AI visibility tracking: how to track your brand in ChatGPT and AI answers (2026)

AI visibility tracking means asking ChatGPT, Gemini, Perplexity, Copilot and Google's AI Overviews the same buyer questions on a schedule, several runs each. You log three things: whether you're named, which pages get cited and who gets named instead. Report your share of answers across all those runs, not your place in one answer. Then add the free first-party data from Search Console's generative AI report, Bing's AI Performance report and GA4, checked for bots. Tools automate the asking. They don't change the answer.

I'm Brenden House, and I run Searchmaxxed, an SEO and AI SEO agency where AI answer tracking is part of our managed packages. Sources checked 27 October 2026 (AEST).

In summary

• Track a share, not a rank. AI answers reshuffle on every run, so measure how often you're named across many runs of a fixed question set, beside named competitors.
• Use the free official data first. Google Search Console now reports AI Overview and AI Mode impressions for every site, and Bing reports Copilot citations.
• Log the sources, not just the mentions. The pages AI engines cite are your target list, and that list is where tracking turns into real work.
• Distrust a sudden "AI traffic" spike in GA4. Check it for bots before you report it, and remember that ChatGPT's tracking tag broke for a whole week in October 2026.
• Tracking is the scoreboard. Pages, facts and sources move it.

What is AI visibility tracking?

AI visibility tracking is the regular measurement of how often AI answer engines name, cite and recommend your brand for the questions your buyers ask. People also call it LLM visibility tracking, AI brand monitoring or AI share of voice, and the method is the same under every label.

Four terms do most of the work:

• A mention is your brand named in the answer text, with or without a link.
• A citation is a link to one of your pages, shown as a source for the answer.
• A recommendation is your brand put forward as an option on a shortlist, which is the outcome you actually want from all of this.
• Share of voice is your mentions divided by all brand mentions in the tracked answers.

A prompt is the question typed into the AI tool. A prompt set, sometimes called a prompt library, is the fixed list you rerun every cycle so numbers stay comparable month to month.

Classic rank tracking checks one position on one results page. AI answers have no page two. Ask the same question twice and you'll often get a different answer, so AI visibility tracking samples many answers and reports a rate. For the background on how these engines choose their sources, read what AI SEO is and how ChatGPT picks sources.

Can you see your brand mentions in ChatGPT?

Not directly. OpenAI gives brands no report of mentions or citations in ChatGPT, so you sample the answers yourself or pay a tool to do it. Google and Microsoft are further ahead, because both of them now give site owners free first-party data about their own AI features.

Google announced its report on 3 June 2026. The help page now says: "As of August 31, 2026, we've rolled out these insights to all websites worldwide". An impression counts whenever a link to your site shows in an AI feature. No report in your account? Google says that usually means too few AI impressions so far.

Bing's AI Performance report arrived as a public preview in February 2026. The grounding queries are the gold. They're the phrases the AI searched when it pulled your page, about as close as you'll get to the question behind an answer.

Two big engines now tell you something for free, but for ChatGPT, Perplexity, Gemini and Claude, somebody still has to ask and write it down.

How do you track AI visibility, step by step?

Build a fixed set of real buyer questions and run each one several times in each engine, every month. Log who's named and what's cited. Then compare the trend with Search Console, Bing and GA4. Seven steps:

1. Write 30 to 60 buyer questions. Pull them from sales calls, tenders and RFPs, Search Console queries, People Also Ask and competitors' comparison pages, written as full sentences.
2. Group them by buying stage. Use four groups: problem questions, category questions ("best industrial sensor suppliers in Australia"), comparisons and brand checks.
3. Pick the engines your buyers use. For most B2B teams that's ChatGPT, Google AI Overviews and AI Mode, Gemini, Perplexity and Copilot.
4. Run every question at least three times per engine. Use a clean, logged-out session with search on, record the country, and keep the exact wording of every question the same.
5. Log each answer in one sheet. Record the named brands and their order, how you're described, every cited URL and the date, all in the same columns.
6. Join it with first-party data. That means Search Console AI impressions, Bing citations, and GA4 sessions and conversions from AI referrers.
7. Review monthly, judge quarterly. Read the rate across the whole question set, because any single answer is mostly noise.

A sheet with one row per answer looks like this:

Most dashboards skim those last two columns, which is a shame, because they're the ones that tell you what to build next.

How many times should you run each AI prompt?

Run each question at least three times per engine, and judge the rate across your whole question set rather than any single prompt. AI tools almost never repeat a list.

