How to Track Your Brand's AI Mentions for Free (Our Own Setup)

Most “AI visibility” platforms charge $300 to $1,000 a month to show you citation counts and referral traffic. We built both of our AI citation reports without one. Here is the exact free setup we use on topickz.com, with our own real numbers as the worked example at every step.

What you need, all free:

• A Microsoft Clarity account, already free, with the AI Visibility (Citation) feature enabled.
• A GA4 property on your site, already standard and free.
• GA4’s BigQuery export linked to that property, free on GA4’s standard tier, no GA4 360 required.

Step 1: turn on Clarity’s AI Visibility

• Create a free project at clarity.microsoft.com if you do not already track your site there.
• Open AI Visibility (BETA) → Citation. This is the tool that showed us 15,178 citations on Copilot and Bing in our last 30-day pull.
• It backfills history on day one, so you get real data immediately, not a blank dashboard waiting to accumulate.
• Pull the Citations, Share of Authority, and Grounding queries numbers. The grounding-query table is the one we used to find our single biggest content gap, a query sitting at 2.33% Share of Authority for three straight months.

Step 2: turn on GA4’s BigQuery export

• In GA4, go to Admin → Product links → BigQuery Links and link a Google Cloud project. This is free up to BigQuery’s standard query quota, which a site our size never comes close to hitting.
• Once linked, GA4 writes a new events_YYYYMMDD table to BigQuery every day, holding every raw event, not the sampled, pre-aggregated view the GA4 UI shows you.
• This is the step most guides skip, and it is the one that actually matters. The GA4 UI’s own channel reports often misclassify or undercount AI referral traffic. The raw export does not.

Step 3: isolate AI-referred sessions with one query

Once the export is running, this single query pulls every session referred by a named AI source:

SELECT traffic_source.source, traffic_source.medium, COUNT(DISTINCT
  CONCAT(user_pseudo_id,'-',(SELECT value.int_value FROM UNNEST(event_params)
  WHERE key='ga_session_id'))) AS sessions
FROM `your-project.analytics_XXXXXXXXX.events_*`
WHERE _TABLE_SUFFIX BETWEEN '20260926' AND '20261017'
  AND event_name = 'session_start'
  AND LOWER(traffic_source.source) IN ('chatgpt.com','perplexity.ai','gemini.google.com')
GROUP BY 1, 2
ORDER BY sessions DESC

• Swap in your own project and dataset ID, found under BigQuery Links in GA4 admin.
• This is the exact query shape we used to find our own real number: 42 ChatGPT-referred sessions over 21 days, spread across 26 different pages, not concentrated on one or two.

Step 4: check engagement before you trust the session count

• Pull engagement_time_msec for those same sessions. We caught a real problem this way: one of our pages showed over 100 “AI-referred” sessions that turned out to be bot or drive-by traffic, averaging under one second of engaged time, with incoherent referrers like Roblox mixed in.
• A session with real engaged seconds attached is a real human. A session with zero is probably not, no matter what the source field says.
• This single check is the difference between an honest AI-referral number and a number that will fall apart the first time someone asks you to defend it.

Step 5: cross-reference citation against referral

This is the step that turns two separate dashboards into one finding. Citation data tells you who Copilot is naming. Referral data tells you whether that citation actually sends anyone. They are not the same number, and conflating them is the most common mistake in AI-visibility reporting right now.

• We did this by content type and found a real gap: our buyer guides earn 18.6% of our citations but only 2% of real referred sessions, while listicles lead on both.
• You can run the same cross-reference for your own content types, your own categories, or individual pages.

What this setup cannot show you

• Citation data from Clarity covers Microsoft Copilot and Bing only. ChatGPT, Perplexity, Gemini, and Claude do not expose citation data to any tool, free or paid, including the $500 to $1,000 a month platforms. No free setup closes that gap, because the data does not exist outside those companies.
• Referral data only sees humans. A page can be heavily crawled by AI bots and show zero referred sessions, which is a different, separate signal worth tracking at the server level, not something this setup covers.
• This is descriptive, not predictive. It tells you what happened, not what will happen next. We learned that the hard way: a page we logged as a citation miss in July became our most-cited page by October, for reasons we still cannot fully isolate.

See it in action

Both of our AI citation reports were built entirely from this setup, with no other data source:

• The State of AI Citations 2026 used the Clarity side alone, across three dated snapshots.
• Which Content Gets AI to Send You Real Humans is the full cross-reference from Step 5, citation and referral data side by side by content type.

Topickz may earn affiliate commissions from some tools covered across our listicles. Commissions never influence a ranking or a citation finding. See our affiliate disclosure for details.

Related reading: the state of AI citations 2026 , which content gets AI to send humans , and our review methodology .