AI Visibility Scan: Is Your Business Showing in AI Answers?
An AI visibility scan answers one question: when a customer asks an AI system for a business like yours, does your name come up, and if it does, is the answer accurate? More people now get their first answer from AI Overviews, AI Mode and chat-based assistants before they ever see a list of links, so a business can hold decent search rankings and still be invisible in AI answers. This guide explains what AI visibility is, how to run the scan yourself in an afternoon, how to score what you find, and what actually moves the score, with the same checks a professional AI visibility audit would run.
- AI visibility scan at a glance
- What AI visibility means
- Why it matters now
- Where AI answers get their information
- The scan you can run yourself
- The prompts to test
- How to score AI visibility
- Reading the results: five common findings
- What makes AI answers mention a business
- The fix list, in priority order
- Service businesses and local questions
- Free checkers and trackers vs a manual scan
- How often to rescan and what to track
- AI visibility, GEO, AEO and SEO: the terms
- Mistakes that lower AI visibility
- What a professional AI visibility audit adds
- A worked example: a two-van heating firm
- The AI visibility scan checklist
- Frequently asked questions
- Get your AI visibility scan
AI visibility scan at a glance
The scan is a structured test of how AI systems answer the questions that lead to enquiries in your category. It is not a keyword ranking report and it is not a website crawl, although both feed into it. The short version is in the table below; the rest of the guide is the long version.
| Question | Short answer |
|---|---|
| What is being measured? | Whether AI answers mention, cite, recommend or misdescribe your business for a set list of customer questions, and how that compares with competitors. |
| Where do you test? | Google AI Overviews and AI Mode, the general-purpose AI assistants your customers use, and any answer engine with a real share of your audience. |
| What counts as “showing”? | A named mention, a citation link to your site, or a direct recommendation. Each is scored separately because each is earned differently. |
| How long does it take? | A first manual scan of 20 to 30 questions across three or four AI platforms takes two to three hours. A full audit with fixes takes days. |
| What is a good result? | There is no industry standard score. The useful benchmark is your share of mentions against the three or four competitors who keep appearing instead of you. |
| What changes the result? | Clear, factual pages about your services and areas, consistent business details everywhere, reviews and mentions on third-party sites, structured data, crawlable pages and a real presence in the places AI systems draw from. |
| How often to repeat? | Monthly for the core questions, quarterly for the full list, and again after any site change, rebrand or new location. |
What AI visibility means
AI visibility is a measure of how often, and how favourably, a brand appears inside AI-generated answers. When someone asks an AI system a question, the answer is assembled from what the model already knows and, increasingly, from live search results and web pages it retrieves at the moment of asking. AI visibility measures whether your business is part of that answer: named in it, linked from it, recommended by it, or described correctly in it.
It sits next to search visibility rather than replacing it. Classic search visibility asks where your pages rank in a list of links. AI visibility asks whether you are in the answer that now sits above, or instead of, that list. The two overlap because AI systems lean heavily on the same signals that search engines use, but they are not identical, and a business can do well in one and badly in the other.
The four things a scan records
Most AI visibility audits reduce to four observations for every question tested:
- Mentions. Your business name appears in the answer. A mention is the basic unit of AI brand visibility.
- Citations. The answer links to one of your pages, or shows it as a source. Citations are what send traffic; mentions alone often do not.
- Recommendations. The answer actively suggests you for the task, sometimes in a ranked or shortlisted form, sometimes with reasons.
- Accuracy and sentiment. What the answer says about you: services, prices, areas, opening hours, reputation. Wrong details and lukewarm descriptions are their own problem, separate from not appearing at all.
The AI visibility score that free checkers produce is usually some combination of the first three, counted across a sample of prompts. That is useful as a snapshot, and misleading if you treat it as the whole picture, for reasons covered in the scoring section.
