Book Free Audit
Free tool · no signup

The free AI visibility check you run yourself

Seven tests. About fifteen minutes. Nothing to fill in. Each one uses a free public tool and tells you exactly what a pass looks like, what a failure looks like, and what to do first if it fails. Run them on your own business and keep the answer — you never have to talk to us at all.

No form. No email address. No account. · Works for any local service business
Serving the GTA from Richmond Hill Audits returned in ~5 business days Free audit · No commitment · No upsell AI visibility · Maps · Organic · Content architecture Honest timelines · No guaranteed-ranking promises Built for tradespeople and home service operators
// what you are testing

What exactly is this checking for?

AI visibility is whether an AI assistant names your business when somebody asks it for a local recommendation in your category and area.

That is a narrower question than "how is my SEO", and it is answerable on a Tuesday afternoon without buying anything. The seven tests below split it into two halves: what the assistants currently say about you, and whether the public record they read from is clean enough for them to say anything confident at all. The wider mechanics of how that record gets built are covered on the AI visibility services page — this page is only the diagnostic half.

It is worth knowing how narrow the opening is before you start. The SOCi 2026 Local Visibility Index, which analysed nearly 350,000 locations across 2,751 multi-location brands, found that only 1.2% of those locations were recommended by ChatGPT. That is a study of chains, not independents, so read it as an indication of how narrow the opening is rather than a figure about your own business. A local recommendation query comes back with three or four names, not thirty, and the same handful tends to reappear.

None of that means the shortlist is closed. It means the entry requirement is being legible: one name, one number, one description of what you do and where you do it, repeated in enough places that a machine has no reason to hesitate. Most of the seven failures below are failures of legibility rather than failures of quality, which is why they are worth finding before you spend money on anything.

A machine will not choose between two versions of your business. It will choose the business that only has one version.

// two minutes of setup

Set up so the results mean something

Skipping this makes the whole exercise less reliable, because assistants personalise heavily and because a half-remembered business name will produce a false failure on three separate tests.

  1. Open a private or incognito windowLog out of everything. A signed-in session carries your history, your saved places, and in some tools a memory of previous conversations. All three quietly bias the answer toward businesses you have already looked at, including your own.
  2. Write down your exact registered nameCharacter for character, including the legal suffix and any punctuation. Several of these tests are exact-match searches, and a name typed from memory will fail against a listing that is actually fine.
  3. List the areas you genuinely serveNot the aspirational radius — the places you actually take work in. You will compare this list against what your own site and profiles currently claim, and the gap between the two is often the finding.
  4. Keep one page of notesSeven results is more than you will hold in your head, and the value of this exercise is the pattern across them rather than any single answer. A pass with two adjacent failures usually points somewhere different than the same pass in isolation.
// the seven checks

Seven tests, with pass and fail spelled out

Work through them in order. Each card gives you the exact thing to do, then what a clean result looks like, what a broken one looks like, and the single first move if it comes back broken.

Check 01

Ask the assistants for a recommendation and watch what comes back

Run in ChatGPT, Gemini and Perplexity — three phrasings each:
"Recommend a [your trade] in [your area]"
"Who should I call for [the job you do] near [your area]?"
"Best [your trade] [your area]"

PassYou are named in at least two of the three phrasings on at least one assistant, with the right service and the right area attached to your name.
FailNothing comes back with your name on it. A single appearance that disappears the moment you reword the question is also a fail — that is a loose association, not a position. So is being named with the wrong service, the wrong town, or a number you do not answer.
Check 02

Confirm you exist properly on Microsoft's business index

bingplaces.com → search your exact registered name
Then: bing.com → search the same name in quotes

PassExactly one listing, claimed and verified, with the right category, the right phone number, and a service area that matches the list you wrote down during setup.
FailNo listing at all, an unclaimed listing someone else's data was scraped into, two listings for the same business, or a listing carrying an address you moved out of.
Check 03

Look for structured data in your own homepage source

Load your homepage → right-click → View page source
Ctrl+F (Cmd+F) → search for: schema.org

PassAt least one JSON-LD block containing LocalBusiness or a more specific subtype for your trade, carrying your name, your phone number, and either a street address or a list of areas served.
FailNo match anywhere in the source. Or — quietly worse — a block that is present but stale: an old number, a service you stopped offering, or a service area that does not resemble the one on your visible page.
Check 04

Make sure the crawlers are allowed to read you at all

yourdomain.com/robots.txt
Look for: GPTBot · PerplexityBot · ClaudeBot · Google-Extended · bingbot
Then: view source on a service page and confirm your service copy is in the HTML, not injected later by script

