buying into the acronym circus, built on what the platforms themselves publish.

What AI search optimisation actually means in 2026

AI search optimisation is the work of getting your pages retrieved, cited and recommended inside AI-generated answers: Google AI Overviews and AI Mode, ChatGPT search, Perplexity, Claude and the other LLM-powered search tools. The unit of success is different from classic SEO. Classic SEO chases the click. AI search optimisation chases the citation, the brand mention and the answer slot, knowing that the click may never come.

You will see the work sold under several names: AIO (AI optimisation), AEO (answer engine optimisation), GEO (generative engine optimisation), LLMO (large language model optimisation), plain "AI SEO" and, in American spelling, AI search optimization. Google's own view is blunt. Its guide to optimising for generative AI features on Google Search says that from Google Search's perspective, optimising for generative AI search is optimising for the search experience, "and thus still SEO", and it sends anyone weighing third-party AEO or GEO services to its guidance on evaluating SEO advice. There is no separate algorithm to rank inside AI Overviews, and no secret schema or hidden file that unlocks AI visibility.

For a UK SMB, the practical scope is narrower than the noise suggests. You are not trying to dominate every AI surface. You are trying to win citations on the questions your buyers ask, in the AI features that show up for those questions, on the platforms your customers actually use. That is a manageable list. If you would rather have that list built and worked for you than run it in-house, that is what our answer engine optimisation service does, and the generative engine optimisation side covers the ChatGPT and Perplexity surfaces specifically. If you are still deciding whether any of this is distinct from search work you already pay for, our comparison of AEO and SEO answers that first.

Six moves UK SMBs should make this quarter

1. Get indexed and ranking in ordinary search first

Google says its generative AI features are rooted in its core Search ranking and quality systems, and they retrieve pages from the Search index before writing an answer. To be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to appear with a snippet. If your pages are not indexed or rank nowhere, fix that before anything labelled "AI".

2. Publish what only you can say

Google's guide separates commodity content (common knowledge anyone could write) from non-commodity content built on first-hand experience. A summary of what is already online gives an AI system nothing new to cite. Your own prices, data, case numbers, photographs and opinions do.

3. Let the search crawlers in

Each AI company runs separate bots for search, user requests and model training, and a robots.txt file or CDN setting written for one can block another. Check that the bots that put you in answers can reach you, then decide about training on its own merits.

4. Answer the question near the top

Google says you do not need to chunk content or rewrite it specially for AI. Clear headings with the answer stated early still help readers, and they give every engine that quotes passages a clean sentence to lift. Cite your sources and give real numbers: in the original GEO research, those changes lifted visibility the most.

5. Earn genuine mentions, do not manufacture them

AI answers draw on what is said about you across the web: reviews, forums, videos and press. Google warns that seeking inauthentic mentions is not as helpful as it looks, because its spam systems sit underneath its AI features. Earn coverage, reviews and recommendations the slow way.

6. Measure AI visibility with first-party data

Search Console now has a Generative AI performance report showing impressions from AI Overviews and AI Mode. Add GA4 referral data from the AI assistants and a monthly manual prompt check, and you have a free measurement stack before you spend anything on tools.

There is a seventh move, and it is a "do less" instruction: stop spending on services that promise guaranteed AI rankings, special AI files or "GEO-only" audits. Google's guidance on third-party SEO tools and advice says no tool has access to its ranking data and that AEO and GEO claims deserve the same scrutiny as any other SEO promise. There is real research on what earns AI citations (we cover it below), but it points at better content and sources, not at a separate product. Audit the deliverable, not the acronym.

What Google's official guidance actually says

Two Google documents should anchor every UK SMB's AI search optimisation work. The first is the May 2025 Search Central post "Top ways to ensure your content performs well in Google's AI experiences on Search". The second is the longer guide "Optimizing your website for generative AI features on Google Search", last updated in July 2026, which is now the canonical reference.