What AI visibility is not
It is not a measure of whether your site uses AI, produces AI-written content or has an AI chatbot. It is not the same as being findable by AI crawlers, although crawler access is one of the inputs. And it is not a ranking position in the old sense, because AI answers do not have ten blue links in order; they have a handful of named sources and an answer that changes wording every time it is generated.
Why it matters now
Three shifts made AI visibility a real business question rather than a marketing trend.
The answer moved above the links. Google’s AI Overviews now appear on a large share of informational searches, and AI Mode turns the search box into a conversation. Google’s own guidance on AI features says the same fundamentals that help a page rank help it appear in AI features, and that there is nothing extra to opt into; the point for a business owner is that the answer at the top is now assembled, not listed.
Buyers ask assistants first. A growing share of people start a purchase, a hire or a comparison inside a chat-based assistant, asking things like “who is a good accountant for a small limited company in Leeds” or “which local firm installs heat pumps and handles the grant paperwork”. If the answer names four firms and none of them is you, the enquiry goes elsewhere before your website is ever opened.
Clicks on answered questions are falling. Independent research on AI summaries in search, including work published by the Pew Research Center, found people are less likely to click a result when an AI summary is shown. The exact percentages vary by study and by query type, and the figures should be treated as directional, but the direction is consistent: informational clicks are thinner, and the searches that still send people to websites are the ones where they need a specific business, a price or a booking.
Put together, that means the queries that matter most commercially are exactly the ones where AI answers now shortlist businesses. Being absent from the shortlist is a quiet loss: nothing in your analytics shows the enquiry you never received.
Where AI answers get their information
To improve AI visibility you need to know where the answer comes from, because each source is influenced differently. Broadly, an AI system draws on three layers.
1. Training data
Large language models are trained on a huge snapshot of the public web, books and other text. If your business has been written about for years on many sites, the model may “know” you from training alone. Most small and mid-sized businesses have little or no footprint in training data, which is normal and not something you can change quickly. It is why the next two layers matter far more for local and service businesses.
2. Live retrieval
Modern assistants and AI search features fetch information at the time of the question. They run searches, read pages, pull business listings, reviews and structured data, and then write the answer from what they retrieved. This is where most of your visibility is won or lost, and it is why classic SEO fundamentals still matter: pages that rank, load, explain clearly and carry structured data are the pages that get retrieved and cited.
3. Third-party sources
AI systems weight what other sites say about you: directories, review platforms, industry publications, local news, trade bodies, supplier pages and comparison articles. When an assistant answers “best emergency plumber in Nottingham”, it is often summarising a handful of listicles, review profiles and local guides rather than individual plumbers’ sites. Being present, accurate and well reviewed in those sources is a large part of AI brand visibility, and it is the part most businesses have never looked at.
The practical rule: your own website controls accuracy and citations; third-party sources control whether you are shortlisted at all. A scan has to look at both.
The scan you can run yourself
You do not need a tool to run a first AI visibility scan. You need a spreadsheet, a list of the questions your customers actually ask, two or three hours, and the discipline to record what you see rather than what you hoped to see. The steps below are the manual version of what paid trackers automate.
Step 1: write down the questions that lead to enquiries
Start with buyer questions, not keywords. Keywords are what people type into a search box; prompts are what they say to an assistant, and they are longer, more specific and usually contain the job, the place and a constraint. A plumbing firm’s list might include “who can replace a boiler in Coventry this week”, “is a combi or a system boiler better for a four-bedroom house” and “what does an emergency call-out cost at the weekend”. A dental practice might list “private dentist in Bristol that takes nervous patients” and “how much is a dental implant”. Aim for 20 to 30 prompts covering four groups:
- Category prompts: “best [service] in [town]”, “[service] near [area]”, “recommend a [type of business] for [situation]”.
- Comparison prompts: “[you] vs [competitor]”, “is [service A] or [service B] better for [need]”.
- Cost and process prompts: “how much does [service] cost”, “how long does [service] take”, “what should I ask before hiring [type]”.
- Brand prompts: “what does [your business] do”, “is [your business] any good”, “[your business] reviews”.
