How to Track Your Brand Mentions in ChatGPT

Tracking your brand mentions in ChatGPT means running a fixed set of shopper queries against it on a schedule, recording whether and where your brand appears, and trending the rates. It turns "does ChatGPT know us?" from a vibe into a number.

Here is why we wrote this: before our rebrand, the two highest-demand queries reaching our old site were "chatgpt brand tracker" (64 impressions in 28 days) and "chatgpt visibility tracker" (82 impressions), per our own Search Console data from 2026. Buyers were asking; nobody had built the page. This is that page.

Why the obvious method fails

The obvious method is opening ChatGPT and asking "what are the best [your category] brands?" It fails for four reasons:

1. Sampling. The model draws from a distribution. The same prompt returns different brand lists across sessions. One answer is one draw.
2. Memory. Your account knows you. Founders checking their own brand from their own account get flattered results a real shopper never sees.
3. Fan-out. ChatGPT rewrites one shopping question into 5-7 internal searches, each pulling different sources. What you typed is not what it searched. We tore this apart in our query fan-out research if you want the mechanics.
4. Drift. Model updates and retrieval changes move answers week to week. A January check says nothing about March.

The failure mode is predictable, and we see it constantly: a founder spot-checks five prompts, sees the brand twice, and reports "we're in 40% of answers." Run the math: at a true 40% rate, a 20-prompt sample swings between 25% and 55% by chance alone. Most DIY trackers are measuring noise.

The four numbers that matter

Measure these per query, per surface, over time:

Absent rate is the underrated one. A brand at rank #2 on ten queries and absent on forty has a visibility problem that averages hide. Getting into the answer set at all is the scarce win; our 86% Rule research found that when a retailer page gets cited in a ChatGPT shopping answer, it takes the top recommendation slot 86.03% of the time. Eligibility beats ranking.

The manual protocol

If you want to run this yourself, do it properly:

1. Map the query set. 30-200 queries covering head terms ("best running shoes"), use cases ("shoes for flat feet marathon training"), and comparisons ("Brand A vs Brand B"). Write them the way shoppers talk, not the way your category manager does.
2. Clean sessions. Fresh chats, memory off or a logged-clean browser profile. Note region; answers differ by market.
3. Repeat runs. Each query several times per cycle, weekly cycles. Log brand mentioned yes/no, position, competitors named, sources cited.
4. Trend it. A spreadsheet works at small scale: rows are queries, columns are weekly rates. Watch deltas, not single weeks.

Budget honestly: 50 queries at 3 runs weekly is 150 transcripts a week to read and score. That is a part-time job, which is why most teams do it for two weeks and quit - and why we built this exact loop as a product, Citation Rank, running it continuously across ChatGPT, Gemini, Perplexity, Rufus, Copilot, and Claude.

What acting on the numbers looks like

Measurement only pays when it drives edits. Amazon India mapped queries across six categories, tracked weekly, and pointed content and catalog fixes at the absences. In six weeks: Share of Voice +0.98%, Top-5 Presence +2.4%, Price Competitiveness +5.53%, and absent rate down 2.5%. The full loop is documented in the Amazon India case study.

The pattern generalizes: find where you are absent, fix the specific gap (a missing comparison page, thin product data, no review signal), remeasure. Repeat.

Start with a baseline

You cannot manage a number you have never seen. Get your free Citation Rank scan: work email plus brand URL, and in 24 hours you get your rank across six AI surfaces with three prescriptive fixes. No login, no credit card.