The May 2025 post sets out the basics under these headings:

  • Focus on unique, valuable content for people.
  • Provide a great page experience on every device.
  • Ensure Google can access your content: Googlebot not blocked, the page returning HTTP 200, and indexable content on it.
  • Manage visibility with preview controls such as nosnippet, data-nosnippet, max-snippet and noindex.
  • Make sure structured data matches the visible content.
  • Go beyond text with good images and video.
  • Understand the full value of your visits, not just the click count.

The 2026 guide explains the mechanics. AI Overviews and AI Mode use retrieval-augmented generation (grounding): they retrieve relevant pages from the Search index and support the answer with links to them. They also use query fan-out, where the model runs a set of related searches behind the scenes. Google's example is a search for "how to fix a lawn that's full of weeds" fanning out to queries about herbicides, chemical-free weeding and prevention. That is why a page that answers the questions around a topic, not just the headline query, gets more chances to be retrieved.

The guide also lists what you can ignore for Google Search:

  • llms.txt and other "special" files or markup. Google Search does not use them. Creating one will neither harm nor help your visibility.
  • "Chunking" content into tiny pieces. Google says its systems can understand several topics on one page and there is no ideal page length.
  • Rewriting content just for AI. AI systems understand synonyms, so you do not need every long-tail variation of a phrase.
  • Seeking inauthentic mentions. Its core ranking systems favour quality and its spam systems block the rest, and the AI features depend on both.
  • Overfocusing on structured data. It is not required for generative AI search and there is no special schema to add, though it still helps with rich results.

One warning in the guide deserves underlining: publishing a separate page for every fan-out variation, primarily to manipulate rankings or AI responses, breaches Google's scaled content abuse spam policy. We covered the guide's launch, with what it means for UK marketers, in Google's New Guide to Optimising for AI Search. This playbook is the operational follow-up.

Let the AI crawlers in before you optimise anything

The most common reason a UK SMB is invisible in ChatGPT or Perplexity has nothing to do with content quality: the site is blocking the bots that would put it in an answer. This is the first thing to check and the cheapest thing to fix.

Each provider documents its own bots, and the important detail is that search, user-requested fetches and model training are controlled separately:

Which AI bot does what, from each provider's own documentation (November 2026)

ProviderSearch and citationsFetches for a user's questionModel training
GoogleGooglebot (AI Overviews and AI Mode use the Search index)Answers are grounded in the Search indexGoogle-Extended robots.txt token (Gemini training and grounding outside Search)
OpenAIOAI-SearchBotChatGPT-UserGPTBot
AnthropicClaude-SearchBotClaude-UserClaudeBot
PerplexityPerplexityBotPerplexity-UserNeither bot is used for training

Sources: Google's common crawlers list, OpenAI's bots documentation, Anthropic's crawler help article and Perplexity's crawler documentation.

Three things follow from that table:

  • Blocking GPTBot does not take you out of ChatGPT search. OpenAI says each setting is independent: you can allow OAI-SearchBot to appear in search results while disallowing GPTBot for training. Blocking OAI-SearchBot is what removes you.
  • Google-Extended does not control AI Overviews. Google says it does not affect inclusion in Google Search and is not a ranking signal. To limit what Google shows from your pages in AI Overviews and AI Mode, you use the same preview controls as ordinary search (nosnippet, data-nosnippet, max-snippet or noindex). Google also says it intends to support a shared AI-training preference token in robots.txt from early 2027.
  • Your CDN can block bots your robots.txt welcomes. Cloudflare's documentation says that from 15 September 2026, new domains block bots classified as Training or Agent on pages that display ads by default, while Search bots stay allowed. Cloudflare defines Agent as bots acting in real time on a person's behalf, such as chat fetch bots, which describes ChatGPT-User and Claude-User. Check the dashboard, not just the file.

When the door is open, the bots do come. On our own site, across the 30 days from 4 September to 4 October 2026, OpenAI's bots made 4,372 requests, Googlebot 2,084, Anthropic's 1,128 and PerplexityBot 633. Not one of them fetched our llms.txt file, which is the subject of our write-up on llms.txt and 30 days of server logs.

