field notes · for small business

the 20-question prompt sheet: how to track your own ai visibility every week.

elisabeth hitz · july 26, 2026 · 5 min read

this is the actual template, not the theory behind it. copy the structure, fill in your own category, run it weekly.

the 20 questions, by group

Evaluation questions (8), where buying decisions actually form:

  1. what's the best [your category] for [your specific customer type]?
  2. who should I hire for [the specific job you do]?
  3. what's a good [your category] for [a specific use case]?
  4. i need help with [the problem you solve], who's good at that?
  5. what's the top [your category] in [your city or niche] right now?
  6. who do people recommend for [specific service]?
  7. what should I look for in a [your category]?
  8. is [a specific competitor] worth it for [use case]?

Comparison questions (6), where you get directly measured against a name:

  1. how does [you] compare to [specific competitor]?
  2. [you] vs [specific competitor], which is better for [use case]?
  3. is [you] worth it compared to [doing it yourself / a cheaper alternative]?
  4. what's the difference between [your category] and [adjacent category]?
  5. [competitor] alternatives?
  6. who else does what [you] does?

Definitional questions (6), where first impressions form for people who've never heard of you:

  1. what is [your category]?
  2. what does a [your job title] actually do?
  3. how much should [your category] cost?
  4. what's [your specific method or framework name]?
  5. is [your category] worth paying for?
  6. how do I know if I need [your category]?

what to log for each one

Run every question through ChatGPT, Perplexity, and Google's AI Mode. For each platform, log four things:

  • Appeared? yes / no
  • Linked citation, or just a name-drop?
  • Tone: positive, neutral, or hedged/negative
  • Who else got named, so you can see who you're actually up against for that specific question

Don't run each question once and call it done. Run the same question two or three times over the week, since the same question can return a different answer on different runs. If your brand shows up 2 out of 3 times, that's a real, if incomplete, answer, not a coin flip you should re-run until you get the result you want.

how to grade what you find

what you seewhat it usually means
Strong appearance, weak citationsYour content is generating mentions but not giving models something specific enough to link to
Weak appearance across evaluation questions, fine on definitional onesPeople can find out what your category is, but you're not making the actual shortlist
Fine mentions, flat or hedged toneSomething you don't control is framing you badly, worth tracing to its source
Nothing at all, any questionWorth checking whether AI crawlers can even reach your site before assuming it's a content problem

why this beats checking once and moving on

A single check tells you where you stand today. A weekly log, run over a month or two, tells you whether whatever you're doing, new content, third-party mentions, structural fixes, is actually moving the number. Most of the work that improves AI visibility, earning mentions on sites you don't control, building topical depth over time, compounds slowly. Without a log, you have no way to tell "nothing's working" from "it's working and I can't see it yet."

keep it simple

A spreadsheet with rows for each question and columns for each platform and week is enough. This doesn't need a subscription, it needs the habit of actually running it weekly instead of checking once when you're anxious about it and never again.

want the deeper version of this?

the free 60-second ai visibility scan runs against your site right now, results on the page, no email needed.

run the free scan

browse all free tools →

frequently asked questions

How many prompts should I track for AI visibility?
Fifteen to twenty is enough for a solo operator. Split them across evaluation questions, comparison questions, and definitional questions, since each group reveals a different part of how you are being surfaced.

What should I log for each prompt?
Four fields per platform: whether you appeared, whether it was a linked citation or an unlinked mention, the tone you were described in, and which competitors were named alongside you.

How many times should I run each prompt?
Two or three times across the week. AI answers vary between sessions, so a single run cannot distinguish a real result from noise. If you appear in two of three runs, that is a genuine signal.

Which AI platforms should I track?
ChatGPT, Perplexity, and Google's AI Mode at minimum. They draw on different source pools and frequently disagree, so tracking one gives you a narrow and potentially misleading view.

How long before tracking shows anything useful?
One to two months. Most of the work that improves AI visibility, particularly earning mentions on sites you do not own, compounds slowly. Without a log you cannot distinguish slow progress from no progress.

written by elisabeth hitz, certified in anthropic's ai fluency program (framework & foundations, and ai capabilities & limitations), plus claude 101 and claude cowork. template drafted 25 july 2026; measurement method detailed in the companion pillar post linked above.