Answer Engine Optimisation (AEO): How UK Businesses Get Found in AI Search (2026)

Answer engine optimisation (AEO) is how you get your business named and cited inside AI answers, on ChatGPT, Claude, Perplexity and Google’s AI Overviews, instead of hoping someone scrolls down to your blue link. It is search optimisation for a world where the machine reads the results and hands back one answer, rather than ten links for a person to sift through.

We learned why this matters the embarrassing way. We ran an AI-visibility audit on our own site, thesecondfloor.io, testing 48 questions a prospective client might genuinely ask an AI assistant about agencies like ours. We appeared in zero of them. Zero out of 48. Invisible in our own category, on the exact topic we sell. That result is why we built the playbook on this page, and why we now run the same audit for a client before we promise them anything.

This is the complete, no-fluff version: what AEO is, why it matters in 2026, how AI engines actually choose who to cite, the seven-part playbook we run, how local businesses fit in, how you measure it, and how long it takes. All killer, no filler.

What is answer engine optimisation (AEO)?

AEO is the practice of structuring your content and online presence so AI answer engines retrieve, trust and cite you when they generate an answer. The target is not a ranking position. It is a mention inside the answer itself, ideally with your name attached and a link back to you.

An answer engine is any system that responds to a question with a synthesised answer rather than a list of links. That now includes ChatGPT, Claude, Perplexity, Google’s AI Overviews and AI Mode, Gemini, and the AI-powered answers baked into Bing and other tools. They do not just rank pages. They read across many sources, pull out the facts and phrases they judge most trustworthy, stitch them into a single reply, and (usually) cite a handful of the sources they leaned on.

That changes the unit of optimisation. Classic SEO optimises a page to rank. AEO optimises a claim to be quoted. The neat way to hold it in your head: Google ranks pages, AI cites statements. If a model cannot find a clean, checkable sentence on your site that answers the question, it will happily quote someone who wrote one, even if your page is objectively better.

So AEO works at the level of individual answers and individual facts. Can a machine find the specific claim it needs? Can it confirm that claim from other independent sources? Can it attribute it cleanly to you? Get those three things right, repeatedly, across the questions your customers ask, and you start showing up in the answer. If the acronym soup is already bothering you, our explainer on what GEO is unpacks the wider family of terms in plain English.

AEO, GEO and SEO: same job, different labels

AEO and GEO describe the same core job under two names, and SEO is the older discipline they both build on. Do not let the vocabulary slow you down. The work is more alike than the labels suggest.

Here is the honest breakdown:

• SEO (search engine optimisation) is getting your pages to rank in traditional search results. Still essential, because AI engines mostly draw from the same open web they always have.
• GEO (generative engine optimisation) is the umbrella term for being visible across all generative AI engines: the recommendations, the summaries, the cited sources.
• AEO (answer engine optimisation) focuses on the answer-retrieval layer specifically: being the source a model selects when it needs a fact, a definition or a recommendation.

In practice, the tactics overlap almost completely. Clean structure, clear entity data, credible third-party proof and genuinely useful content help you across all three. We tend to use AEO and GEO interchangeably in client work, because splitting hairs over the acronym is a distraction from the job. If you want the full side-by-side, we wrote a dedicated piece on GEO vs SEO vs AEO that maps exactly where they overlap and where they diverge.

The useful takeaway is this: AEO is not a replacement for SEO, it is a layer on top. The businesses winning in AI answers are largely the ones already doing solid SEO, then adding the specific moves that make their content easy for a machine to lift, verify and attribute.

Why does AEO matter now?

Because a fast-growing share of searches now end inside an AI answer, not on your website. If you are not in the answer, you do not exist for that query, no matter how good your page is.