The second access problem is rendering. Google documents how it crawls, renders and indexes JavaScript in its JavaScript SEO basics. OpenAI, Anthropic and Perplexity publish nothing about rendering, so treat the raw HTML as your baseline for them. If your key content only appears after a click (accordion panels, tabbed specifications, "read more" toggles that load on interaction) or is injected client-side, do not assume it will be read. Fetch your own page with curl and read what comes back; if the answer is not in there, no amount of schema will rescue it. Our technical SEO checklist covers the rest of the crawl and render basics.

How to write pages that AI search can quote

Google says you do not need to write in a special way for its AI features, and that is the right starting point: write for the person asking. But the habits that make a page easy for a person to scan are the same ones that make a passage easy to quote, and there is evidence about what earns citations elsewhere.

The most cited study is "GEO: Generative Engine Optimization", a paper led by researchers at IIT Delhi and Princeton and accepted to the KDD 2024 conference. Testing nine rewriting methods on a benchmark of queries, the authors found that adding citations to sources, quotations and statistics improved a page's visibility in generated answers by around 30 to 40 per cent on their main measure, while keyword stuffing performed worse than leaving the page alone. Lower-ranked pages gained the most. It is a lab benchmark, not a live ranking study, but it points the same way as Google's advice: substance beats tricks.

Habits worth building into every important page:

  • Definition sentences: "X is Y." Open any section that defines a concept with a plain statement.
  • Numbered lists for processes: step one, step two, step three.
  • Tables for data and comparisons: prices, specifications, options side by side.
  • Short paragraphs: two to four sentences, one idea each.
  • Question headings where people ask questions: answer in the first sentence underneath, then expand.
  • Named entities over pronouns: spell out the brand, product or place rather than leaning on "it" or "they", so a lifted sentence still makes sense on its own.
  • Sources and numbers: link the evidence and state the figure, with the date it applies to.

Avoid the opposite habits: long anecdotal openings, vague language ("various factors"), conclusions buried in the final paragraph and headings that hide the topic. Our breakdown of the content types that win in AI search goes further on formats.

Schema markup: worth having, not a lever

Structured data helps search engines understand what a page is, and Google still uses it for rich results. What it is not is an AI switch: Google's guide says structured data is not required for generative AI search and there is no special schema.org markup to add. Its May 2025 post adds one firm rule, that structured data must match the content visible on the page. If WordPress is your CMS, our WordPress SEO guide covers the plugin setup end to end.

A sensible baseline for a UK SMB, usually handled by your CMS or SEO plugin:

  • Organization (or LocalBusiness for a local firm): legal name, logo, address, contact details and links to your official profiles. Set once at site level.
  • Article or BlogPosting with an author: headline, author as a Person with a profile page, publish and update dates.
  • BreadcrumbList: how the page sits in your site.
  • Product and Offer: ecommerce only, with price and availability that match the page.

Validate it in Google's Rich Results Test, then stop. Five types installed correctly beat twenty rendered with errors.

One correction worth making loudly, because a lot of published advice is now out of date: FAQPage no longer earns a rich result in Google Search. Google stopped showing the FAQ rich result on 7 May 2026 and removed the supporting documentation in June 2026; the old FAQPage documentation URL now redirects to the Search Central updates page. If an agency is still selling FAQ schema as a route to rich snippets, it has not read the changelog. That does not mean you should strip FAQ sections off your pages. A clear question with a direct answer is useful to readers and easy to quote. What changed is the payoff: you are writing FAQs for people and answer engines, not for a SERP widget.

Off-site signals: earn mentions, do not manufacture them

Google's guide notes that its AI features can show what is being said about products and services across the web, "including in blogs, videos, and forum discussions". In the same breath it says chasing inauthentic mentions is not as helpful as it seems. For a UK SMB that means the off-site work is real, and that the shortcuts are not.

A practical UK off-site list:

  • Google Business Profile and review platforms: an active profile, genuine reviews and replies to them. Never pay or reward anyone for a Google review.
  • LinkedIn company page and directory listings: the same name, address, founding date and description everywhere.
  • UK trade press: commentary in the publications your buyers read. One genuine quote a quarter is worth more than ten guest posts on low-quality blogs.
  • Forums and communities: real participation where your customers ask questions, not seeded "recommendations".
  • Video: Google's guide calls out images and video as extra ways to appear, and a clear explainer gives AI systems a named expert and a transcript.
  • A consistent brand entity: the name spelt the same way everywhere, one canonical logo, one address.

