For a single-site local business, three kinds of AI visibility product show whether ChatGPT and Google's AI recommend you: your website builder's own tracker (Squarespace has one), a local-search tracker such as BrightLocal's, or a prompt tracker such as Otterly. Semrush and Ahrefs suit SEO teams more than a single shop.
Before paying anyone, spend an afternoon running the check by hand, because the tools automate it and nothing more. Then pick the product by two tests: does it track the assistants your customers use, and does it understand that you are a place with an address, not a brand with a worldwide audience?
What an AI visibility product actually does
An AI visibility product takes a list of questions (called prompts), asks them of one or more AI assistants on a schedule, and records which businesses and websites appear in the answers. It then reports whether you were mentioned, where in the answer, and which competitors beat you. That is all. It does not edit the answers, and it cannot see what real customers typed.
Two limits follow from how it works. First, the prompts are your guess at customer questions, so a tracker is only as good as the list you give it. Second, assistants answer differently each time, so any single result is a sample, not a verdict. Good products repeat each prompt and show a share. If a product shows you a single score without saying how it was calculated, treat it as a rough indicator.
Start with a free manual check
An example with a car repair garage with one site. This is an illustration, with made-up results.
- Write 20 prompts. Mix three kinds: category questions ("best garage for a brake service"), problem questions ("my car pulls to the left when braking, who can fix it") and branded questions (the garage's own name plus "reviews"). Add your town to the category and problem prompts.
- Run each prompt three times in each assistant you care about, in a fresh chat with no history. ChatGPT, Gemini and Google's AI Mode are the sensible three.
- Record the result in a sheet: prompt, assistant, run number, named (yes or no), position (first, middle, last or only), and who else was named.
- Score it. For each assistant, divide the number of runs that named you by the total runs.
Say the garage is named in 14 of 60 ChatGPT runs (23%), 9 of 60 Gemini runs (15%), and 21 of 60 AI Mode runs (35%). Looking closer, the garage appears in almost every branded prompt and almost none of the problem prompts. That is a real finding: customers who already know the name find it, and customers with a problem do not. No product was needed to see that, and it points straight at the work to do (pages and listings that describe the problems you fix).
Keep this sheet even after you buy a tool. It is your check on the tool: if the tracker says 40% and your sheet says 15%, find out why before believing either.
A filled-in prompt list for the garage
The quality of the list decides everything, so here is a realistic extract for the garage. Notice the mix and the plain wording.
| Type | Prompt (illustrative) |
|---|---|
| Category | Which garage should I use for a brake service in [town]? |
| Category | Independent garage in [town] that does annual safety inspections and servicing on older cars |
| Problem | My car shudders when I brake at speed, who can fix it in [town]? |
| Problem | Warning light came on after filling up, where can I get it diagnosed quickly in [town]? |
| Price | How much does a brake pad replacement cost at a garage in [town]? |
| Trust | Is [garage name] a good garage, and what do customers say? |
Twenty prompts of this shape take about half an hour to write. Add six or seven you have heard customers actually say at the counter; those are worth more than anything invented at a desk.
Recording one run: a sample answer and the row it produces
Here is an illustrative answer to the brake-judder prompt, the kind of thing an assistant returns, followed by the row it becomes in the sheet.
Prompt: My car shudders when I brake at speed, who can fix it in [town]?
Answer (illustrative): Brake shudder is often caused by warped discs
or uneven pad wear. Garages in [town] that customers mention for brake
work include [competitor A], which is often praised for fair pricing,
and [competitor B], a chain centre with same-day slots. Check recent
reviews and ask for a quote before agreeing to replace discs.| Prompt | Assistant | Run | Named? | Position | Others named | Sources shown |
|---|---|---|---|---|---|---|
| Brake shudder, [town] | ChatGPT | 2 | No | n/a | Competitor A, competitor B | A review site, competitor A's own page |
The useful column is the last one. The answer leans on a review site and on a competitor page that explains brake judder in plain words; the garage has neither a strong presence on that review site nor a page on the symptom. Write down the sources every time, because after 60 runs they tell you where to spend effort far more clearly than the yes/no column does.
The products, one by one
I checked each product's own pages in October 2026. Prices are deliberately left out of this comparison because they shift often and several vendors show different figures by billing period, currency or location; each section says how the product is sold so you know what to look for.
Squarespace AI Visibility
If your site already runs on Squarespace, this is the easiest start because it sits in the same dashboard as your site. Squarespace's own page says it analyses responses from ChatGPT and Gemini, and that it is currently available in English only. For each prompt it shows whether your site is mentioned, where you appear, whether that position has changed, and which competitor brands and source websites the AI includes. Prompts are generated for you and you can edit, add or delete them. The page shows an overall visibility score without explaining how it is calculated, says every plan has an allocation of AI credits that depends on the plan, and does not state a refresh schedule, so re-running prompts appears to be manual.