Step 2: choose the AI platforms to test
Test where your customers are, not everywhere. For most UK and US service businesses that means Google’s AI Overviews and AI Mode first, because they sit inside the search that people already use, then the two or three general-purpose assistants with the largest consumer audiences, and any answer engine popular in your industry. Run each prompt in a fresh session with no personalisation where the platform allows it, and note the location setting, because local prompts change completely with location.
Step 3: record what the answer actually contains
For every prompt on every platform, record five things in the sheet:
| Column | What to write |
|---|---|
| Mentioned? | Yes or no. If yes, the exact wording used for your business name. |
| Cited? | Yes or no, and which page of yours the answer linked to or listed as a source. |
| Recommended? | Yes or no, and the position if the answer gave a shortlist (first of four is not the same as fourth of four). |
| Accuracy | Any wrong detail: services you do not offer, areas you do not cover, old prices, a closed branch, a misspelt name. |
| Who appeared instead | Every competitor named, and every third-party source the answer leaned on (a directory, a review platform, a local guide, a comparison article). |
Screenshot each answer as well. AI answers change from one generation to the next, and when you rescan in a month you will want the evidence, not a memory.
Step 4: run the same prompts twice
Because generated answers vary, a single run overstates or understates visibility. Run the core prompts at least twice, ideally on different days, and record both. A business that appears in one of two runs has partial visibility, which is a different problem from appearing in neither.
Step 5: scan the third-party sources the answers relied on
This is the step most people skip. Take the list of sources the answers cited for your category prompts and open each one. Are you listed? Is the listing accurate? Do you have reviews there? Is your category correct? If three answers all drew from the same two local guides and one review platform, those three pages are now your most important marketing pages, and none of them is on your website.
Step 6: check the technical basics on your own site
AI systems cannot cite what they cannot read. Check that your robots file is not blocking the AI crawlers you want to allow, that important pages are indexable and load quickly, that each service page states plainly what you do, where and for whom, and that structured data describes the business correctly. A quick technical health check is enough at this stage; a full audit comes later if the scan shows problems.
The prompts to test
A scan is only as useful as its prompt list. These templates cover most service businesses; swap in your services, towns and buyer situations.
| Group | Prompt template | Why it matters |
|---|---|---|
| Local shortlist | “Recommend a [service] in [town] for [situation]” | The highest-intent prompt. Being absent here costs enquiries directly. |
| Near-me equivalent | “Who does [service] near [neighbourhood or postcode area]?” | Tests whether your service area is understood. |
| Trust | “Is [your business] reputable? What do reviews say?” | Shows which review sources the AI trusts and what it repeats. |
| Comparison | “[You] or [competitor], which is better for [need]?” | Reveals how the AI describes your strengths, or invents them. |
| Cost | “How much does [service] cost in [region]?” | If the answer cites a competitor’s pricing page, that is a content gap on yours. |
| Process | “What happens when I hire a [type of business] for [job]?” | Explainer content earns citations even from non-local prompts. |
| Qualification | “What should I check before choosing a [type of business]?” | Authority prompts: the sources cited here shape shortlists. |
| Brand facts | “What services does [your business] offer and where?” | Pure accuracy test of how the AI describes you. |
Keep the list stable between scans. The value comes from comparing the same prompts over time, not from testing new ones each month.
How to score AI visibility
Turn the sheet into numbers so you can track change. The metrics below are the ones that matter for a business rather than for a marketing report.
| Metric | How to calculate it | What it tells you |
|---|---|---|
| Mention rate | Prompts where you were named, divided by prompts tested | Whether AI systems know you exist for your category |
| Citation rate | Prompts where one of your pages was cited, divided by prompts tested | Whether your content is being used as a source (and can send traffic) |
| Recommendation rate | Prompts where you were actively suggested, divided by shortlist-type prompts | Whether you are on the buyer’s shortlist |
| Share of voice | Your mentions divided by total mentions of you plus named competitors | Your position relative to the businesses that appear instead of you |
| Average position | Your mean place in ranked shortlists where you appeared | Whether you are the first suggestion or an afterthought |
| Accuracy rate | Answers with no factual errors about you, divided by answers that mentioned you | Reputation and data-consistency problems |
| Sentiment | Positive, neutral or negative tone of each description | How the AI summarises your reputation from reviews and mentions |
What is a good AI visibility score?
