Written for two readers: you, and the answer engines. Each answer opens with the short version in plain English, then the detail underneath. Nothing here is gated and nothing is padded. If your question isn't here, email it to me and I'll add it.
Start here if AEO is new to you.
AEO is the practice of getting your business named and recommended by AI platforms — ChatGPT, Claude, Perplexity, Google AI Overviews, Gemini and Copilot — when someone asks for a recommendation.
Traditional SEO competes for a position on a page of ten blue links. AEO competes to be the answer itself. When a patient asks “who is the best implant dentist near me?”, the AI does not return ten options — it names one to three. AEO is the work that decides whether you are one of them: structured data the engines can parse, named and credentialled authors, consistent entity information across every register and directory, answer-shaped content, and third-party citations.
You need both, and they have come apart. SEO gets you ranked; AEO gets you cited. A first-page Google position is no longer a reliable proxy for being named in the answer above it.
Roughly 60% of AI Overview citations come from sources outside the traditional top three organic results, which means a firm that never ranked on page one can now be cited — and a firm that ranks well can be invisible. Good SEO foundations help AEO, but the additional work is different in kind: extractable answers, entity consistency, verifiable credentials, and structured data.
It is happening, and the gap between customer behaviour and business preparedness is the widest I have seen in any channel.
Roughly 47% of UK adults have used an AI-powered search tool, and about one in three UK Google searches now returns an AI Overview above the blue links. Over 55% of UK consumers have used AI to research a legal question before contacting a solicitor, and around 40% have asked a chatbot a health question. Meanwhile fewer than 5% of UK SMEs have taken any deliberate action on AI visibility. That gap is the whole opportunity, and it closes.
ChatGPT, Claude, Perplexity, Google AI Overviews and AI Mode, Gemini, and Microsoft Copilot — plus agent-readiness checks, included in every audit.
They do not share data sources or citation logic, so being visible in one does not mean being visible in another. Your audit reports each engine separately, with screenshots, because the fix for Perplexity is often not the fix for ChatGPT.
Very possibly. In a January 2026 study of 100 UK small businesses, 74 were actively blocking GPTBot in their own robots.txt — most had no idea, usually because a developer or plugin added it by default.
Only 3 of those 100 had any AI discovery files at all, and the average visibility score was 31 out of 100. This is the cheapest fix in the entire discipline and the one almost nobody checks. It is the first thing the free check tests.
Published, because you should not have to book a call to learn the cost.
Audit £595 one-off, credited in full against implementation. AI compliance review £450, or £895 for the audit and review together. Four-week implementation from £2,400. Watchdog maintenance from £350 a month, rolling. Sector Authority from £1,800 a month, by application.
There is no hidden pricing and no long-term contract on anything except Sector Authority, which is six months because the work genuinely takes that long. The free check comes first so that neither of us wastes a call.
Because those tools are a smoke alarm and I am the fire brigade. Buy the tool if you have someone in-house to act on it — genuinely. If you do not, it is a subscription to bad news.
Otterly starts around $29 a month, Semrush’s AI toolkit is about $99 per domain, Profound realistically $399–$499. Every one of them is a measurement product: they will tell you, with excellent charts, that ChatGPT names your competitor. None of them writes your schema, unblocks your crawlers, builds your named-practitioner pages or earns you a third-party citation. That is the work, and the work is what changes the answer.
That criticism is largely fair, which is exactly why I do not sell a £350 programme. The programme is the four-week build, paid once. £350 a month is what it costs to keep it working afterwards.
Most of an agency’s retainer floor is software cost: per-client licence fees to two or three tracking platforms, which is why a published “$99 starter plan” becomes £5,000–£15,000 a month across 25 clients. I built my own scanner, so I do not carry that. It is a cost structure, not a corner cut — and I publish my own score from that scanner so you can see I use it on myself.
UK AEO and GEO retainers run about £2,000–£10,000 a month for genuine programmes, with a one-off onboarding or audit fee of £1,500–£5,000 on top.
