How to Measure Your Visibility in ChatGPT, Gemini and Perplexity

You measure AI search visibility by asking the same fixed set of buyer questions in each AI assistant every month and logging whether you're mentioned, where, how accurately and who else is named. Then you add the hard data the platforms do give you: the AI Assistant channel in Google Analytics 4, the Generative AI report in Google Search Console, Bing Webmaster Tools' AI Performance report, and a "how did you hear about us?" question for every new customer.

No single tool gives you the full picture. The method below takes about two hours a month for a typical business and produces a trend you can act on.

What should you actually measure?

AI visibility has four parts: being mentioned, being mentioned early, being described correctly and being cited as a source. Traffic and enquiries come after those.

The first four are visibility. The last two are commercial results. Report both, because visibility can rise for months before the enquiries follow.

How do you build a fixed prompt set?

A fixed prompt set is a list of 15 to 25 questions you ask in exactly the same words every month, so the only thing that changes is the answer. If you change the questions, you can't compare months.

Build it from five types of question:

1. Category questions. "What are the best solar installers in Accra?" "Who does commercial cleaning in Kumasi?"
2. Problem questions. "My inverter keeps tripping, who can fix it in East Legon?"
3. Comparison questions. "Is [your brand] or [competitor] better for a small office?"
4. Brand questions. "What is [your brand]?" "Is [your brand] legit?" "How much does [your brand] charge?"
5. Decision questions. "Which solar company in Ghana offers instalment payments?"

Take the wording from real conversations. Your WhatsApp chats, sales calls and Google Search Console queries are full of the exact phrases buyers use. See keyword research for the Ghanaian market for where to find them.

Give each prompt an ID (P01, P02 and so on) and never edit a prompt once it's in use. If you need a new one, add it as a new ID.

How do you run the monthly check?

Run every prompt in every assistant you care about, on the same few days each month, under the same conditions. AI answers vary from one run to the next, so consistency in how you ask is what makes the trend meaningful.

1. Pick your assistants. Most businesses should check ChatGPT, Gemini, Perplexity, Google (AI Overviews and AI Mode), Microsoft Copilot and Claude.
2. Use clean sessions. Use a private browser window or a test account with memory and personalisation switched off. An assistant that remembers you asked about your own brand last week is not a fair test.
3. Keep location steady. Answers change by country. If your buyers are in Ghana, check from Ghana, or note the location you checked from.
4. Run important prompts twice. For your top five prompts, run each one twice in a fresh session. If you appear once out of two, log it as 0.5.
5. Screenshot everything. Save the answer with the date in the file name. Screenshots settle arguments later.
6. Log the result straight away in the tracking sheet below.

What does the tracking sheet look like?

The tracking sheet has one row per prompt, per assistant, per month. A spreadsheet in Google Sheets or Excel is enough. Use these columns:

At the end of each month, calculate three numbers:

• Mention rate: prompts where you were mentioned, divided by total prompt runs.
• Share of voice: your mentions divided by all business mentions across your prompts. If 60 businesses were named in total and 9 were you, that's 15%.
• Accuracy rate: correct mentions divided by total mentions.

The "Other sources cited" column is the most useful one for strategy. It tells you which directories, lists and articles the assistants trust in your category. Those are the places to get listed or featured. See why AI recommends brands from "best of" lists.

How does AI referral traffic show up in Google Analytics 4?

Google Analytics 4 now has a built-in AI Assistant channel. Google's Analytics Help documentation for the default channel group describes it as the channel by which users arrive "from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok", and says it sets the medium to ai-assistant and the campaign to (ai-assistant) when the referrer matches Google's list of AI assistants. Google's release notes date the feature to 13 May 2026.

Two details matter:

• It excludes Google's own AI features. Google's documentation says the AI Assistant channel excludes AI Overviews and AI Mode. Clicks from those arrive as normal Google organic traffic.
• It's not retroactive. Older AI visits stay in Referral, Direct or Unassigned.

To see it, open Reports, then Acquisition, then Traffic acquisition, and look for the AI Assistant row. Then switch the dimension to "Session source" to split it by assistant.

Should you still build a custom channel group?

Yes, if you want control over exactly which sources count. Google hasn't published its full list of recognised assistants, so a custom group lets you add any that are missing. In GA4, go to Admin, then Data display, then Channel groups, and create a new channel group. Add a channel called "AI assistants" with the condition Session source matches regex:

chatgpt\.com|chat\.openai\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai

Then drag it above Referral, because GA4 assigns each session to the first channel rule it matches. Don't add a "medium equals referral" condition: ChatGPT adds utm_source=chatgpt.com to many links without a medium, and those visits would slip past a referral-only rule.

Why does GA4 undercount AI traffic?