It suits: a single site that is already on Squarespace, an English-speaking audience, and an owner who wants a first view without another login. It does not suit: a site on another platform, a non-English audience, or anyone who needs Google AI Overviews or AI Mode coverage, which the page does not claim.
BrightLocal Local AI Visibility Tracker
BrightLocal is a local-search specialist, and its pricing page labels a Local AI Visibility Tracker as new, described as showing how your business appears across ChatGPT and Google's AIs. It is included from its lowest plan (Track) and carries up into the higher ones, priced for one location, with other tiers for multiple locations. The page did not detail what it measures beyond that, so ask for a demo account or a sample report. Its footer also lists a free Local AI Visibility Checker; I could not read its details, so test it yourself.
It suits: a business with a physical location, particularly one that already cares about listings and reviews, because the tracker lives beside that data. It does not suit: a business that needs per-prompt control across many assistants.
Otterly
Otterly is a prompt-tracking tool sold by the number of prompts. Its pricing page shows plans with prompt counts rising from 15 to 400, and enterprise above that. ChatGPT, Google AI Overviews, Perplexity and Microsoft Copilot are included in all plans; Claude, Google AI Mode and Gemini are listed as add-ons. There is a free trial, but the page does not state how long it lasts. Prompt-based pricing means you control cost: a single-site business with 15 to 20 prompts sits at the bottom of the range.
It suits: an owner who wants to choose assistants and prompts, and does not mind a tool designed for marketers. It does not suit: someone who wants local features such as review and listing data.
Semrush
Semrush builds AI visibility tracking (AI search performance, sentiment and custom prompts) into its paid toolkits rather than selling a cheap standalone tracker, and the tier structure changes often, so check its pricing page for which plan carries prompt tracking and how many prompts a day it allows. That makes it a sensible choice for a business already paying for Semrush's SEO tools, and a heavy one otherwise. For an agency view see Semrush AI Visibility against Ahrefs Brand Radar.
Ahrefs Brand Radar
Ahrefs splits Brand Radar in two. Its AI Visibility Index covers AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot; Custom Prompts, where you write the buyer questions yourself, covers the same assistants plus Claude, with a small daily prompt allowance included on paid Ahrefs plans and larger check-based packages on top. The sign-up link suggests Brand Radar is bought as an add-on to an Ahrefs plan, so confirm the full cost on the live page. It is aimed at brands treating AI search as a revenue channel; a single-site local business is a stretch fit.
Which product for which kind of business
| Business | Start with | Why |
|---|---|---|
| One site on Squarespace, English-speaking customers | Squarespace AI Visibility | Already in the dashboard; covers ChatGPT and Gemini |
| One location, listings and reviews matter most | BrightLocal tracker | Local focus, one-location pricing |
| Any site, want assistants chosen by you | Otterly | Prompt-based plans, wide assistant list |
| Already pays for Semrush or Ahrefs | Their own AI add-ons | One login, shared data |
| Under 20 prompts, tight budget | The manual sheet | Free, and a baseline for any tool |
What to compare before paying
- Assistants covered. Make sure the ones your customers use are included. Several products list Google AI Mode or Gemini as an extra.
- Languages and markets. Squarespace's tracker is English-only for now. A business with customers in other languages needs a different tool or the manual check.
- How a result is calculated. Repeated runs and a share are better than a single run. If a vendor will not say, discount the score.
- Location handling. Does a result reflect a person standing in your town, or a generic location? Ask for an example.
- Prompt limits and who writes them. Auto-generated prompts are a decent starting point; your own customer questions are better.
- Export. Can you download results to compare with your sheet?
- Trial. A trial lets you check the tool against your manual numbers before you commit.
A worked comparison with numbers
An illustrative case: a specialist food importer with one website sells to cafes and restaurants within a few hours' drive. It wants to know if ChatGPT and Google's AI name it when a chef asks for a supplier of a particular product.
It starts with the manual check: 20 prompts, three runs, two assistants, so 120 runs. Result: named in 18 runs (15%), nearly all on prompts that mention a product it is known for. It then trials a prompt tracker with the same 20 prompts and compares. The tool reports 20% for the same assistants. The gap, five points, is within what run-to-run variation explains, so the importer trusts the tool's direction, not its decimal.
The important decision comes next. The monthly cost of a tracker is not the main expense; the main expense is acting on the report. The importer reads the competitor and source lists: the sources the assistants quote are two trade directories and a food-service forum, none of which mention the importer. The action is to be listed in those directories with accurate details, which costs time, not software. That is the general pattern: a tracker identifies which sources the assistants lean on, and the job is to be present and accurate on them. How to get your business recommended by ChatGPT and AI search covers that work.
Reading the report: what to do with each finding
A tracker report, or your own sheet, yields four kinds of finding. Each one leads to a different job, and only the first needs the tool.
