How to measure your AI visibility without paying for a tool
Short answer: You can check whether ChatGPT and other AI assistants mention your business without buying a tracker. Start with five genuine customer questions, ask them on a few assistants and record which brands are named, and whether any is clearly recommended. Then check AI referrals in Google Analytics 4 and use Bing Webmaster Tools for Microsoft’s citation data. These methods measure different things, so keep their results separate.
A paid tracker becomes useful when you need the same measurement repeated every week, across several engines, against named competitors.
Start here: a five-minute check
- Write five questions a real customer would ask before choosing a business like yours. Use their words, not your service names, and leave your brand name out.
- Ask each one in ChatGPT with web search on, and in at least one other assistant such as Gemini, Claude or Perplexity.
- Log every answer in a simple table: the date, question, assistant, whether you were named, who else was named, and whether your site was linked. Download our free answer log template (CSV), which opens in Excel or Google Sheets.
- Ask each question a second time. If the two answers disagree, you have learned something important: one reading is not a result.
- Note the conditions: signed in or not, search on or off, and roughly where you are. Next month you will want to repeat the check under the same conditions.
That is enough to answer the first question most businesses have: does AI name us at all, and who does it name instead?
A worked example (illustrative)
This is a constructed example, not a real business or a real answer. A boiler installer in Leeds writes down “Who fits a new boiler in Leeds?” and asks it twice in ChatGPT with search on, signed out.
| Run | Installer named? | Competitors named | Link to installer’s site? |
|---|---|---|---|
| 1 | No | Competitor A, Competitor B, a national comparison site | No |
| 2 | No | Competitor A, Competitor C | No |
Two runs, two different lists, and the installer is in neither. That is a useful finding: Competitor A appears both times and is worth studying. It is not yet a “rate”. Two readings of one question cannot tell you how often the installer would be named across the questions its customers actually ask.
What exactly are you measuring?
“AI visibility” covers several different events. Mixing them up is a common mistake in free and paid reports alike.
| What | What counts | Where you see it free | What it does not tell you |
|---|---|---|---|
| Mention | Your brand appears anywhere in the answer | Your own answer log | Whether you were recommended, or linked |
| Recommendation | You are put forward as a choice on a question that did not name you | Your own answer log, judged by you | How often across all buyer questions |
| Source citation | The answer links to your site, or names you as its source | Your answer log; Bing AI Performance for Microsoft surfaces | Whether anyone clicked |
| Referral visit | Someone clicked from an assistant to your site | GA4 | How often you were shown without a click |
| Website readiness | Whether AI crawlers can reach and understand your site | A free website audit | Whether any assistant actually names you |
Our guide to AI mentions, citations and recommendations defines each one precisely, including the awkward cases.
Method 1: Ask the assistants yourself
This is the most direct free method. Done carefully, it is also more useful than many people expect.
- Use buyer questions without your name. “Who are the best accountants for small charities?” tests whether you are recommended. “Is [your firm] any good?” only tests whether the engine has heard of you.
- Repeat each question. AI answers vary from one run to the next. Two or three runs per question shows you how stable an answer is.
- Record the conditions. Whether web search is on, which model or mode answered, whether you were signed in and where you are can all change the answer. A private or signed-out window reduces the effect of your own history, but it does not make the other conditions the same.
- Judge recommendations honestly. Being listed in a neutral list, being endorsed, being mentioned negatively and being quoted as a source are different outcomes. Note which one you saw.
What it cannot tell you: a stable rate or a trend. A handful of runs is too small a sample. If next month’s check looks different, you cannot tell whether your visibility changed or the answers simply varied. We explain that variation in why AI models give different answers. If ChatGPT is the assistant you care about most, our guide on how to track ChatGPT brand mentions week to week builds on the same manual method.
Our free AI recommendation check runs one version of this for you: it asks one real customer question across ChatGPT, Gemini, Claude, Perplexity, Grok and DeepSeek and shows whether you, or a competitor, are named. It is free and needs no card.
Method 2: Check AI referrals in Google Analytics 4
GA4 can show visits that arrive from AI assistants, such as chatgpt.com or perplexity.ai, as referral traffic. Our GA4 guide walks through the report and the filter to use.
What it tells you: that assistants are sending real visitors, which pages they land on, and whether those visitors enquire or buy. Set up a conversion event for enquiries if you have not already, so you can see what AI visits are worth.
What it cannot tell you: how often you are mentioned or shown. GA4 only sees visits where someone clicked through and the referrer survived. Many answers produce no click. Some apps and browsers send no referrer, so a visit from an assistant can land in Direct.
Treat GA4 AI referrals as a floor, never a total. Do not divide them by anything to make a “citation rate”: the two count different events.
Method 3: Read Bing Webmaster Tools’ AI Performance report
Microsoft’s AI Performance report in Bing Webmaster Tools, launched in public preview on 10 February 2026, shows “how publisher content appears across Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations” (Bing Webmaster blog). It reports citation counts, cited pages and the grounding queries behind them. On 16 June 2026 Microsoft said four more views were “beginning to roll out in preview” (Bing blog).
What it tells you: real citation data from Microsoft’s own AI surfaces, for your verified site, free.
What it cannot tell you: anything about ChatGPT, Gemini, Claude or Perplexity, which Microsoft does not say the report covers. Microsoft also says the count “reflects how often pages are cited, not page importance, ranking, or placement”. It is a preview, so expect changes.