There is no agreed benchmark, and anyone quoting one is describing their own tool’s scale. AI answers are probabilistic, platforms differ, and a national brand and a two-van plumbing business are not on the same curve. The benchmark that is actually useful is relative: compare your share of voice with the three or four competitors who keep appearing for your category prompts. If they are mentioned in 60 percent of local shortlist prompts and you are mentioned in 10 percent, that gap is the number to work on. If nobody in your town is mentioned and answers fall back to national directories, the opportunity is to become the first local business the AI can confidently name.
Weighting the score
If you want a single AI visibility score for reporting, weight the metrics by commercial value rather than averaging them. A reasonable weighting for a service business is recommendation rate on shortlist prompts first, share of voice second, accuracy third, mention rate and citation rate after that. A high mention rate with poor accuracy is not a good result; being named for the wrong services attracts the wrong enquiries.
Reading the results: five common findings
The results of a scan tend to fall into a small number of patterns. Each has a different cause and a different fix.
Finding 1: not mentioned at all, competitors are
The most common result for a small business. The AI knows the category and the town, names three or four firms, and none is you. Usually the cause is thin third-party presence: you are not in the local guides, directories and review platforms the answers draw from, or your listings there are incomplete. Sometimes the cause is on your own site: no page states plainly that you provide that service in that place, so there is nothing to retrieve.
Finding 2: mentioned, but never cited
The AI names you, often because a directory or review site mentions you, but never links to your website. Your pages are not being retrieved as sources. Typically the service pages are vague, image-heavy or missing the facts an answer needs: prices, areas, process, qualifications. Sometimes the pages are blocked to AI crawlers or are slow enough to be skipped.
Finding 3: mentioned with wrong details
You appear, but with an old address, a service you stopped offering, a price from years ago or a branch that closed. The AI is repeating inconsistent data across the web. The fix is data hygiene: make the name, address, phone, services and hours identical everywhere, correct the third-party listings, and give the correct version in structured data on your own site.
Finding 4: cited from someone else’s page
The answer talks about you but the source it cites is a comparison article, a review platform or a supplier’s page. That is not bad, and often it is how the shortlist works, but it means someone else controls the wording. The response is to make your own page the better source for the same question, and to make sure the third-party page is accurate.
Finding 5: visible for informational prompts, invisible for buying prompts
Your explainer blog posts are cited for “how does X work” questions, but the “recommend a firm” prompts never name you. Content authority and business authority are different things. The second comes from reviews, local presence, mentions on trusted sites and clear service pages, not from more articles.
What makes AI answers mention a business
AI visibility is earned through inputs that AI systems can read, verify and cross-reference. Most of them are ordinary marketing and SEO work done properly; the difference is that AI answers punish vagueness and inconsistency more sharply than a list of links ever did.
Entity clarity
AI systems work with entities: a named business, with a type, a location, services and relationships to other entities. If your site never states what you are in plain terms, or your name appears in three variations across the web, the entity is fuzzy and the AI hedges by leaving you out. State the business name, what it does, where it does it and who it serves in the first paragraph of the homepage and every service page, and keep it identical everywhere else.
Clear, factual service pages
AI answers are built from facts that can be lifted: what is included, what it costs, how long it takes, which areas are covered, what qualifications apply. Pages written as adjectives and slogans give the AI nothing to quote. Pages written as plain statements, short definitions, tables and question-and-answer sections get retrieved and cited. This is the single biggest on-site lever for citation rate.