- You are named, but with wrong details. Say the garage is described as open on Sundays when it is not. Fix the source: the business listing, the website opening-hours text, and any directory entry. Then re-run the prompt in two or three weeks. Accuracy beats volume.
- You are absent on problem prompts. The garage has no page describing brake judder or warning-light diagnosis. Write one plain page per common problem, with the symptoms, what you check, a price range and how to book.
- A competitor is named on every category prompt. Look at the sources cited beside that competitor. If it is a review site or directory, check your presence and reviews there; if it is the competitor's own page, read what it answers that yours does not.
- The sources are sites you are not on. Claim the listing, complete it, and ask recent customers for honest reviews there. Do not buy reviews, and do not mass-produce pages for every suburb.
Most owners who run the manual check once end up with a to-do list of three or four of these. That list is the real deliverable, and a tracker earns its keep only by telling you, month by month, whether the list is working.
A monthly routine that takes under an hour
- Week one of the month, 20 minutes. Re-run the same prompts, three runs each, in the assistants you track (or let the tool do it and spot-check five).
- 10 minutes. Update the sheet and compare with last month for each prompt type, not the whole average. A rise on problem prompts is the signal you want.
- 15 minutes. Read two or three full answers aloud. Look for wrong facts, outdated prices and missing services.
- 10 minutes. Pick one fix for the month. One page improved, one listing corrected, one batch of review requests sent.
Keep the prompt list stable for at least three months so that changes in the numbers mean something. If you change prompts, note the date and treat that month as a new baseline.
The 100-prompt plan that reported a misleading 4%
An illustrative one: a single-site furniture maker bought a tracker plan with 100 prompts because the bigger number looked safer. Most of the prompts were auto-generated and generic ("best furniture makers"), the tracker reported a visibility score of 4%, and the owner concluded AI was hopeless. A colleague rewrote 20 prompts in the words customers use ("oak dining table made to measure, seats eight, who makes them"), and the score on those was 30%.
The lesson is that the score measured the prompts, not the business. Before paying for more prompts, write better ones. A cheaper plan with a well-chosen list beats an expensive plan with a vague one, which is another reason to start with the manual sheet.
Things that make a tracker mislead you
- Prompts nobody types. A prompt like "top-rated multi-service automotive repair provider" is not how a driver asks. Use plain phrasing from real enquiries.
- One run, one conclusion. A single mention is luck; a rate over many runs is evidence.
- Branded-prompt inflation. If most of your prompts include your name, your score will be high and meaningless. Keep branded prompts to about a fifth of the list.
- Treating the score as a ranking. A visibility score is not a position in a list. AI answers have no fixed order across users.
- Ignoring wrong information. A tracker may show the assistant naming your business with the wrong hours or services. Fixing that (listings, website, reviews) is worth more than raising a score.
Google's own guidance says AI Overviews and AI Mode need no special optimisation beyond normal search practice, and that you do not need new machine-readable files, AI text files (such as llms.txt) or special markup to appear in them. No tracker changes that, which is a good reason to keep a tool's cost proportionate. If you want to measure customers, not mentions, how to track whether AI search is sending you customers is the companion tutorial, and for Google's AI Overviews alone see tracking software for AI Overviews.
If you want help building your prompt list and reading the results, that is what my AI implementation consultation covers; start from the contact page. The AI hub lists related tutorials.
Questions about tracking AI recommendations
Can I check my AI visibility for free?
Yes. Write 15 to 20 questions a customer would ask, run each one in ChatGPT, Gemini and Google's AI Mode, and record whether your business is named. Repeat every month. It is slower than a tool but costs nothing and teaches you what the answers look like.
Why do the same questions give different answers each time?
AI answers are generated fresh and vary with wording, location settings and the model's own randomness. That is why serious tools run each prompt repeatedly and report a share, not a single yes or no. Do the same by asking each question three times.
Does a visibility tool improve my ranking in AI answers?
No. A tracker measures; it does not change what an assistant says. Google's guidance says AI Overviews and AI Mode need no special optimisation beyond normal search practice. Improvement comes from accurate listings, reviews, clear pages and being mentioned by sources the assistants read.
Is one tool enough if I serve several towns?
Check how the tool handles location. Many track one set of prompts, not results as they appear in each area. If you serve several areas, add the area to the prompt itself and confirm whether the tool supports more than one business location before paying.
Further reads
- AI SEO for Plumbers: Get Named by ChatGPT and Google AI Overviews — The fixes worked through for one local trade.
- How AI Overviews Are Changing Local Search for Small Businesses — How AI Overviews are changing local search.
- AI Visibility for Hotels: Get Recommended in ChatGPT Answers — The same question answered for a hotel.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Squarespace, BrightLocal, Otterly, Semrush and Ahrefs product and pricing pages, and Google's guidance on AI features in Search, checked 27 October 2026.