AI Search
How to Get Your Business Cited by ChatGPT and Google AI Overviews
Answer engines summarise your market and cite three sources. Here is how UK SMEs become one of them — entity clarity, question-led structure, quotable evidence and honest measurement.
Samuel McGarrigle · 31 July 2026 · 9 min read
A growing share of buyers now form a shortlist before they ever see a list of blue links. They ask ChatGPT, Gemini, Copilot or Perplexity who the credible options are, read a synthesised answer citing three or four sources, and act on it. If your business is not in that answer, you are not in the decision.
This is a representation problem, not a ranking problem
Search engines rank pages. Answer engines describe entities. They assemble a working picture of your business from your website, your structured data, your Google Business Profile, directories, trade publications, review platforms and old press mentions — then answer from that picture with a confidence level attached.
When those sources disagree, or say very little, the model hedges. Hedged answers do not contain recommendations. So the first job is rarely content: it is making sure every place your business is described says the same thing about what you do, who you serve, where you operate and what qualifies you to do it.
The five signals that decide who gets quoted
Direct answers. A page that answers one specific question completely, in its opening two sentences, is quotable. A page that warms up for four paragraphs is not. Give every real buying question its own heading and its own self-contained answer.
Entity clarity. Consistent name, location, service list, credentials and named authorship, backed by Organization, Person, Service and FAQPage schema. Schema does not manufacture authority — it removes ambiguity, and ambiguity is what keeps you out of answers.
Quotable evidence. Specifications, tolerances, lead times, pricing logic, comparison tables, process detail, honest limitations. Models cannot extract anything useful from "innovative, industry-leading solutions", which is exactly why that language never appears in citations.
Third-party corroboration. Answer engines lean on the sources their training and retrieval already trust for your sector: trade press, distributor and supplier sites, standards bodies, associations, review and comparison platforms. Being present there is closer to PR and partnership work than to link building.
Recency. These systems favour information that has been recently confirmed. A page last updated in 2021 is a weaker source than the same page reviewed last month.
What to publish if you want to be cited
Start with the questions your sales team answers every week and nobody has written down. How much does it cost, and what drives the price up. How long does it take. What are the alternatives and when is each one wrong. What standards or certifications apply. What goes wrong on badly specified projects. Who is this not suitable for.
That last category is the most under-used. Content that says plainly where you are not the right answer is unusually citable, because it is the kind of qualified statement a model can quote without risk — and it filters your enquiries before they reach you.
Format matters more than it used to. Tables, defined terms, explicit units, numbered processes and named components are easy to lift accurately. Information trapped in images, PDFs behind forms, or content that only renders after a click is effectively invisible.
How to measure something with no rank tracker
Build a fixed list of thirty to fifty real buying questions and test them across the major assistants at a set cadence. Record three things: whether you are cited, how you are described, and who is cited instead. That last column is the most instructive — it tells you exactly which third-party sources your sector's answers are built from.
Then add AI Overview presence for your commercial terms, referral traffic from assistant domains in analytics, and the signal that actually matters commercially: enquiries arriving already informed, saying they were pointed your way. Log the source of every enquiry in your CRM and the picture assembles itself within a quarter.
What to stop measuring
Total organic sessions will drift downwards as informational queries resolve without a click. Treating that as failure leads businesses to defend traffic that never produced revenue. Cost per qualified enquiry, enquiry quality and pipeline contribution remain meaningful no matter how discovery changes.
The window is open, briefly
Almost none of your competitors have done this work. Entity alignment takes days, structural fixes take weeks, and citation share builds over three to six months of consistent publishing. That is an achievable head start for an SME — and it will be considerably harder to buy in two years.