Analysis of 34 UK agencies in early 2026 put the tiers at £800–£2,000 (starter), £2,000–£5,000 (foundation), £5,000–£10,000 (authority) and £10,000–£20,000+ (enterprise). If you need a team, multi-brand coverage and 1,000+ tracked prompts, hire one of them — I will tell you so on the call.
If your tracked prompts have not moved in 90 days, the next three months are free — I keep working at my cost until the needle moves.
“Moved” is defined in writing before I start: named in an answer where you previously were not, on prompts we agree together. Not impressions, not a chart that trends up. I can offer this because I choose who I take on — if the free check shows nothing here worth fixing, I will say so and charge you nothing.
Who this works for, and who it does not.
UK clinics — private dentistry, aesthetics, specialist healthcare — and law firms. Practices where one new client is worth £1,500 or more and where clients research before they ever pick up the phone.
These are the sectors where AI-mediated research is already normal and where the fixes compound: named practitioners with verifiable credentials, published fee bases, and regulated identity that engines can confirm. I also take specialist B2B services, surveyors and recruitment, where 65% of buyers now use AI before first contact with a vendor.
Because the evidence is stark. An independent 2026 audit of 80 UK firms across 200 queries and five platforms found 87% had zero AI citation share for the queries their own clients ask.
Meanwhile 77% of UK legal queries trigger AI Overviews, and firms with practice-area content bylined by named, credentialled solicitors were cited four times more often than firms publishing under “the team”. That last finding is a fortnight of work, not a rebuild. The 13% of firms already visible are quietly taking the AI-driven enquiry market in their areas.
Technically yes, commercially probably not yet. At £150–£600 a job, my fees are hard to justify and I would rather tell you than take the money.
Do these three things yourself, free: check your robots.txt is not blocking GPTBot and ClaudeBot, put your services, areas covered and price bands on the page as real text rather than in images, and make sure your name, address and phone number match exactly across your website, Google Business Profile and Companies House. That is most of the available gain for a local trade. Run the free check first — it will tell you which of the three you are failing.
Almost never. In most audits, the majority of the fix is restructuring and marking up content you have already paid to have written.
The cheapest engagements I see are firms that already rank well on Google and are absent from the answer above it — their guides are good, they just do not lead with the answer and carry no structured data. If a rebuild genuinely is the right call, I will say so and I do not take a commission from web designers.
What to expect, month by month.
Technical fixes can register within days — a crawler unblock sometimes shows inside a fortnight. Consistent naming across engines takes 8–12 weeks. In regulated sectors, authority signals typically land in months four to seven.
Anyone promising faster in law or healthcare is guessing. The engines build recommendations from accumulated trust signals — citations, mentions, reviews, verifiable credentials — and those accumulate on their own schedule. You will see the technical scores move first, then the citations, then the enquiries.
A monthly page in plain English: your score, the prompts you now appear in, the prompts you do not, screenshots of who is named instead of you, and what I did that month.
The prompt list is agreed with you at the start and does not change without your say-so, because a moving target is how this discipline gets abused. You get the raw screenshots, not just a summary — and if a competitor overtakes you, that is in the report too.
Because AEO-REX is early stage and I would rather show you something checkable than a wall of quotes you cannot verify.
The teardowns are audits I ran on live, publicly accessible UK websites with identifying details removed. They are not clients, and I claim no results on their behalf — every “before” line is a fact I recorded on the date shown and can reproduce in front of you. As client work matures and clients consent, dated before-and-afters will replace them.
41 in March, 62 in July, targeting 85 by October. Re-scanned monthly on my own scanner and published whether it flatters me or not.
A consultancy that hides its own number is asking you to take the whole discipline on faith. The direction matters more than the figure, and you can watch mine.
Accurate, and deliberately unalarming. Nothing here is legal advice.
In the UK, no. There is no UK AI Act and no general duty to label AI use on a website. Anyone telling you otherwise is selling fear.