Consistent business details
Name, address, phone, opening hours and service list should match on your website, your Google Business Profile, every directory and every social profile. Inconsistency is read as uncertainty, and uncertainty is exactly what an AI system avoids when recommending a business.
Reviews and reputation signals
When an AI describes whether a business is reputable, it is summarising reviews and mentions it can find. Volume, recency and the wording of reviews shape the sentiment of the answer. A steady flow of genuine reviews on the platforms your customers use matters more here than almost anything on your own site.
Third-party mentions and citations
Being listed in local guides, industry directories, trade association members’ lists, supplier partner pages and local press builds the web of corroboration that AI systems rely on. This overlaps with classic link building, but the goal is different: a mention that describes you accurately is valuable even if the link is nofollow, because it is the description the AI will repeat.
Structured data
Schema markup describing the organisation, its local business details, services, prices, reviews and FAQs gives AI systems a machine-readable version of your facts. It does not force inclusion, but it removes ambiguity and it is one of the cheapest fixes on the list.
Crawlability and speed
Pages that AI crawlers can fetch quickly and parse easily are the pages that get used. Check the robots file, avoid rendering critical content only through scripts, keep pages fast and make sure the key facts are in the HTML rather than in images. A page that takes six seconds to load may simply be skipped during retrieval.
Topical authority
A site that covers its subject thoroughly, with connected pages answering the questions buyers ask before, during and after a purchase, is more likely to be treated as a reliable source than a five-page brochure site. This is where a content plan and internal linking earn their keep, and where a blog that answers real questions pays off in citations for informational prompts.
The fix list, in priority order
Most scans produce a long list of findings. Fix them in this order; the early items are cheap and move several metrics at once.
| Priority | Fix | Metric it moves | Effort |
|---|---|---|---|
| 1 | Correct every inconsistent business detail across your site, Business Profile, directories and social profiles | Accuracy, mention rate | Hours |
| 2 | Rewrite the homepage opening and each service page so the first paragraph states what, where and for whom in plain words | Mention rate, citation rate | Days |
| 3 | Add the facts AI answers need to each service page: inclusions, price ranges, timescales, areas, qualifications, a short FAQ | Citation rate | Days |
| 4 | Add or fix Organization, LocalBusiness, Service and FAQ structured data | Accuracy, citation rate | Hours |
| 5 | Check robots rules, indexability and speed on the pages you want cited | Citation rate | Hours |
| 6 | Claim and complete listings on the third-party sources the scan showed answers rely on | Recommendation rate, share of voice | Days |
| 7 | Build a steady review process on the platforms that were cited for trust prompts | Sentiment, recommendation rate | Ongoing |
| 8 | Earn accurate mentions: local press, trade bodies, supplier pages, comparison articles, guest contributions | Share of voice, mention rate | Weeks |
| 9 | Publish explainer content that answers the process, cost and qualification prompts better than the current cited sources | Citation rate for informational prompts | Weeks |
| 10 | Rescan, compare against the first sheet, and repeat for the prompts that did not move | All | Monthly |
Notice what is not on the list: producing more AI-written articles, adding “AI” to your page titles, or paying for placements that promise instant inclusion. None of those changes what an AI system can verify about your business.
Service businesses and local questions
The headline that prompted this guide announced a free AI visibility scan aimed at service businesses in one US state, and the framing is right: trades, clinics, practices, agencies and local specialists are the businesses most exposed to AI shortlists, because their customers ask location-specific, trust-heavy questions that AI systems now answer with names.
How local prompts are answered
Ask an assistant for a plumber in a specific town and the answer usually combines three sources: the local pack and Business Profile data for the area, the review platforms and directories that list plumbers there, and any local guide or comparison article that ranks them. Your own website is often the fourth source, used to confirm details. That order explains why a firm with a mediocre website but a complete Business Profile, two hundred recent reviews and listings on the right local sites can outscore a firm with a beautiful site and nothing else.
