Skip to content
Vouch
Join the waitlist
Menu

All posts

How to Track Brand Mentions in AI Search

By Vouch5 min read

Vouch blog cover with the title "How to Track Brand Mentions in AI Search" beside a line drawing of a half-circle radar with three rings and small circles on them, one of them a filled red dot.

When a buyer asks ChatGPT for the best options in your category, whether your brand comes up, and what is said about it, shapes the shortlist. Your analytics see none of it: no impression, no click, no referrer. Tracking brand mentions in AI search is how you make that conversation visible.

This is the how-to. For what each metric means and why it matters, start with AI search visibility; this post is about collecting the data, free or with a tool, and acting on it.

What to record

Four things are worth tracking, in rising order of value:

  1. Mentions: does the assistant name your brand for the questions that matter?
  2. Citations: does it link to your site as a source?
  3. Position and sentiment: are you named first or last, praised, hedged or criticised?
  4. Survival through the conversation: are you still named when the buyer asks the follow-ups about downsides, trust, price and "which one should I buy"?

Most tracking stops at the first three. The fourth is where sales are won and lost, as what happens after the first answer explains.

The free method: a structured manual log

An honest baseline costs nothing but a few hours.

1. Build the question set. Write 15 to 30 questions buyers really ask, covering the whole conversation:

  • Discovery: "what is the best X for Y?"
  • Shortlist: "compare A, B and C"
  • Objection: "what are the downsides of A?"
  • Trust: "is A reliable? what do owners say?"
  • Value: "is A worth the price?"
  • Decision: "which one should I buy?"

The buyer conversation prompts give you a ready set for each stage, and the objection prompts cover the hardest follow-ups.

2. Ask in fresh sessions. Use ChatGPT, Perplexity, Gemini and Copilot at least, signed out or in a temporary chat where you can, so your own history does not colour the answer.

3. Keep the follow-ups in the same chat. Ask the Discovery question first, then the rest in order without starting a new conversation. That is how a buyer asks, and it is the only way to see where your brand drops out.

4. Log every answer in a spreadsheet. One row per answer, with these columns: date, assistant, country, question, stage, named (yes or no), position, sentiment, sources cited, brands named instead, notes.

5. Repeat on a fixed schedule. Monthly is a sensible minimum for a manual log.

Be honest about the limits. Answers vary by session, wording, location and model version, so a manual log tells you whether you have a problem, not its exact size. Treat a change as noise until it holds for a few weeks.

What tracking tools add

AI visibility tools automate the question set across assistants, countries and time: consistent sampling, trend lines, captured citations and rival benchmarks. Most measure the first three levels in single answers; fewer follow the conversation into the follow-up stages. Best AI visibility tools in 2026 compares the main ones, and the comparison pages list each vendor's prices as checked on its own site.

When you compare them, ask four questions:

  • Which assistants are covered on the plan you would buy, not on the top plan?
  • How are answers collected? Answers fetched through an API can differ from what a signed-in user sees in the app. A good vendor tells you plainly.
  • How often does it run, and how does it handle variance? One answer a week is an anecdote.
  • Does it keep follow-up context, or score every question on its own?

What to do with the data

Tracking without a response is reporting for its own sake. Run the loop:

  1. Find the questions where you are missing or drop out.
  2. Work out why: missing content, weak coverage on other sites, facts that disagree, or blocked crawlers. The free AI crawler checker rules out the last one in a minute.
  3. Publish the fix: the comparison page, the proof of after-sales service, the clear pricing. Killer questions shows how to match the fix to the stage.
  4. Measure again next cycle.

For a one-off, deeper diagnosis that also checks your pages and third-party coverage, see the AI visibility audit.

Common questions

Can Google Analytics track AI mentions? No. It sees only visits that reach your site. AI referral traffic in GA4 shows how to find those, but what assistants say in answers that never lead to a click needs tracking of the answers themselves.

How often should I check? Monthly is a sensible manual cadence. Where AI answers drive real pipeline, daily tracking with a weekly trend is better, because answers move with model updates and source changes.

Why do I get a different answer each time? Assistants rarely give the same answer twice. Session context, wording, location and model version all change the result. That is why Vouch's methodology runs each Thread three times a week and shows no score until enough conversations have completed.

Automate the whole conversation

Vouch runs your questions every day across seven assistants, follows six-question buyer conversations on ChatGPT, Claude, Gemini and Perplexity, and shows the stage where your brand drops out, who is recommended instead and what to publish next. Every result sits on your dashboard, and paid plans can download an HTML report.

Vouch is invite-only for now, and every invitation starts with a 14-day free trial: join the waitlist.

Read next