How to track your brand mentions in AI answers

You can ask ChatGPT what it thinks of your brand right now — and tomorrow the answer may differ. Tracking brand mentions in AI answers is less like checking a ranking and more like running a small survey, repeatedly, with the results written down. Here is how to do it by hand, what to record, and when a tool earns its keep.

Why a one-off check misleads

Two reasons. First, answers are not stable: Google documents that its AI Mode and AI Overviews can run on different models and techniques, so their answers and links will differ¹. One good answer today proves little about the answers your buyers get this month. Second, your analytics will not fill the gap: Cloudflare's data puts it plainly: "Over the past 12 months, 80% of AI crawling was for training, compared with 18% for search and just 2% for user actions"³. AI systems read your site far more than they send visitors back, so the answers themselves are the only place to look.

Step 1: write the questions buyers actually ask

Do not start from your brand name. Start from the moment before a buyer knows you exist: "best X for small teams", "how do I choose a Y", "is Z worth it". Add the branded ones afterwards ("What is Acme?", "Acme vs alternatives", with your own name in place of Acme). Ten to twenty questions is a workable starting set; your support inbox and sales calls are the best source.

Step 2: ask them on a schedule, in the right place

Ask each question in the assistants your market uses, from the market you sell to and in its language. Keep the wording identical between runs, or the comparison means nothing.

Step 3: record four things per answer

For every run, note:

1. Mentioned or not — is your brand in the answer at all?
2. Position and tone — first recommendation or an afterthought; recommended, neutral, or warned about.
3. Competitors named — who appears when you do not.
4. Citations — which pages the answer leaned on. OpenAI's documentation describes web search as giving models current information from the internet, with answers that come "with sourced citations"². In our view, those cited pages are where next month's answers come from, so they are the most actionable column in your log.

A spreadsheet with question, date, assistant, and these four columns is a real monitoring system. Its weakness is not the format — it is that someone has to do it every day, identically, forever.

Step 4: turn the log into work

The log is only useful if it changes what you do. Each "competitor named, you absent" row points at cited pages to read; each wrong fact points at a source to fix. Without this step, tracking is a scoreboard for a game nobody is playing.

When a tool earns its keep

OLAB Assist automates this exact loop: it records the AI answers to your questions every day, keeps position, tone and citations per answer, reads the cited pages and flags the ones that mention a competitor but not you — then turns each finding into a task, with context, evidence, steps and code, that you or your AI agent can run. The spreadsheet above is the manual version of what the dashboard does on a schedule.

Start your own log

Start on the free plan, one site for 7 days with no card needed, or try any paid plan free for 7 days. Add your first questions today; the trend line starts tomorrow.