None of this is new. It is ordinary digital PR and housekeeping, and it compounds across SEO, AI search and conversion. Our guides to getting your brand mentioned in ChatGPT and Perplexity and to generative engine optimisation and LLM visibility go deeper on the brand-mention side.

How to measure AI search performance without burning budget

The free starting stack covers most of what a UK SMB needs in the first six months.

  • Search Console's Generative AI performance report: Google's help page says it was rolled out to all websites worldwide by 31 August 2026. It shows impressions for links to your site in AI Overviews and AI Mode, by page, country and device. If you cannot see it, your site has not yet had enough AI feature impressions. Our news piece on the Search Console generative AI reports covers the launch.
  • GA4 referral data: filter for chatgpt.com, perplexity.ai, claude.ai, copilot.microsoft.com and gemini.google.com. The traffic is small, and a good share arrives as direct rather than referral. Our guide to tracking AI search traffic in GA4 sets it up step by step.
  • A monthly manual prompt check: run your ten most commercial questions through Google AI Mode, ChatGPT search and Perplexity. Record whether you are cited, what the citation says and who appears next to you. Answers vary from run to run, so compare months, not single checks.
  • Brand alerts: a Google Alert on your brand name catches new third-party mentions.

Paid tools can automate AI citation tracking and share-of-voice reports across hundreds of prompts. They are useful at scale, but Google's own guide says no third-party tool has access to its internal ranking or AI systems, so judge them on what they actually observe. For most UK SMBs that is a year-two purchase, not a launch-day one.

What not to do (the snake oil checklist)

The AI search space is full of paid playbooks that will not move the needle. Patterns worth refusing:

  • Paying for an llms.txt file as a Google lever. Google Search ignores it, and in our own logs no AI company's bot requested it in a month.
  • Chopping pages into fragments "for AI". Google says chunking is not required and there is no ideal page length.
  • Mass-producing pages for every fan-out query. Doing it to manipulate rankings or AI answers is scaled content abuse under Google's spam policies.
  • Buying mentions, reviews or "AI placements". Inauthentic mentions are exactly what Google's guide tells you to ignore, and incentivised Google reviews break its rules outright.
  • Buying FAQ schema as a rich-result play. That result no longer exists in Google Search.
  • Hidden text. Text and links hidden from visitors are named in Google's spam policies, whether the target is classic search or AI extraction.
  • Guaranteed AI rankings. Nobody controls what an AI system cites, and Google says no tool can see its ranking systems.

Most of the rest of the work is what any honest SEO programme would tell you to do anyway: write better than the competition, get cited by reputable sources, and make sure the crawlers can reach your site. If you want a structured way to check your own site against all of this, our answer engine optimisation audit checklist turns it into a list.

A 90-day AI search optimisation plan for a UK SMB

If you are starting from a standing position and want to optimise for AI search without specialist tooling, this sequence is enough to move the needle.

  1. Days 1 to 15: check access first. Read robots.txt against the bot table above, check your CDN's AI bot settings, confirm your key pages are indexed in Search Console, and curl your top pages to confirm the substance is in the initial HTML.
  2. Days 16 to 30: rework your 20 most important commercial pages so each heading is answered early, claims carry a source or a number, and structured data matches what is on the page.
  3. Days 31 to 45: off-site sweep. Update your Google Business Profile, LinkedIn page, review profiles and trade directories so names, addresses and descriptions match.
  4. Days 46 to 60: publish two pieces with genuine first-party substance: your own data, prices, process or results. These are the pages AI systems have a reason to cite.
  5. Days 61 to 75: set your baseline. Run your ten most commercial questions through AI Mode, ChatGPT search and Perplexity, save the answers, and note your figures in the Generative AI performance report.
  6. Days 76 to 90: review. Which pages gained AI feature impressions? Which questions now cite you, and who sits next to you? Use the answers to plan the next quarter.

This is the work. There is no shortcut, no secret file and no certified GEO badge that changes the outcome. There is only a disciplined 90-day cycle, repeated.