The best public test is SparkToro's January 2026 study, where 600 volunteers ran 12 prompts through ChatGPT, Claude and Google's AI 2,961 times. The finding was blunt. Rand Fishkin put the odds of ChatGPT or Google's AI giving the same brand list twice, in 100 runs, at "<1 in 100". The order was even less stable.

Frequency held up, though. In one test, 142 different human-written prompts about headphones still named Bose, Sony, Sennheiser and Apple 55 to 77% of the time. SparkToro's write-up says measuring visibility percentage is "(probably) a reasonable way" to know how prominent you are in the AI's consideration set. Position in the list told them very little.

Our own panel ran every question three times, and we read what came back as rates across the whole set, never as positions. We put 300 Australian agency-buyer questions to ChatGPT three times each, with search on. It pulled 17,870 search results into those 900 answers and kept 4,389 as sources. Being found is the easy part. About three in four results it reads never make the answer.

Now the volume. Forty questions across four engines, three runs each, comes to 480 answers a month. That's days of copy and paste. Or one setting in a tool.

What metrics should you track for AI visibility?

Track visibility rate, share of voice, citation rate, source coverage, accuracy and AI-referred conversions. Skip average position in the answer. The order changes on nearly every run, so a position average mostly measures luck, and it will swing for no reason you can act on.

Split every metric by engine. Engines lean on different sources, as the directory figures in the next section show, so one blended number across all of them hides that.

Sentiment scores look good in a demo. I treat them as a rough flag, then read the actual wording of any answer that matters to a live deal.

Which sources do AI answers cite, and why does tracking them matter?

AI answers lean on a small set of outside pages. Mostly lists, directories and review sites. That makes the cited sources in your log the most useful thing you'll collect, because being named usually means being on those pages.

Our own testing shows how tight that set is. Across 900 ChatGPT answers about Australian agencies, a single third-party "best agencies" site was cited in 168 of them. The next three were a second listicle, Semrush's agency directory and Clutch. If you want to be recommended, the job is getting onto the handful of pages the model already trusts.

The engines differ, too. Perplexity cited a review directory (Semrush Agencies, Clutch, DesignRush and the like) in 36 of 44 agency answers. ChatGPT did it in 20 and Claude in 15. One agency's Clutch profile had six reviews and was cited 20 times.

We tested agencies. Your category's names will differ. In B2B software the cited set might be G2 and analyst lists, and in manufacturing it's more likely trade association directories and industry publications. Your log tells you which.

We record four separate states for every source: the page is live, the engine fetched it, the engine cited it, and the page mentions you. A page can be fetched and never cited, or cited without naming you at all. Keep them apart. Otherwise one win gets counted four times.

How do you track ChatGPT and AI traffic in GA4?

Build a custom channel group in GA4 that matches AI referrers, then check that channel for bots before you report it. ChatGPT tags its outbound links with utm_source=chatgpt.com, which GA4 files as a referral unless you group it.

A starting regex for session source: chatgpt\.com|chat\.openai\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai. Put the AI channel above Referral in the channel order, or GA4 will file those visits as ordinary referrals before your rule ever runs.

Then watch for three traps.

Bots dressed as AI visitors. Here's what we found on an enterprise B2B client's account. A client's GA4 showed AI-assistant visits jump from 354 to 6,050 in a month. Almost none of it was real. About 5,700 came in three one-hour bursts: same 1920x1080 desktop Chrome, all new users, no referrer, about 20 seconds each, landing on two pages. Real AI traffic was 385 sessions against 354 the month before, roughly flat.

Tags that vanish. In October 2026, ChatGPT links stopped carrying the utm_source tag after the GPT-6 release. Search Engine Land reported that OpenAI confirmed a bug and began rolling out a fix on 15 October. Because referrer data is missing from most ChatGPT clicks, those untagged visits landed in Direct, where they looked like people typing your URL. Annotate that week and move on.

Conversions that never fire. The same account had a second fault. A client's four enquiry types never counted as conversions. Their event names ran 41 to 50 characters; GA4 cuts at 40. The 40-character "career" event counted every time. Nobody had noticed, because the events still showed up. So check that your key events actually fire before you credit, or blame, any channel for the enquiries it seems to bring in.

Server logs add one more view. OpenAI's crawler documentation says ChatGPT-User may visit a page when someone asks ChatGPT a question. Those hits are your pages being read live. OAI-SearchBot is the crawler that decides whether you can appear in ChatGPT search at all.

Do you need an AI visibility tool, or can you track it manually?