The UK regulates AI through existing regimes — UK GDPR, the automated decision-making rules at Articles 22A–22D inserted by the Data (Use and Access) Act 2025, and sector regulators including the ICO, FCA, MHRA and SRA. What that means in practice is narrower and duller than the headlines: if AI processes personal data, your privacy notice has to say so and explain how; if customers talk to a chatbot, the ICO expects that to be obvious from the interface rather than buried in a policy.
Only if your AI reaches EU users. The transparency duties in Article 50 apply from 2 August 2026, and they follow the users and the content rather than where you are based.
The four duties are specific: AI systems that interact with people must make clear they are machines; generative outputs must be marked as artificially generated; people exposed to emotion-recognition or biometric categorisation must be told; and deepfakes, plus AI-generated text published to inform the public on matters of public interest, must be disclosed. Machine-readable marking of synthetic content follows on 2 December 2026. A Birmingham dental practice with UK-only patients and no chatbot is very unlikely to be in scope. A UK firm running an AI assistant for EU clients is. Penalties reach €15 million or 3% of worldwide turnover — which is precisely why the honest answer matters more than the scary one.
Yes, and it overlaps almost perfectly with AEO: the things regulators require you to publish are the same things AI engines need in order to cite you confidently.
SRA transparency rules require firms to publish price and complaints information. Health advertising rules require accuracy and restrict claims. FCA rules require financial promotions to be fair, clear and not misleading. Consumer law bans fake and incentivised reviews. All of that is publication work — clear prices, verifiable credentials, accurate service descriptions, honest reviews — and all of it is exactly what makes an engine willing to name you. One piece of work, two reasons to do it.
Practically, yes — and it is more common than invisibility. Research published in July 2026 found AI chatbots returning false information about roughly half of London’s small businesses.
Wrong opening hours, services you no longer offer, prices years out of date. In a regulated sector that is worse than a marketing problem, because publicity accuracy is a regulatory expectation. Correcting the ground truth the engines draw on — your site, the relevant registers, directories and profiles — is a compliance improvement as much as a visibility one.
No. I am not a solicitor, a compliance officer or a regulated adviser, and I do not pretend to be.
What I do is technical and editorial: check what your website publishes against what your regulator requires you to publish, fix the technical layer, and flag anything that needs your compliance officer or solicitor. Everything I draft goes to you for approval before it is published, and you remain responsible for the content on your own site.
Where this is heading, without the futurism.
It is making your business legible and transactable to AI agents acting for customers rather than to customers directly. If you sell products online, it matters now. If you are a clinic or a law firm, it is a 2027 conversation.
Google and Shopify launched the Universal Commerce Protocol in January 2026; OpenAI’s Agentic Commerce Protocol, built with Stripe, is live with major merchants; Shopify’s agentic storefronts sell inside ChatGPT, Copilot, Gemini and Google AI Mode. The pattern is consistent: a merchant with a basic catalogue and a clean, structured feed gets recommended, while a sophisticated one with a poor feed gets skipped. Every audit I run includes an agent-readiness score at no extra cost, so you know where you stand — but I will not sell you a protocol you have nothing to plug into.
No. Paid placement will arrive and it will be expensive. The organic citation profile is what the engines weigh, and it compounds while you wait.
The cost of starting in 2026 is lower than the cost of starting in 2028, because the firms building citation authority now are extending a lead that is measured in months of accumulated signal. When ads do arrive, the businesses already cited organically will pay less for them, not more.
They are neutral at best and often harmful. Answer engines read crawlable HTML text; anything drawn by JavaScript after load is frequently invisible to them.
I have audited practices whose entire treatment menu was injected by script — beautiful to a human, absent to every crawler tested. Heavy 3D and WebGL also damage mobile load performance, which affects both search and conversion. Keep motion light and decorative, make sure every claim, price and credential exists as real text, and spend the budget on real photography instead of renders.
Put your domain in and see your score across the major engines, whether you're blocking the crawlers, and the three fixes that matter most. Sixty seconds, no card, no call unless you ask.
Check my practice free →Founder, AEO-REX. Ten years in STEM education, most recently Applied Science lecturer at BMet College, Birmingham. Speaker at Saïd Business School, Oxford. Every answer on this page is written and reviewed by me, and dated.