Look at where search actually sits in 2026. AI Overviews now appear on roughly 48% of Google searches, up sharply over the past year, according to analysis drawing on BrightEdge data. Around 68% of Google searches now end without a click to any website, per a 2026 zero-click study reported by Search Engine Land. SparkToro’s read of the same shift puts it bluntly: less than a third of Google searches still send a click to the open web. Gartner projected traditional search volume would fall 25% by 2026, and hundreds of millions of people now open ChatGPT, Claude or Perplexity as their first move rather than a search box.

Put those together and a simple truth falls out: the answer box is the new front page. The prime real estate is no longer position one in a list of links. It is the two or three sources a model chooses to cite, and the business it names when someone asks for a recommendation.

This matters for three reasons that hit revenue:

1. Zero-click means zero visibility if you are not cited. When the answer satisfies the question on the spot, the only businesses that get noticed are the ones inside the answer. Everyone else is invisible for that search.
2. Citations compound. Brands cited in AI answers pick up more clicks and more trust, and being named once makes a model more likely to name you again. Absence compounds too.
3. Your competitors are mostly asleep on this. Most businesses have done nothing deliberate about AEO. That is a problem if it is you, and an opportunity if it is your rivals.

Google’s own AI surface is the biggest single battleground, so it is worth understanding on its own terms. We go deep on that in our guide to ranking in Google AI Overviews.

How do AI answer engines decide what to cite and recommend?

They recommend what they can retrieve, verify and attribute. Everything in the playbook below is really just a way of making those three things easy for a machine. Get all three right and you become quotable. Miss one and you tend to fall out of the answer.

Break the decision into its three moving parts:

• Retrieve. The engine has to be able to reach and read your content in the first place. If AI crawlers are blocked, if the useful facts are buried in an image or a script, or if your page is a wall of vague marketing prose, there is nothing clean to pull. No retrieval, no citation.
• Verify. Models are tuned to avoid getting things wrong, so they favour claims they can corroborate from several independent sources. A fact that appears only on your own website is weaker than the same fact echoed across reviews, directories, press and other reputable pages. Corroboration is the currency.
• Attribute. The engine wants a clean, self-contained statement it can point back to. A crisp sentence that answers the question outright is far easier to quote than the same idea spread across three rambling paragraphs.

The research backs this up from several angles. The foundational academic work, the GEO study from researchers at Princeton and IIT Delhi, tested nine content tactics and found that adding citations, quotations and statistics from credible sources lifted a page’s visibility in generative answers by up to 40%. Freshness is decisive too: eMarketer’s summary of the research shows recently updated pages are far likelier to be cited, and analyses of commercial queries have found roughly 83% of AI citations go to pages refreshed within the past twelve months.

Then there is proof from other people, which turns out to matter enormously. A Seer Interactive study of more than 800,000 AI responses found businesses actively gathering and responding to reviews were cited in around 75% of relevant AI answers, against roughly 1% for those with no active review presence. AirOps’ analysis of off-site signals reached a similar place, finding third-party sources drive the large majority of brand discovery in AI search, not your own marketing copy.

So the businesses AI recommends are not the ones with the slickest homepage. They are the ones a machine can reach, cross-check against the wider web, and quote in a single clean line. That is the whole game, and the playbook below is how you win it.

The AEO playbook: seven moves that get you cited

AEO comes down to seven repeatable moves: fix your entity data, write answer-shaped content, add structured data, give AI plain-text proof, earn reviews and third-party validation, get authoritative coverage, and open the door to AI crawlers. None of them is exotic. Done together and kept up, they are what shifts a business from invisible to cited. This is the exact spine we run in our growth engagements, one move at a time.

1. Nail entity consistency and your NAP

An entity is the thing a model understands you to be, and it recommends entities it can pin down with confidence. Your first job is to make your business unmistakable and consistent everywhere it appears.

Start with your NAP: name, address and phone number. These must be byte-for-byte identical across your website, Google Business Profile, Companies House record, social profiles and every directory you sit in. “The Second Floor Ltd” in one place and “Second Floor Agency” in another reads, to a machine, like two different companies with weaker evidence for each. Inconsistency dilutes the very corroboration that gets you cited.