You can track AI visibility manually. A spreadsheet and the free reports from Google and Bing will give you a solid baseline. A tool earns its fee past a few hundred answers a month, or once you track several markets. Want the numbers to move, not just be measured? That's an agency's job.

Tools you'll see in this search include Semrush's AI Visibility Toolkit, Ahrefs Brand Radar, Profound, Peec AI, Otterly, Scrunch and plenty more. I compared 12 of them in our guide to AI visibility tools.

Whichever you pick, ask to see the stored answer text, engine, date and country behind every score. If you can't open the answer, you can't check the number.

What does AI visibility tracking look like for a B2B company?

Here's an illustrative example. Take a mid-sized manufacturer that sells industrial flow sensors into mining and water utilities in Australia and the US. The CEO asks one question: do we show up when buyers ask ChatGPT?

The marketing lead writes 40 questions. Ten are problem questions, like "why do flow sensors fail in slurry". Fifteen are category questions, such as "best industrial flow sensor suppliers in Australia". Ten compare named competitors, five are brand checks, and each runs three times in ChatGPT, AI Overviews, Perplexity and Copilot, once per market.

A first month's sheet like that might read:

• The brand checks come back fine, because engines know the company exists.
• The category questions name three larger competitors and rarely the manufacturer.
• The cited sources are two trade directories, an industry magazine's supplier list and a competitor's comparison page.
• One answer lists a discontinued product as the current range.

Now the tracking has done its job, because it's become a work list. Fix the out-of-date product facts on the site, then build a comparison page and a page for each application. Get into the industry list and the magazine's supplier roundup, then rerun the same 40 questions next quarter to see whether the share moved.

That's the loop. Enterprise just adds markets, product lines and people who want the report.

How do you improve AI visibility once you're tracking it?

Fix what the log shows: wrong facts, missing pages, blocked crawlers and absent sources. Google's guidance says "specific optimization isn't required for AI Overviews and AI Mode", so most fixes are good SEO aimed at sharper sources.

1. Let the crawlers in. Allow Googlebot, Bingbot and OAI-SearchBot in robots.txt and at your CDN.
2. Put plain facts on your own pages. Products, markets, specs and who you serve belong in text, not locked inside images or PDFs where an engine may skip them.
3. Build a page for each tracked question you lose. Comparison, application and cost pages answer the questions shortlists are built from.
4. Make structured data match the page. Organisation, product and FAQ markup should state what the copy states.
5. Get onto the sources the engines already cite. Think digital PR, industry lists, reviews and earned links, all aimed at the pages that already show up in your citation log.
6. Rerun the same questions. Same set, same engines, same runs. Change the questions halfway through and you've lost the comparison you were paying to build.

Most of that happens on your website. The rest happens on the sites your log says the engines trust.

How does Searchmaxxed track and improve AI visibility?

Searchmaxxed is an SEO and AI SEO agency. We track what ChatGPT, Gemini, Perplexity and Google's AI answers say about you, then build the pages, fixes and outside sources that change it. One team does the technical SEO, content, GEO, AEO, LLM SEO, link building and digital PR.

Each month we ask the engines a fixed set of your buyer questions and log who's named, which pages are cited and the date. We also log the competitors named instead. That sits next to your Search Console and GA4 data, and enquiries are the number we report against. Our AI search optimisation service sets out the full scope.

A page release is an approved new page or a substantial rebuild of one. Every package includes a monthly decision memo. You own the site, repository, content and data from day one, and our enterprise SEO service covers large and multi-market sites.

Here's how we do it. We build on a website you own, our Agentic Website. AI agents pull your Search Console and GA4 data, read the Google results and check what AI answers say about you and your competitors. That evidence feeds the Managed Search Loop, which shows which page, fix or link should bring the most enquiries next. We build it. Then we check the live page and measure the change. Off your site, the Off-Page Source Layer builds the links, digital PR and mentions that AI answers draw on.

Searchmaxxed's work has generated more than A$10 million in revenue through search for its clients.

Is Searchmaxxed right for you?

Searchmaxxed suits marketing and growth teams at enterprise, B2B SaaS, venture-backed, manufacturing and public-sector companies. It fits teams that want AI visibility to move, not just to be measured, and want the work done for them. If you only want a dashboard, a tool will serve you better and cost less.

Frequently asked questions

What should you do next?

This week, write 30 buyer questions and run each three times in ChatGPT and Google's AI Mode. Log who's named and what's cited, and switch on Search Console's generative AI report and Bing's AI Performance report beside it. If you'd like us to track it and build what moves it, let's talk pricing.