How to get your local business cited in AI Overviews (without gaming it)

To get quoted in AI Overviews, make each page the clearest answer on the web for one local question: a direct answer up top, a real price range with the factors that change it, plain question-and-answer sections, and consistent business facts. Thin, near-duplicate pages get skipped, not cited. There is no AI-only trick underneath that, and this guide is the do-this-on-your-page version of it.

Figures cited below are current as of 31 July 2026.

Is there a special trick for AI Overviews?

No. The honest answer is that there is no separate checklist for Google’s AI features beyond being indexed and eligible to appear in a normal snippet. The same things that make a page genuinely useful to a person in Uckfield are the things that make it liftable by a model: a clear answer, specific numbers, and facts a machine can match against the rest of the page.

Google is unusually blunt about this. In its guide to optimising for generative AI features, published in May 2026, it states that you do not need to create machine-readable files, AI text files, special markup, or Markdown to appear in generative AI search, and that breaking your content into tiny chunks for AI systems is unnecessary because its systems already extract relevant passages from ordinary pages. See Google’s guide to optimising for generative AI features for the exact wording. In other words, good people-first SEO is the optimisation. There is no llms.txt hack, no AI-only schema, and no chunking ritual that beats simply being the best answer.

What makes a page quotable?

A quotable local page is one a model can lift a clean, correct passage from without having to reassemble it. Four things do most of the work, and they are all plain server-side text, never an image:

A direct answer in the first 200 words, written so it stands alone if it is pulled out of the page.
A real price range with the factors that move it, so the number means something rather than reading as a guess.
Plain question-and-answer sections that match how people actually ask.
Consistent business facts: the same name, service area, and phone number everywhere they appear.

Take a worked example. An engineer covering Uckfield and Lewes opens a boiler-repair page with: “Most boiler repairs in the Lewes area run £90 to £280, depending on the fault, the age of the boiler, and whether parts are in stock. A standard diagnostic visit is a £45 call-out within ten miles of Lewes.” That passage is liftable on its own. It carries a named area, a real band, the factors behind it, and a concrete call-out figure. A page that instead says “we offer competitive boiler repair across Sussex” gives a model nothing to quote, because it says nothing a hundred other pages do not.

There is good third-party evidence that this kind of specificity helps. The Princeton-led GEO paper (Aggarwal et al., introduced November 2023 and published in the KDD 2024 proceedings) tested nine tactics across thousands of queries and found that adding relevant statistics, citations, and direct quotations raised a source’s visibility in AI answers by up to around 40 per cent. Treat that as directional rather than a promise: it tells you which way to push, not what you will get.

Why do thin location pages get skipped, not cited?

A model citing a local answer is choosing the clearest source, and a near-duplicate page is the opposite of clear. If forty area pages share the same paragraphs with only the town name swapped, there is no single page that is obviously the best answer for any one of them, so none gets quoted. Worse, Google’s scaled-content policy treats pages mass-produced without adding value as spam, which is a ranking problem before it is ever a citation problem. See Google’s spam policies for the current definition, and our plain-English scaled content policy guide for what it means in practice.

The fix is substance, not volume. Each page you keep needs something genuinely local in it: a named area you actually cover, the housing stock you see most around the High Street, a job you did last winter, your call-out radius. That is exactly the difference between a page worth citing and a page worth collapsing. For the underlying retrieval theory, read how AI search finds local businesses first, then come back here for the template.

A service-page template you can copy

Use the same skeleton on every service or location page and let the local detail vary. In order, top to bottom:

An H1 that names the service and the area in plain words.
A two-to-three-sentence direct answer immediately under it, with a real price band and the factors that move it.
A short section on what the job actually involves locally, with at least one detail that could only be true of your area.
Plain question-and-answer sections covering the things people email and ring to ask.
Your name, service area, and phone number, identical to how they appear on every other page and on your Google Business Profile.

For the writing itself, the local landing page content guide covers what to put on a page when every town feels the same, and the guide to FAQs on location pages without duplicate content covers how to vary the answers so the pages do not read as copies of each other.

What about schema?

Structured data is not required for AI Overviews or AI Mode, and there is no special AI-only markup that earns a citation. Keep the ordinary schema a normal local page already wants: an Article reference for the guide itself, and the LocalBusiness and Service markup for the service pages, with the service area and contact details in the markup matching exactly what is visible on the page. Mismatched or invisible markup is a structured-data manual-action risk, not an advantage.

FAQ markup is a special case worth being clear about. FAQ rich results stopped appearing in Google Search on 7 May 2026, with the search-appearance filter, rich-result report, and Rich Results Test support removed through the rest of the year, per the deprecation notice on Google’s FAQ structured-data documentation. The practical takeaway is to keep your FAQs as plain on-page text. It is still excellent for readers and easy for a model to extract; it just no longer earns a rich result.

Where Townsmith fits

Townsmith is a free WordPress plugin that builds real, editable location pages and scores each one out of 100 before you publish, holding a weak page back with an advisory gate so a thin near-duplicate does not go live by accident. It runs entirely on your own server, makes no external calls, and writes nothing for you. It complements Rank Math, Yoast, AIOSEO, or SEOPress rather than replacing them. The score rewards the same things this guide does: a real local paragraph, a genuine price band, distinct text from page to page. You can see how the score is built in the quality score documentation, and what the engine does on the features page.

None of this guarantees a citation. Nothing can. What it does is make each page the clearest, most specific answer to one local question, which is the only thing inside your control and the thing both Google’s own guidance and the independent research point at.

Common questions

Do you need FAQ schema for AI to read your answers?

No. Models read the visible text on the page, so plain question-and-answer sections are enough. Adding FAQPage markup will not earn you stars, and claiming a rich result you cannot get risks a structured-data manual action if the markup does not match what is visible. Keep the questions, drop the expectation of a rich result.

Is there any markup that gets you into AI Overviews?

No. Google states plainly that no special markup, AI text file, or Markdown is needed to appear in its generative AI features. Ordinary, accurate schema that matches your visible content is worth having for normal search, but it is not an AI cheat code.

Will more location pages get me cited more often?

Usually the opposite. A pile of near-duplicate pages gives a model no single best answer to quote and risks tripping the scaled-content policy. Fewer pages, each with real local substance, are far more likely to be the source an AI answer reaches for.

Should you put a price on the page if it varies?

Yes, as a band with the factors behind it. A range like £90 to £280 with the reasons it moves is specific enough to be quotable and honest enough to set expectations, which is exactly what makes it useful to both a reader and a model.