If you want a partner to run it for your site, our team handles SEO services for UK businesses, including the technical audits, schema rebuilds and content programmes that put pages into AI citation range. For broader context on whether outside SEO support is the right call at all, our guide to measuring SEO ROI walks through the honest answer, and UK SMBs starting from scratch will get value from our SEO for small business UK owner's guide.

Frequently asked questions

Is AI search optimisation different from SEO?

Mostly no. Google says that from its perspective, optimising for generative AI search is optimising for the search experience, and so still SEO: technical health, useful content and a credible reputation. The parts that need extra attention are AI crawler access (each AI company runs its own bots), passages that answer a question clearly, genuine off-site mentions and measuring citations as well as clicks. Be wary of anyone selling it as a separate discipline with its own guarantees.

How does AI search optimisation actually work?

Google describes two techniques. Retrieval-augmented generation means the AI feature retrieves relevant pages from the Search index and supports its answer with links to them. Query fan-out means the model runs a set of related searches behind the scenes to fill in the answer. Other AI search tools work in a similar way with their own crawlers. So the work runs in that order: make sure the right crawlers can reach your content, publish pages that answer the question and the questions around it, and earn a reputation that makes you a credible source.

Will AI Overviews kill my organic traffic?

Probably reduce, not kill, but the effect is real. Pew Research Center tracked the browsing of 900 US adults in March 2025: on searches with an AI summary, people clicked a traditional result in 8% of visits, against 15% without one, and clicked a link inside the summary in just 1%. No equivalent UK study has been published. The sensible response is to win the citation, strengthen brand searches and put more effort into the commercial pages people still click.

Is FAQ schema still worth adding in 2026?

For AI extraction, yes. For rich results, no. Google retired the FAQ rich result on 7 May 2026 and removed the documentation the following month, so FAQPage markup will not put a dropdown under your listing any more. The question-and-answer content itself is still among the most citable formats in AI Overviews, so keep writing FAQs, just stop treating the schema as a SERP feature play.

How long does it take to appear in AI Overviews?

Google publishes no timeline. A page becomes eligible once it is indexed and can be shown with a snippet, and from then on it competes like any other page in Search. Watch the Generative AI performance report in Search Console to see when your pages start earning AI Overview and AI Mode impressions, and treat anyone promising a fixed timeframe with caution.

Do I need to optimise separately for ChatGPT and Google AI Overviews?

The underlying work overlaps heavily: useful content, clear structure, credible sources and genuine mentions carry across. The differences are mostly access and measurement. Google's AI features use the Search index, so ordinary indexing applies, while ChatGPT search depends on OpenAI's OAI-SearchBot being allowed to crawl you. Check each platform's bot separately, and run your monthly prompt check on each platform, because the answers and citations differ.

Is SEO dead in 2026 now that AI search has arrived?

No. Google says its generative AI features are rooted in its core Search ranking and quality systems, so the work that wins organic positions is the same work that earns AI citations. The discipline is widening to include crawler access and citation tracking, but demand for trustworthy, well-structured content is rising, not falling. Our is SEO dead reality check goes through the evidence.

Do I need an llms.txt file for AI search?

Not for Google. Its guide to generative AI features says Google Search does not use llms.txt or similar files, and creating one will neither help nor harm your visibility. In our own server logs, no bot belonging to an AI company requested the file in 30 days, while those companies' bots made more than 11,500 requests to the rest of the site. The full write-up is in our post on llms.txt.

Does blocking GPTBot remove my site from ChatGPT search?

No. OpenAI documents GPTBot as its training crawler and OAI-SearchBot as the bot that surfaces sites in ChatGPT search, and says each setting is independent. You can allow OAI-SearchBot and disallow GPTBot. Blocking OAI-SearchBot is what keeps you out of ChatGPT search answers.

Filed underAI Search
Written byAidan van Vuuren

Head of Digital at Peak Digital. More than 12 years delivering results for clients across web, SEO and paid media.

Want help with your own marketing?

Get a free audit, we will tell you what is working, what is not, and what to do about it.

Get a free audit