Then reinforce the entity itself:

• Use one canonical business name and stick to it.
• Keep an About page that states plainly who you are, what you do, where you operate and who you serve.
• Connect your profiles to each other (site to social, site to Google Business Profile, and so on) so the model can join the dots.
• Make sure the basic facts (founding, location, services, people) agree across every source.

Entity clarity is unglamorous and it is the foundation everything else sits on. If a model is not sure who you are, it will not risk recommending you.

2. Write answer-shaped content

Answer-shaped content leads with a direct, self-contained answer, then expands, so a machine can lift the answer without needing the rest. This single habit does more for AEO than almost anything else on the page.

The pattern is simple. For every question your customers ask, give the page a heading that mirrors the question, then open with one or two sentences that answer it outright, in plain language, before you elaborate. That opening line is what gets quoted. If your answer only emerges after four paragraphs of throat-clearing, the model will grab a cleaner version from somewhere else.

Practical rules we hold to:

• Front-load the answer. First sentence answers the question. Everything after is supporting depth.
• Use question-shaped headings. “How much does X cost?” beats “Pricing” because it matches how people actually ask.
• Chunk it. Short sections, clear subheads, bullet lists and numbered steps. Retrievable beats pretty.
• Be specific. Numbers, named methods, concrete examples and dates give a model something checkable to cite. Vague copy gives it nothing.
• One idea per chunk. Self-contained sections travel better into an answer than sprawling ones.

This is why FAQ sections earn their keep. A tight question-and-answer block is basically pre-packaged AEO fuel, each pair a quotable unit ready to be lifted.

3. Add schema and structured data

Structured data is code that tells a machine exactly what each part of your page means, removing the guesswork before an engine even reads the words. It is one of the most reliable ways to make content easy to retrieve and understand.

Schema markup (from Schema.org, added as JSON-LD) labels the entities and facts on your page: your organisation, your location, your products and prices, your FAQs, your articles and authors, your reviews and ratings. The types that pull the most weight for AEO are:

• Organization and LocalBusiness for your core entity and NAP.
• FAQPage for question-and-answer blocks.
• Article with clear author and date for editorial content.
• Product, Service and Review where relevant.
• BreadcrumbList to make site structure explicit.

Structured data does not force an engine to cite you, but it removes ambiguity, and ambiguity is what stops machines quoting you. It also feeds the freshness signal: a visible, accurate “last updated” date helps, given how strongly recency correlates with citation. If your markup is patchy or broken, fix it as part of a proper technical SEO audit, which is where we usually catch these issues. We go deeper in schema and structured data for AI search.

4. Give AI plain-text proof

AI engines read text, not marketing gloss, so the facts that matter must exist as clean, extractable words on the page. If a key detail lives only inside an image, a video, a PDF or a piece of interactive JavaScript, assume the model cannot see it.

This one trips up beautifully designed sites constantly. Your prices sit in a graphic. Your opening hours are baked into an image. Your best proof point is a number on a slide with no text equivalent. To a person it looks great. To a machine it is a blank space.

The fix is to make every important fact available as real, selectable text:

• Put prices, specifications, hours, service areas and key claims in the page copy, not just in graphics.
• Add genuine text alternatives for anything visual that carries information.
• State your credentials, results and differentiators in plain sentences a model can quote.
• Where you cite a statistic or a source, write it out. Quotations and cited figures are among the strongest citation triggers there are.

Think of it as writing for a very literal reader who cannot see, only read. Give that reader the facts in words and you give the engine something to lift.

5. Earn reviews and third-party validation

Reviews and independent validation are the single biggest off-site lever in AEO, because models trust what others say about you far more than what you say about yourself. This is where a lot of the outcome is actually decided.