The service-area problem
AI answers are literal about places. A business based in one town that serves six others will be shortlisted for the home town and ignored for the rest unless its pages and listings say, in words, that it covers those places. Service-area pages that name the areas, describe the work done there and carry consistent contact details fix this, and they are the same pages that local SEO needs anyway.
Trust prompts for service businesses
“Is this firm reliable”, “are they Gas Safe registered”, “do they turn up on time” are answered from reviews and from qualification pages. Publish the credentials, registration numbers, insurance and guarantees on a page an AI can read, and collect reviews that mention the things buyers ask about. Sentiment in AI answers is largely borrowed from review wording.
US and UK differences
The mechanics are the same on both sides of the Atlantic; the sources differ. US answers lean more on large review platforms and city guides; UK answers lean more on the Business Profile, trade-body registers and regional directories. If you serve both markets, scan them separately with the location set appropriately, because a single scan from one country tells you nothing about the other.
Free checkers and trackers vs a manual scan
A number of free AI visibility checkers now offer an instant score from a domain name, and paid AI visibility trackers monitor brand mentions across platforms daily. They are useful, with limits worth knowing before you trust the number.
| Approach | What it does well | What it misses |
|---|---|---|
| Free instant checker | Quick snapshot from a small prompt sample; good for a first look and for comparing a few competitors | Generic prompts rather than your customers’ questions; often one platform; no accuracy or sentiment check; no view of third-party sources |
| Paid AI visibility tracker | Daily monitoring, share-of-voice trends, citation tracking across several platforms, alerts | Still depends on the prompt list you give it; cost adds up for a small business; does not fix anything |
| Manual scan (this guide) | Your real prompts, your locations, accuracy and sentiment recorded, the third-party source map that explains the results | Time-consuming; sample size limited; needs repeating to smooth out answer variation |
| Professional audit | Everything above plus the technical, content and data fixes, prioritised and carried out | Costs money; choose one that shows you the raw prompt sheet, not just a score |
A sensible sequence for most businesses is a free checker to confirm the problem exists, a manual scan to understand it, and a tracker only once you are actively working on it and want to see the trend.
How often to rescan and what to track
AI answers change with model updates, index changes and your own work, so a single scan is a starting point rather than a verdict. A workable cadence:
- Monthly: the ten highest-intent prompts on the two platforms that matter most, two runs each, same sheet.
- Quarterly: the full list on every platform tested in the first scan, plus a fresh look at which third-party sources are being cited.
- After changes: any rebrand, new location, new service line, site migration or big content update, because each can change how the entity is understood.
Track five numbers over time: mention rate, citation rate, recommendation rate on shortlist prompts, share of voice against the same competitors, and accuracy rate. Alongside them, watch referral traffic from AI platforms in your analytics and enquiries that say “the AI recommended you”; ask on the enquiry form, because that is the only place the effect becomes visible.
AI visibility, GEO, AEO and SEO: the terms
The industry has produced several overlapping labels. They describe the same work from different angles.
| Term | Meaning | How it relates |
|---|---|---|
| AI visibility | The measurement: how often and how well a brand appears in AI-generated answers | The outcome the others are trying to improve |
| Generative engine optimisation (GEO) | Optimising content and presence so generative AI systems draw on and cite it | The content and authority side of the work |
| Answer engine optimisation (AEO) | Structuring information so answer engines can extract direct answers | The formatting and structured-data side |
| Search engine optimisation (SEO) | Earning visibility in search results through relevance, authority and technical quality | The foundation the others stand on; AI systems retrieve from search |
| AI search optimisation | Umbrella term for GEO and AEO applied to AI search features | Marketing shorthand; same inputs |
| Share of voice | Your mentions as a share of all brand mentions in a set of answers | The most useful comparative metric |
| Citation | A source link or reference in an AI answer | The metric tied to traffic |
A business does not need separate GEO, AEO and SEO programmes. It needs clear pages, consistent data, real reputation signals and technical health, then a scan to check whether AI systems are picking them up.
