How to measure whether AI assistants mention your brand

When someone asks an AI assistant which tool to buy, the answer usually names a handful of products. If yours isn't among them, that buyer may never reach your site. The good news is that you can measure this, as long as you treat it like a survey rather than a search ranking.

Start from buyer questions, not keywords

Search rank tracking starts from keywords. AI answers start from questions, and people phrase them as they would to a colleague: "what's a good payroll tool for a 20 person company in Texas?" rather than "payroll software".

Write down the questions a buyer asks at each stage:

• Comparisons: "best X for Y", "X vs Y", "alternatives to X".
• Buying guides: "what should I look for in X?"
• Use cases: "how do I do Z with a small team?"
• Pricing: "cheapest X that does Y".

Twenty to fifty questions spread over these groups is enough to see a pattern. Fewer than ten and a single odd answer swings your numbers too much.

Ask every engine the same way

Each assistant answers differently, and the same assistant answers differently from one run to the next. To make results comparable:

1. Ask every question on every engine you care about, with web search on where the engine offers it.
2. Ask from the country your buyers are in, because answers cite local sources.
3. Keep the wording fixed. Changing a question resets its history.
4. Repeat on a schedule. Daily is ideal; weekly still shows a trend.

Record more than "mentioned or not"

A yes/no is the headline, but the detail is where the work is. For each answer, note:

• whether your brand is named, and in what position among the brands named;
• which competitors appear alongside you;
• whether the answer describes you accurately;
• which sources the engine cites, and which of them are Reddit threads, review sites or your own pages.

The cited sources tell you where to act. If an engine keeps citing a particular Reddit thread when it recommends a competitor, that thread is part of what shapes the answer.

Read the results as a trend

Expect noise. A brand that appears in 40% of answers this week and 35% next week has probably not lost ground; one that drops from 40% to 10% over a month has. Look at the share of answers across all questions and engines, and only drill into single questions to explain a change.

Act where the answers come from

Measuring is only useful if it points to something to do. The usual levers are:

• Your own pages: clear, specific pages for each use case and comparison buyers ask about.
• Third-party sources the engines cite: review sites, directories and community discussions.
• Reddit: taking part in relevant threads honestly, as yourself, and within each community's rules. Disclose your affiliation, answer the question that was asked, and skip threads where a mention wouldn't help the person asking.

If you'd like to see where you stand today, the free AI visibility check asks five buyer questions across four engines and emails you the results. To find the communities where your buyers talk, try the subreddit finder.