The numbers are hard to ignore. The Seer Interactive study of more than 800,000 AI responses found review-active businesses cited in around 75% of relevant answers versus about 1% for those with no review presence. Reviews give a model exactly what it craves: independent, corroborating, frequently refreshed evidence that you exist, that you are legitimate, and that you are any good.

What to actually do:

• Build a steady flow of Google reviews. Ask every happy customer, make it easy, keep the flow constant rather than a one-off burst. Fresh reviews double as a freshness signal.
• Respond to reviews, the good and the bad, because engagement is itself a trust cue.
• Spread across platforms that matter for your sector: Google, Trustpilot, industry-specific sites, sector directories.
• Encourage specific, detailed reviews. A review that names the service and the outcome gives a model quotable substance.

Reviews pull double duty, feeding both local search and AI answers at once. We dug into that overlap in our piece on whether Google reviews affect local rankings. Short version: they affect a lot more than rankings now.

6. Get authoritative coverage and UK directory listings

Being written about, listed and referenced on reputable third-party sites gives AI engines the independent corroboration they need to recommend you with confidence. One tidy website is rarely enough on its own. Our guide to the UK directories that influence AI recommendations names the ones worth your time.

If reviews are the loudest third-party signal, everything else you appear on is the supporting chorus. The more reputable, independent places confirm who you are and what you do, the stronger your entity and the safer a model feels quoting you. That means:

• Solid UK directory listings. Get listed accurately in the directories that matter for your sector and region, with consistent NAP throughout. Quality and relevance over sheer volume.
• Genuine coverage and mentions. Trade press, local news, partner sites, guest articles, podcasts and industry roundups all build the web of references a model reads.
• Wikipedia and knowledge sources, where you legitimately qualify, because these feed how entities are understood.
• Consistent facts everywhere, so every new source reinforces the same story rather than muddying it.

You are building a web of independent references that all point back to the same clear, consistent business. That web is what a model checks before it puts your name in an answer. It also happens to be genuine reputation-building, which is why it holds up as the ground shifts.

7. Open the door to AI crawlers

If AI crawlers cannot access your site, none of the rest matters, so check that your robots.txt actually lets them in. This is the most common own goal we find, and among the fastest to fix.

AI engines use their own crawlers, separate from the traditional Googlebot. The ones that matter most in 2026 include GPTBot, OAI-SearchBot and ChatGPT-User (OpenAI), ClaudeBot and Claude-SearchBot (Anthropic), PerplexityBot (Perplexity), Google-Extended (Google’s AI), and Applebot-Extended (Apple). Plenty of sites block some or all of these without realising, sometimes via a plugin default, sometimes because a blanket rule got copied in years ago.

A sensible approach:

1. Read your current robots.txt. Find out which of these agents you are allowing or blocking today. Most people are surprised.
2. Decide deliberately. There is a real distinction between crawlers that gather training data and those that fetch pages to answer a live query. You can, if you want, allow the search-and-answer agents (like OAI-SearchBot, Claude-SearchBot and PerplexityBot) while making a separate call on the training crawlers. What you should not do is block the answer engines by accident.
3. Confirm nothing else blocks them. Firewalls, bot-protection services and CDN rules can quietly turn AI crawlers away even when robots.txt looks fine.
4. Check rendering. Make sure your important content appears in the raw HTML, not only after JavaScript runs, so a crawler sees it.

We treat crawler access as step zero of any AEO project. Businesses that have opened their sites to the main AI crawlers have reported meaningful jumps in AI-attributed traffic within about 90 days, which lines up with the timelines we see. It is a small change with an outsized effect, and it belongs in the same sweep as your broader technical SEO audit.

How does AEO work for local businesses?

Local AEO is about showing up when someone asks an AI assistant for a business near them, and right now almost nobody does. The gap between traditional local search and AI recommendation is enormous, which makes this both the harshest and the most winnable arena.