Setup and interpretation are covered step by step in our Bing Webmaster Tools AI Performance guide.
Method 4: Check your website is ready to be named
The three methods above observe outcomes. A website audit checks conditions on your side: whether AI crawlers can reach your pages, whether your content answers questions clearly, and whether your site states plainly who you are.
The free SearchScore audit checks 150+ AI visibility signals across 8 categories and gives you a Fix-List of what to change. It reads your own site. It does not ask any AI engine anything, so it tells you how ready your site is, not whether you are being named. A good readiness score does not guarantee mentions; our guide to reading your audit findings goes into that.
If you only want to know whether ChatGPT’s crawlers can reach you, the free ChatGPT visibility checker covers that, and our guide to GPTBot vs OAI-SearchBot explains which crawler matters. To check whether ChatGPT can find your website step by step, see our ChatGPT visibility guide.
How do the free methods compare?
| Method | Measures | Engines covered | Repeatable over time? |
|---|---|---|---|
| Manual answer log | Mentions, recommendations, links | Any you test | Only if you keep the questions and conditions fixed |
| GA4 AI referrals | Clicks from assistants | Those that pass a referrer | Yes, for the clicked subset |
| Bing AI Performance | Citations of your pages | Microsoft Copilot, Bing AI summaries, select partners | Yes, within Microsoft, while in preview |
| Free website audit | Your site’s readiness | None queried | Yes, for your own site |
When does a paid tracker become useful?
When the question changes from “where do I stand?” to “is it changing, and against whom?”. Typical signs:
- You need comparable readings over time. Manual checks drift: different days, different settings, different people. A tracker asks the same questions the same way every week.
- You want competitor movement. Logging who else is named, across every question and engine, every week, is slow by hand.
- You need several engines. Ten questions, asked twice, on six assistants is 120 answers to read and log each time.
- You report to someone. A client, a board or a manager needs a number measured the same way each period.
A tracker reduces the manual work and keeps the method consistent. It does not remove the variation in AI answers, so any good one should tell you how much a number can move by chance.
What does the SearchScore Tracker add?
The SearchScore Tracker asks the questions customers ask before choosing, across 6 AI engines (ChatGPT, Claude, Gemini, DeepSeek, Grok and Perplexity) every week. Its headline is a mention rate: how often you are named across your questions. It also shows who is named instead and how that moves over time.
| Plan | Price | Questions | Asks per question each scan | Competitors tracked |
|---|---|---|---|---|
| Starter | £49/mo | 10 | 2 | 5 |
| Pro | £99/mo | 25 | 2 | 10 |
| Scale | £199/mo | 50 | 2 | 15 |
| Enterprise | £499/mo | 100 | 3 | 20 |
Prices in pounds. See Tracker plans for your currency and current terms.
A sensible order to do this in
- Run the free audit and fix anything stopping AI crawlers reaching your site.
- Ask your five customer questions by hand, twice each, on at least two assistants, and log the answers.
- Set up GA4 AI referrals and Bing Webmaster Tools so a free baseline runs in the background.
- Repeat step 2 in a month under the same conditions.
- If you need a trend, more engines or a competitor comparison, consider a tracker.
When you are ready to go further, our guide to the full three-layer measurement method sets out how to combine a readiness audit, regular question testing and referral traffic into one routine.
Sources
- Microsoft Bing Webmaster blog, “Introducing AI Performance in Bing Webmaster Tools Public Preview”, 10 February 2026: https://blogs.bing.com/webmaster/February-2026/Introducing-AI-Performance-in-Bing-Webmaster-Tools-Public-Preview (checked 12 November 2026)
- Microsoft Bing blog, “New AI Visibility Insights in Bing Webmaster Tools: Intents, Topics, Citation Share, Compare”, 16 June 2026: https://blogs.bing.com/search/June-2026/New-AI-Visibility-Insights-in-Bing-Webmaster-Tools-Intents-Topics-Citation-Share-Compare (checked 12 November 2026)
Product details on this page (engines, plans, prices and limits) are rendered from SearchScore’s product data at build time. Platform details were checked on 12 November 2026.
Frequently asked questions
Can I check AI visibility without paying for a tool?
Yes. Ask AI assistants your customers' questions by hand and log the answers, check AI referrals in Google Analytics 4, read Bing Webmaster Tools' AI Performance report for Microsoft's AI surfaces, and run a free website audit. Each measures something different. Together they give a fair first picture, but not a number you can compare reliably week to week.
Why do I get a different answer each time I ask ChatGPT the same question?
AI answers vary from run to run, between engines, and with settings such as whether web search is on, which model answers, whether you are signed in and where you are. One manual check is a single reading, so ask each question more than once and log the conditions before drawing a conclusion.
Is a private browser window enough to get a neutral answer?
It helps, because your own history is less likely to shape the answer, but it does not standardise anything else. Location, model, search mode and ordinary run-to-run variation still change the result, so write those conditions down next to every answer.
Is GA4 AI referral traffic a measure of AI visibility?
It is a floor, not a total. GA4 only sees visits where someone clicked through from an assistant and the referrer survived. Many answers that name you produce no click at all, and some AI visits arrive with no referrer and land in Direct.
When is a paid AI visibility tracker worth it?
When you need the same questions asked the same way every week, across several engines, against named competitors, or you need to report a number to someone else. A tracker removes the manual work and keeps the method consistent. It does not remove the variation in AI answers.