Three ways to find out, what each one actually measures, and why the same questions asked every week beat a spot-check every quarter.
The short answer. There are three ways to find out whether an AI assistant names your brand: ask it yourself and read the answer; run a one-off check that asks several buyer questions for you and quotes the answers back; or track a fixed set of questions on a schedule, so you can see a trend and work on what it shows. All three measure the same thing: whether your name appears when a buyer asks a question your product answers. They differ in whether the result is a snapshot or something you can act on.
Settle one thing first, because it decides which approach you need: a single answer is not a measurement. Ask Claude the same question twice and you can get two different shortlists. The model samples, so run-to-run variation is real and it isn’t about you. The useful question is not “did Claude mention us?” but “in how many of N answers, and is that number going up?”
Open ChatGPT, Claude or Gemini and ask what a buyer would ask: “what’s the best CRM for a 20-person sales team?”, not “tell me about [your brand]”. Leave your brand name out, since a leading question defeats the point. It costs nothing and it’s a fine way to feel the problem. It doesn’t scale: you see one run of one question on one assistant, and the consumer apps add memory and personalisation, so a colleague can get a different answer to the same prompt.
A one-off check automates the spot-check. It writes several realistic buyer questions for your category, asks each one independently, and reports how many answers named you, with the quoted sentence rather than a checkmark. Mentioned’s free check does this on Claude, ChatGPT and Gemini, on our keys. DeepSeek, Llama 3.3 and Perplexity run on your own API key.
A one-off check is the right baseline. Its limit is that it writes fresh questions every time, so two one-off checks a month apart can’t tell you whether your visibility changed or the questions did.
Tracking fixes that. You choose a brand and a category once, the questions are written once, and the same questions are asked on every run. When the number moves, it’s because the answers changed.
Several tools sell this. Mentioned is one: a free account tracks 5 queries, run weekly on Claude and Gemini on us, and paid plans from $39/month add more queries, Actions (a ranked list of what to fix, each with a playbook of the usual fixes), sentiment and an MCP connection to Claude. Pro and Agency add all six assistants per check and daily runs on our key. Profound, Peec AI and Otterly.AI also sell AI visibility tracking across several assistants. Coverage and pricing change often, so check each one’s own site; our comparison of seven AI visibility tools is dated and links every source. We have no commercial relationship with any of them.
Track the same questions weekly. A single run is noisy; a locked set of questions asked every week isn’t. Read the trend across runs rather than reacting to one, and look closer when it moves after something you shipped: a new comparison page, a launch, fresh coverage on a review site. Daily runs make sense for the few questions you’re actively working on, where you want to see a change within days rather than weeks.
Run a free check with your brand and the category a buyer would actually type. It asks five questions independently on Claude, ChatGPT and Gemini, and shows the sentence from each answer that named you, or who got named instead. Then track your three to five most important buyer questions weekly, which a free account covers. When you want the ranked list of what to fix, that’s the Starter plan.