Mistakes that lower AI visibility
- Testing from a personalised, logged-in session and concluding you are visible because the assistant remembered your earlier chats.
- Scanning once and treating the result as fixed, when answers vary run to run.
- Chasing a tool’s score instead of the share of voice on the prompts that produce enquiries.
- Publishing volume for its own sake: dozens of thin AI-written posts add nothing an AI system can verify and can dilute the pages that matter.
- Keyword stuffing “AI” phrases into titles and headings; AI systems retrieve for meaning, not for the word AI.
- Blocking AI crawlers by default because of a general worry about scraping, then wondering why nothing is cited.
- Buying reviews or mentions: inconsistent, obviously artificial signals get discounted, and they damage the accuracy of what is said about you.
- Ignoring the third-party layer and fixing only the website, when the shortlist was assembled from directories and review sites.
- Letting old details live on across profiles you forgot you had; the AI has not forgotten them.
What a professional AI visibility audit adds
The manual scan tells you where you stand. A professional audit adds the diagnosis and the fixes: the prompt sheet run at scale across platforms and locations, the third-party source map for your category, a technical review of crawler access, indexability, speed and structured data, a content review of every service page against the facts AI answers need, an entity and data-consistency review across the web, and a prioritised plan with the work carried out rather than listed. It is the AI layer of the same work covered by a full SEO audit, and it is part of what an AI SEO service does month to month rather than once.
If you are weighing whether the AI answer layer is worth attention at all, the earlier piece on AI Overviews and keeping clicks covers which query types are losing traffic to AI answers and which are not, which is the context an audit sits in.
Whichever route you take, insist on seeing the raw prompt sheet. A score without the prompts behind it cannot be checked, and an audit you cannot check is a sales document.
A worked example: scanning a two-van heating firm
To make the method concrete, here is how the scan plays out for a typical small heating and plumbing business: two engineers, one town as a base, five surrounding towns served, a ten-page website, a Business Profile with forty reviews and listings on a couple of trade directories. The example is illustrative rather than a client case.
Prompt list. Twenty-four prompts: six local shortlist prompts across the base town and the two biggest surrounding towns, four trust prompts, four cost prompts (boiler replacement, annual service, emergency call-out, power flush), four process prompts, four comparison prompts against the two firms that dominate the local pack, and two brand prompts.
First-run results. Mentioned in 4 of 24 prompts, all for the base town. Cited in 1 (the boiler-service price page). Recommended in 2 shortlist prompts, both in third position. Accuracy problems in 2 answers: one listed a town the firm no longer covers, one quoted a call-out fee from an old directory listing. Competitors appeared in 15 prompts; the two local-pack leaders appeared in 11 each. Third-party sources cited: one regional trade directory in nine answers, one national review platform in seven, one “best boiler installers in [county]” article in five.
Diagnosis. The firm is a fuzzy entity outside its base town, its service pages carry no facts an answer can lift, the old directory listing is feeding wrong prices, and the three sources driving the shortlists do not list it at all or list it thinly.
Fixes, in order. Correct the old listing and align details everywhere. Rewrite the five service pages with plain opening paragraphs, price ranges, timescales and a short FAQ each, and add service-area sections naming every town. Add LocalBusiness, Service and FAQ structured data. Claim and complete the regional directory and the review platform profile, and approach the county article’s publisher with accurate details. Start asking every completed job for a review that mentions the town and the job type.
Rescan after eight weeks. The realistic expectation is not first place everywhere. It is: mentions rising from 4 to somewhere around 10 of 24, citations from 1 to 3 or 4 as the rewritten pages get retrieved, accuracy errors gone, and appearance in shortlists for at least one surrounding town. The comparison prompts move last, because they depend on reviews accumulating. Numbers like these are what an honest AI visibility programme reports: movement on the prompts that matter, with the sheet to prove it.
The AI visibility scan checklist
Everything above, in one list to work through.