PassNo blanket disallow rule, no rule naming those agents, and your actual service and location wording is visible in the raw source rather than appearing only once scripts have run.
FailA disallow you did not knowingly add — often left over from a staging site or added by a privacy plugin. Or a page whose source is nearly empty because the content is assembled in the browser.
Check 05

Count the independent sources that confirm you exist

Search your exact name on the places that matter in your trade — review platforms, trade directories, your industry association, your chamber, a licensing or certification register

PassYou appear on three or more independent sources beyond your own website and your Google profile, each one claimed, each carrying the same name and the same number.
FailEverything traces back to one profile. A business that only exists in one index reads as unconfirmed, no matter how good that single profile is. Sources that exist but disagree with each other score worse than sources that do not exist.
Check 06

Sweep for the old versions of you that are still online

google.com → "Your Exact Business Name" — read three pages of results
Then: google.com → "your phone number" in quotes

PassEvery result across both searches tells the same story: one trading name, one number, one description of what you do. Nothing on those pages surprises you.
FailA unit number from a previous shop. A line you disconnected. A name you traded under before the rebrand. A duplicate profile from a directory that scraped you in 2019 and has never been touched since. Each one is a small contradiction, and they compound.
Check 07

Read your own homepage the way a machine would

Open your homepage and find one sentence that states, in plain words, what you do and where you do it — without a slogan, without a metaphor, and without needing the picture beside it

PassSuch a sentence exists, it is real text rather than words baked into an image, and it sits near the top of the page. A stranger reading only that sentence could tell you the trade and the territory.
FailThe clearest statement on the page is a tagline like "quality you can trust since 1998", or the service list is a graphic, or the geography is implied by a photo of a local landmark and never actually written anywhere.
// reading your result

What does your tally actually mean?

Count the failures, not the passes. The number tells you which of three very different situations you are in, and the right next move is different in each one.

0

Nothing broken

Your record is legible and consistent. Whatever is holding you back is comparative rather than technical, and no amount of tidying will move it. Re-run the check next quarter and spend your effort on published depth instead.

1–2

Ordinary gaps

The common result, and the easiest to act on. One or two specific things are missing or stale. A motivated owner can close both over a couple of evenings without hiring anyone, then wait six weeks before re-testing.

3+

Structural

Failures that cluster usually share one cause — a rebrand that was never propagated, a move, a site rebuild that dropped things. Fix the shared cause rather than the seven symptoms, or you will do the same work twice.

One thing the tally cannot do is tell you whether the failures you found are the failures that matter most for your particular category. That ordering question is what a proper local SEO audit exists to answer, and it is the one part of this that genuinely does not compress into a fifteen-minute page.

Found three or more failures and not sure which one to start with?

// where this check is wrong

What these seven tests cannot tell you

This section is not a disclaimer. It is the part that makes the rest of the page usable, because a diagnostic you trust too far is more expensive than no diagnostic at all.

  • It is a sample, not a measurement

    Assistants do not return the same answer twice. Ask the same question an hour apart and the names can shift. Three phrasings across three tools is enough to see a pattern and nowhere near enough to call a result definitive. Treat one absence as weak evidence and a consistent absence across every phrasing as strong evidence.

  • Where you are sitting changes the answer

    A private window strips your account history but not your location. Running the check from your own premises biases every result toward businesses near you, including yours. That is the single most common way this check produces a falsely comfortable pass.

  • The answers lag their sources by weeks or months

    Claiming a listing today changes nothing today. Some systems read live web results and update within weeks; others lean on material absorbed much earlier and move slowly and unevenly. A re-test the following morning is measuring noise, not progress.

  • A clean sweep is not a promise of anything

    Passing all seven means your record is readable. It does not mean an assistant will name you, and it does not mean the phone will ring. Nobody can guarantee either — not us, not anyone selling you a package that says otherwise. These systems have no published ranking rules and no scoreboard to move.

  • It is blind to the things that need volume to see

    Whether another company trading under a name close to yours is quietly absorbing mentions meant for you. How each assistant describes what you do when asked in a hundred different ways rather than three. Whether your pages agree with each other about geography across the whole site. What the businesses named instead of you have published that you have not. None of that is visible from a spot check, because none of it shows up until you run the same test at scale and compare the sources behind each answer.

  • It can fail you for a reason that does not matter

    Some trades genuinely have no meaningful directory presence, and a thin result on check 05 means less there than it would elsewhere. A brand-new business will fail several of these simply for being new. Read the failures in the context of your own trade rather than as a score out of seven.

