How to Check If AI Mentions Your Brand (and Track It Over Time)

There's no Search Console for AI answers — you check by asking. Build a fixed set of the questions your buyers actually ask, run them across ChatGPT, Perplexity, and Google's AI results on a regular cadence, and log whether your brand appears, in what position, and with what sentiment. Consistency of the prompt set is what turns anecdotes into a trend line.

Build the prompt set first

Ten to twenty prompts, written the way real buyers phrase them, frozen so month-over-month results compare cleanly:

• Category prompts: "best [your category] tools," "how do I solve [the problem you solve]" — the queries where being mentioned wins new demand.
• Comparison prompts: "[you] vs [competitor]," "alternatives to [category leader]" — where sentiment matters as much as presence.
• Brand prompts: "what is [your brand]," "is [your brand] legit" — where you check that AI describes you accurately.

Our AI citation checker generates a starter prompt set from your brand and target queries.

Run it across engines — they disagree

Each engine draws on different sources (the full breakdown is in where to get cited first): Perplexity leans on live web retrieval with visible citations, ChatGPT blends training data with search and licensed sources, Google's AI results draw on its index and SERP features. A brand can be prominent in one and invisible in another — which is diagnostic information, not noise: it tells you whether your gap is crawlability, community presence, or content structure.

What to log

• Mentioned / not mentioned per prompt per engine — the core binary.
• Position and framing: first recommendation or afterthought? Recommended or merely listed?
• Sentiment and accuracy: AI repeating outdated pricing or a wrong positioning sentence is a fixable content problem — usually traceable to what your own site and llms.txt say.
• Cited sources: when engines show citations, record which pages earned them — those sources (often Reddit threads and comparison posts, per our 26,000-citation analysis) are the surface area to grow.

Cadence and automation

Monthly is the right manual cadence — AI answers shift with model updates and index refreshes, but not daily. The workflow above costs an hour a month at 15 prompts × 3 engines; past that scale, or past one brand, automation wins. This is exactly the visibility loop CommunityMentions runs as a service — tracking where you appear, where competitors do, and which community mentions move the numbers — so the trend line builds itself while you work on the inputs.