- Write 20 to 30 real customer prompts across shortlist, comparison, cost, process, qualification and brand questions.
- Pick the platforms your customers use; test Google AI Overviews and AI Mode first for search-led enquiries.
- Run every prompt in a clean session with the right location, twice, on different days.
- Record mention, citation, recommendation and position, accuracy errors, sentiment, and every competitor and source named.
- Calculate mention rate, citation rate, recommendation rate, share of voice, average position and accuracy rate.
- Benchmark against the three or four competitors who appear most, not against a tool’s generic score.
- Open every third-party source the answers cited and check whether you are listed accurately.
- Check crawler access, indexability, speed and structured data on the pages you want cited.
- Fix data consistency first, then service-page facts, then structured data, then third-party listings, then reviews, then earned mentions.
- Rescan monthly on the core prompts and quarterly on the full list, keeping the same sheet.
Frequently asked questions
What is an AI visibility audit?
A structured review of whether AI systems mention, cite, recommend and correctly describe a business for the questions its customers ask, followed by the technical, content and data fixes that change the result. The scan described in this guide is the measurement half; the audit adds diagnosis and fixes.
How can I check my AI visibility for free?
Run your own prompts in the AI platforms your customers use, in a clean session with the correct location, and record whether you are mentioned, cited or recommended. Free instant checkers give a quick snapshot from a generic prompt sample, but a manual scan with your real questions is more accurate and costs only time.
What does AI visibility mean?
How often and how favourably a brand appears inside AI-generated answers: named, linked as a source, recommended or described accurately. It is the AI-answer equivalent of search visibility.
What is a good AI visibility score?
There is no standard scale. The meaningful measure is your share of voice against the competitors who appear for your category prompts, tracked over time on the same prompt list. A score from a tool is only comparable with the same tool’s score for the same prompts next month.
How is AI visibility calculated?
Most methods count mentions, citations and recommendations across a sample of prompts on one or more platforms, then express each as a rate and combine them. Better methods add accuracy, sentiment and position in shortlists, and compare the result with named competitors.
Why should I track AI brand visibility?
Because AI answers now shortlist businesses before people open a website, and nothing in your analytics shows the enquiry you never received. Tracking reveals whether you are on the shortlist, who is on it instead, and whether what is said about you is correct.
How do I increase my AI visibility?
Make the business a clear entity with consistent details everywhere, write factual service pages an answer can quote, add structured data, keep pages crawlable and fast, be listed accurately on the third-party sources AI answers rely on, collect genuine reviews, and earn mentions on trusted sites. Then rescan to confirm the effect.
Does ranking well on Google mean I am visible in AI answers?
Not automatically. AI systems retrieve from search, so ranking helps, but shortlists are assembled from directories, reviews and comparison content as much as from your pages. A business can rank on page one and be absent from AI answers, and the reverse also happens.
Should I block AI crawlers?
If you want to be cited, no. Blocking the crawlers used for AI search features removes your pages from retrieval. Businesses that block for content-protection reasons should at least allow the crawlers tied to the search features their customers use, and check the robots file after any security or hosting change.
How long does it take to improve AI visibility?
Data corrections and page rewrites can show in AI answers within weeks as pages are re-crawled. Third-party listings and reviews take one to three months to accumulate. Comparison and trust prompts move last. Rescan monthly and expect steady movement rather than an overnight jump.
Is AI visibility only relevant to big brands?
No. Local and service businesses are the most affected, because their customers ask location-specific, trust-heavy questions that AI answers now resolve with a short list of names. Small businesses with complete profiles, clear pages and steady reviews routinely appear ahead of larger firms in local prompts.
Get your AI visibility scan
Send your website and the three or four questions you most want customers to find you for, and you will get a written summary of how AI answers currently treat your business: whether you are mentioned, who appears instead, what is wrong in the descriptions and which fixes come first. No score without the prompt sheet behind it, and no obligation.
Send the details through the contact form, or message me on WhatsApp at +92 309 6405463.