The reality check is stark. SOCi’s 2026 Local Visibility Index, which looked at more than 350,000 locations across 2,751 brands, found ChatGPT surfaced just 1.2% of local business locations. Perplexity managed 7.4% and Gemini 11%, while Google’s local three-pack showed a given location 35.9% of the time. So when someone asks an assistant “who’s the best plumber in Leicester?”, the overwhelming majority of genuinely good local plumbers simply are not in the running. They are invisible.

That invisibility is expensive, because demand is moving fast. BrightLocal’s 2026 Local Consumer Review Survey reported that 45% of consumers now use AI tools to find local services, up from 6% a year earlier. The customers are already asking. Most businesses just are not among the possible answers.

The good news: because so few local businesses have done anything deliberate, the moves that work are not yet crowded. To become recommendable locally:

• Get your Google Business Profile complete and pristine. Correct categories, services, hours, photos, attributes, and NAP that matches your site exactly. This remains foundational, and we walk through it fully in our guide to local SEO and Google Business Profile in the UK.
• Feed the local entity signals. Precise location details, the areas you serve, nearby landmarks and neighbourhood context, all as plain text on relevant pages.
• Build local reviews relentlessly. For local AEO, reviews are close to everything. Volume, recency, responses and specifics all count.
• Get into local and sector directories with consistent details, so the wider web corroborates where you are and what you do.
• Publish location-specific content that answers real local questions, so there is something concrete for an assistant to retrieve and cite.

The full local playbook is in local AEO. If you want the focused version for the biggest assistant, we wrote a whole spoke on how to get recommended by ChatGPT in the UK. For a lot of local businesses, closing this gap is the highest-return marketing work available right now, precisely because the field is so empty.

How do you measure AI visibility?

You measure AI visibility with a tracked panel of real category questions, run monthly across ChatGPT, Claude, Perplexity and Google AI Overviews, recording whether you appear, where, and which sources get cited. Without a fixed method you are guessing, and AI answers vary too much to judge by the odd spot check.

That variance is the whole reason for a disciplined approach. Ask the same question twice and you can get different answers, different sources, even a different recommendation, depending on the model, the phrasing, the session and the day. One lucky mention proves nothing, and one unlucky miss proves nothing either. A fixed question set tested repeatedly is what turns noise into a trend you can act on.

Here is the method we use:

1. Build the question panel. Assemble 30 to 60 questions a real customer would ask an assistant in your category: problems, comparisons, “best X in [town]”, buying questions, and the specific things your ideal client types. This set becomes your benchmark.
2. Run it across the main engines. Put every question to ChatGPT, Claude, Perplexity and Google AI Overviews (add Gemini and others as needed) on a set cadence, monthly at minimum.
3. Record what matters. For each question: do you appear at all, how prominently, is your link cited, and what is the tone of the mention? Note who does get cited when you do not, because that is your competitor and source list in one.
4. Track the trend. Watch your presence rate climb (or not) month on month as the playbook takes effect. The direction over time is the real signal, not any single answer.
5. Confirm it in your analytics. Watch for referral traffic from AI assistants landing on your site. Proper measurement setup makes this visible, which is one more reason to get your GA4 setup right.

Two honest caveats. There are no fixed rankings in AI answers, so measurement is about presence and share of voice over time, not a number one spot. And attribution is still maturing, because a lot of AI influence shows up as someone later searching your brand directly rather than as a clean click from the assistant. Measure the panel and the referrals together and you get a fair read on whether AEO is working.

How long does AEO take to work?

Most businesses see AI visibility start to shift within two to three months of fixing crawler access, entity data and reviews. It is quicker than traditional SEO for the foundational wins, and like SEO it keeps compounding after that.