GA4 can only label a visit if it receives a referrer or a tag. Visits from mobile apps often arrive with neither and land in Direct. People also read an AI answer, then type your name into Google later, which shows up as branded search.

Tagging can break, too. In October 2026, Search Engine Roundtable and others reported several days when ChatGPT clicks arrived without the usual tag and were counted as Direct, until a fix rolled out. Add an annotation in GA4 when you see a sudden swap between AI Assistant and Direct, and treat AI traffic numbers as a floor, not a total.

What does Google Search Console show for AI Overviews and AI Mode?

Google Search Console counts AI Overviews and AI Mode inside your normal Performance report, and it now has a separate Generative AI performance report that shows impressions only.

• Main Performance report. Google's "AI features and your website" documentation says sites appearing in AI features are included in the overall search traffic and reported in the Performance report under the "Web" search type. There is no filter to split AI clicks from ordinary result clicks.
• Generative AI performance report. Google's Search Console Help says this report shows how many times links to your site were shown in a generative AI feature on Google Search, currently AI Overviews and AI Mode, broken down by pages, countries, devices and dates. It doesn't show clicks or queries. Google says it had rolled out to all websites worldwide as of 31 August 2026, though it may not appear if your site has had too few AI impressions.

Use the Generative AI report to see which pages Google uses in AI answers. If a service page gets AI impressions but few clicks in the main report, its answer is probably being read without a visit, which is a cue to make the next step on that page more compelling. For more on how Google's AI features affect local businesses, see Google AI Overviews in Ghana.

While you're in Search Console, check Settings for the Search generative AI control. Google's Help says the default is Include, and a site set to Exclude won't appear in AI Overviews or AI Mode. Make sure nobody has switched it off.

Does Bing Webmaster Tools report AI citations?

Yes. Microsoft added an AI Performance report to Bing Webmaster Tools as a public preview in February 2026. Microsoft's announcement says it shows how your content is cited as a source in Microsoft Copilot, AI-generated summaries in Bing and selected partner integrations.

According to Microsoft and coverage of the launch, the report shows:

• Total citations: how often your pages appeared as a source in AI answers.
• Average cited pages: how many of your pages were cited per day.
• Grounding queries: a sample of the phrases the AI used when it retrieved your content.
• Page-level citations: which URLs are cited most.

It doesn't tell you where in the answer you appeared, and it covers Microsoft's surfaces, not ChatGPT. Still, the grounding queries are gold: they show the questions AI systems are already matching to your pages. Add the best ones to your prompt set. For why Bing matters again, see where AI assistants get their information.

Why should you ask customers how they heard about you?

Asking every new customer how they found you catches the AI influence that analytics misses. Someone who read a ChatGPT recommendation, then searched your name, then messaged you on WhatsApp looks like a branded search or a direct message in your data.

• Add a required "How did you hear about us?" field to your enquiry form, with "ChatGPT or another AI assistant" as an option alongside Google, Instagram, a friend and so on.
• Add the same question to your WhatsApp welcome flow and your sales call script.
• Log the answer next to the lead in your CRM or sheet, so you can see revenue by source, not just enquiries.
• When someone says AI, ask which one and what they asked. That question often belongs in your prompt set.

See how to track sales that close on WhatsApp for setting this up properly.

What goes in the monthly report?

A one-page monthly report should show the trend for each signal, not a pile of screenshots.

Add three lines underneath: what changed, why you think it changed, and what you'll fix next month. Wrong facts come first, because an assistant describing your business inaccurately costs you sales even when it mentions you. See entity SEO for fixing them at the source.

Common questions

How often should I check AI visibility?

Monthly is right for most businesses. AI answers vary day to day, so weekly checks mostly measure noise, while quarterly checks are too slow to catch a wrong fact.

Are paid AI visibility tools worth it?

They save time once you track more than a few dozen prompts, because they run them automatically and at volume. Google's own guidance notes that no third-party tool can see inside its ranking or AI systems, so treat tool scores as estimates and keep your own prompt set as the baseline.

Why does ChatGPT give a different answer each time?

AI assistants generate answers fresh on each request and may search the web differently each time. That's why you use clean sessions, fixed wording and repeat runs for your most important prompts.

Can I see which ChatGPT prompts mention my business?

Not directly. At the time of writing, OpenAI doesn't offer site owners a report of the prompts that surface their pages. Your own prompt set, the GA4 AI Assistant channel and customer answers are the practical substitutes.

What's a good mention rate?

There's no universal benchmark. Measure your starting point, then judge progress against your own trend and against the competitors who appear in the same answers.

If you'd like us to set up this tracking and run the monthly check for you, see our AI search optimisation service. If GA4 itself needs fixing first, our tracking setup service covers it.