Check an AI answer for your brand

What are the six signals?

Each one is mechanical, and each is reported with the substring that produced it, so you can disagree with the rule rather than with the tool:

1. Mentioned — the brand name, or an alias you supplied, appears anywhere in the answer. Latin names are matched on word boundaries, so “Ramp” is not found inside “ramp-up”; names containing non-Latin characters are matched as substrings, because word boundaries do not exist in those scripts and a boundary pattern would match nothing at all.
2. Mentioned early — the first mention lands within 300 characters, roughly two sentences. An answer that names you in its closing paragraph is not an answer that surfaces you.
3. In a list of options — the mention sits on a bulleted or numbered line, which is where a reader actually makes a choice.
4. Carries a recommending word — within 120 characters of the mention there is a word from this published list: recommend, recommended, best for, a good fit, worth considering, top pick, strong option, leading, 优先, 推荐, 首选, 适合, 值得考虑. A mention without one of these is a name, not a suggestion.
5. Sources cited — the domains the answer links to. If your own domain is not among them, the engine is describing you from somewhere you do not control.
6. States what it cannot confirm — detected from this published list: cannot confirm, can't confirm, cannot verify, not sure, don't have, do not have, unclear, may vary, 无法确认, 不确定, 需核实, 请核实, 建议核实. On a factual question this is a pass rather than a failure: a blank is recoverable, a confident wrong number that a customer quotes back is not.

Which questions should I ask?

Twenty to start with, in four groups that measure different things. Replace the bracketed parts with your own words — and use the language your buyers use, because the same question in English and in Chinese are two different measurements.

Category (No brand name in the question)
1.  Which companies offer [product type] for [audience]?
2.  What are the best [category] tools in 2026?
3.  Who are the main vendors in [category]?
4.  Which [product type] is suitable for [size / region / industry]?
5.  What should I look for when choosing [category]?

Scenario (A problem, no brand name)
1.  How do I solve [problem the product solves]?
2.  What is the cheapest way to [outcome]?
3.  How do teams like mine handle [workflow]?
4.  What alternatives exist when [common constraint]?
5.  How do I compare [approach A] and [approach B]?

Comparison (Your brand against a named alternative)
1.  [Your brand] vs [competitor]: which is better for [use case]?
2.  Is [your brand] worth it compared with [competitor]?
3.  What are the disadvantages of [your brand]?
4.  Which is easier to set up, [your brand] or [competitor]?
5.  If I need [requirement], should I choose [your brand]?

Fact (Your brand by name, against verifiable detail)
1.  What does [your brand] do?
2.  Who owns [your brand] and when was it founded?
3.  What does [your brand] cost?
4.  Which company is behind [your product name]?
5.  Does [your brand] handle [specific requirement]?

What this tool cannot tell you

• It cannot produce a rate. It scores one answer at a time. Twenty questions asked three times is sixty samples — enough to notice a change, nowhere near enough for a percentage that describes your market.
• It does not know what any engine would answer. It has never spoken to one. If you want that automated, it costs money per query and you should buy it from somebody who is explicit about that.
• A brand whose name is an ordinary word will sometimes match the word. “we ramp up quickly” contains the standalone token “ramp”, and no string rule can tell it from the company. Every verdict shows the substring so you can see it, and a more specific alias fixes most cases.
• One answer is an anecdote. Engines change between sessions, versions and regions. Record the date, the model and three runs per question, or the second measurement will not be comparable to the first.