The timeline splits into fast movers and slow burners:

• Weeks, not months (the fast wins). Unblocking AI crawlers, fixing NAP and entity inconsistencies, adding schema, and reshaping key pages into answer-first format. These are read quickly, and they are often the difference between total invisibility and starting to appear.
• Two to three months (the shift). As fresh reviews accumulate, content gets restructured, and engines re-crawl and re-index the changes, your presence across the answer panel typically begins to move. This is where most businesses first see themselves named.
• Beyond three months (the compounding). Authoritative coverage, deeper content, and a steady review flow build slowly and then keep paying off. Entity strength and third-party corroboration are the parts that compound hardest over time.

Two things to keep honest. First, nobody can guarantee a placement or a ranking in AI answers, and anyone who does is selling you something. The systems are probabilistic and they change. Second, this is maintenance, not a project with an end date. Freshness matters so much that content left to rot loses citations, so AEO is an ongoing discipline rather than a one-off fix. That is exactly why we work in scoped monthly engagements with weekly updates and no long-term lock-in: it keeps the work moving and it keeps us honest about the results.

What AEO mistakes should you avoid?

The most common AEO mistakes are all versions of the same thing: making yourself hard to retrieve, verify or attribute. Avoid these and you are ahead of most of your market before you have written a single line of new content.

• Blocking AI crawlers by accident. The number one own goal. A plugin default or an old robots.txt rule quietly shuts the answer engines out while you wonder why nothing works. Check this first, every time.
• Inconsistent business details. Different names, addresses or phone numbers across the web split your evidence and weaken your entity. A machine reads the mismatch as uncertainty and plays it safe by leaving you out.
• Burying facts in images and scripts. Prices, hours and proof points that exist only as graphics or post-load JavaScript are invisible to the one reader that decides whether you get cited.
• Vague, benefit-soup copy. “We deliver world-class solutions for forward-thinking businesses” gives a model nothing to quote. Specific claims, real numbers and named methods do.
• Treating reviews as optional. Ignoring reviews is the difference between roughly 75% presence in relevant answers and roughly 1%. It is the biggest lever most businesses leave untouched.
• Writing for your org chart, not the question. Content organised around your service names rather than around what customers actually ask is hard for an engine to match to a real query.
• Letting content go stale. Freshness is decisive, so pages left untouched for a year quietly lose their citations to more recently updated rivals. Set a refresh cadence and keep it.
• Chasing a ranking that does not exist. There is no position one in an AI answer. Presence and share of voice over time are the scoreboard, and anyone promising you a fixed spot is guessing.

None of these is hard to fix. They are common precisely because most businesses have never looked, which is exactly why the opportunity is still open.

How to get started with AEO

Start by finding out whether AI can see you at all, then fix the foundations before you chase the clever stuff. The order matters, because there is no point optimising content a crawler cannot reach for a business a model cannot identify.

A sensible first 90 days looks like this:

1. Run an AI-visibility audit. Build your question panel and test it across the main assistants to get an honest baseline. Our step-by-step is how to check whether AI recommends your brand. Most businesses have never done this, and the result tends to be a shock. You cannot improve a number you have never measured.
2. Clear the technical blockers. Check robots.txt and your bot-protection for anything turning AI crawlers away, and confirm your key content renders as real text.
3. Fix your entity and NAP. Make your business unmistakable and consistent everywhere it appears.
4. Reshape your core pages into answer-first format with question-shaped headings, plain-text facts and proper schema.
5. Turn on the review engine. Start a steady, ongoing flow of genuine reviews and respond to them.
6. Build the off-site web. Accurate directory listings and reputable coverage that all corroborate the same story.
7. Re-run the panel monthly and follow the trend, adjusting where you are still absent.

If that reads like a lot, it is, but it is also the same foundational work that makes you stronger in ordinary search at the same time. You do not have to do it alone, and you do not have to do it all at once.

We built this playbook because we needed it ourselves, and now we run it for clients as part of our growth work. If you want to know exactly where you stand in AI answers today, we will run the audit on your business and show you the gaps, no jargon and no obligation. Grab a free growth review and we will start with the same 48-question test that put us on this path in the first place.