// what to do next

The order to fix things in, and why that order

If more than one check failed, sequence matters more than speed. Each step below assumes the one above it is already done, because doing them out of order means redoing work.

  1. Unblock accessNothing else can be read while a crawler is being turned away, so any failure on check 04 comes first. It is usually a one-line change and it is the only item here that can nullify all the others on its own.
  2. Settle on one version of the truthDecide the exact name, number, and area list you will use everywhere, and write it down. Every later step copies from this. Skipping it is how businesses end up with four slightly different records they created themselves.
  3. Remove the contradictionsWork through the sweep from check 06, starting with whatever is most widely linked. Correcting old records before adding new ones stops you from propagating a problem you have not resolved yet.
  4. Claim the missing listingsMicrosoft's index first, then the two or three sources that carry weight in your specific trade. Copy the agreed wording exactly rather than retyping it, since retyping is where drift starts.
  5. Make the site say it plainly, then mark it upFix the visible sentence from check 07 before touching the structured data from check 03. Markup describes the page; if the page is vague, correct markup only describes vagueness accurately.
  6. Wait, then re-test properlySix weeks minimum, then run all seven again from a private window and compare against your notes. If the result has not moved and the record is genuinely clean, the remaining gap is comparative — and that is the point where an outside read on how you compare to the businesses being named is worth having.
// straight answers

Questions people ask before running it

No. There is no form, no email address, and no account anywhere in the seven tests. Every one of them uses a browser tab and a free public tool you already have access to. Write your results on paper if you like. The optional full audit at the bottom of the page is a separate thing you can ignore entirely and still walk away with a usable answer.
Once a quarter is plenty for most local businesses, plus one extra run about six weeks after you change anything. Running it weekly mostly produces noise, because assistant answers vary between sessions even when nothing about your business has changed. Give any fix time to be crawled, indexed, and absorbed before you judge it.
That is a normal outcome in a busy category, and it is useful information rather than a dead end. Passing means your public record is clean, readable, and free of contradictions. It does not mean your published footprint is deeper than the businesses currently being named. The gap at that point is usually comparative, not technical, and it lives in things a spot check cannot see: how consistently you are described across sources, whether a near-identical trading name elsewhere is muddying your record, and what the assistants say about the businesses they pick instead.
Yes, and the checks were written with that case in mind. A business that drives to its customers usually does not want a street address published anywhere, which is fine. What matters is that the areas you cover are stated in plain words on your own site, listed the same way on every profile you control, and reflected in your structured data. A hidden address is not a problem. A vague or contradictory service area is.
Not on its own. A strong profile is the single most valuable listing you can hold, and it feeds Google's own surfaces first. Several assistants are built on other indexes and on a general crawl of the open web, so a business that exists properly in one place and nowhere else reads as a single unconfirmed source. Breadth of agreement is what raises confidence, not depth in one listing.
Yes. Viewing source works on every platform, hosted or self-built, and the search for the phrase schema.org takes a few seconds. Adding or correcting the markup afterwards is where platforms differ. Some builders emit a basic block automatically, some need a plugin, and some require pasting a snippet into a header field. The check itself is platform-independent.
For drafting a structured data block, that works reasonably well as a starting point. Read it against your live page before publishing it. The common failure is confident markup that describes a business slightly different from the one on the page — an area you do not cover, a service you dropped last year, an old phone number. Markup that contradicts the visible page is worse than no markup, because it gives a machine a documented reason to distrust everything else on the site.
Because a contact form is a bad first step for someone who only wants to know whether they have a problem. Most owners who land here are not ready to talk to anyone, and asking them to book a call before they have any information is the wrong order. If the seven checks come back clean, you have saved yourself a phone call. That is a fair trade.
It costs nothing and it is returned in about five business days. It repeats this ground properly rather than as a spot check — many queries instead of a handful, across several assistants, with the sources traced back to whatever each answer was built from. It also covers the parts you cannot self-test: whether a close trading-name match elsewhere is absorbing mentions meant for you, how each assistant describes what you do when asked directly, whether your pages reinforce or contradict each other on geography, and how your published record compares with the businesses being named in your place. You can start it by calling (437) 799-0606 or emailing [email protected].
Optional · free · ~5 business days

Want the version that does not fit on a page?

The seven checks above are yours to keep either way. If you would rather have the same ground covered at volume — many queries across several assistants, with the sources behind each answer traced and the gaps ranked in the order worth fixing — that is the full audit. It costs nothing and comes back in roughly five business days. If we cannot help, we will say so.

No commitment · No upsell · Written report
[email protected] · Serving the GTA and Barrie from Richmond Hill, Ontario