ai-visibility-check
Skill0 starsChanged 49 days agoMITView on GitHub
Measure whether AI engines recommend a brand for its buying questions, and where competitors win instead. Use when someone asks if ChatGPT recommends them, wants to track AI search visibility, asks how they compare to a competitor in AI answers, or wants to know which questions to track.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- fabio-bergmann/seo-geo-skillstheir own version, first seen Nov 16, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- fabio-bergmann/seo-geo-skills★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository4 other files it gives its agents.
- answer-gap-contentSkill
- answer-readinessSkill
- geo-auditSkill
- llms-txtSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 36
Record the whole answer,: not a yes or no. Position in the list, what it was credited with, and every source cited are the parts that tell …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AI visibility check
Whether a brand is named when a buyer asks an engine for a recommendation. This is a measurement task, and almost everything that makes it useful is in how the questions are chosen and how the results are read.
The questions are the measurement
A visibility number is only as good as the question set behind it. Get this wrong and the rest is noise.
Ask what a buyer asks, not what a brand wants to rank for. "Best session replay tool for B2B SaaS" is a buying question. "Acme Analytics" is a branded lookup and tells you nothing: the engine will name them because they were named.
Cover the shapes buyers actually use:
| Shape | Example |
| Category best | best CRM for a two person agency |
| Constrained | CRM with a free tier and Gmail sync |
| Alternative | alternatives to HubSpot for a small team |
| Comparison | Pipedrive vs Close for outbound sales |
| Job to be done | how do I stop leads going cold after a demo |
| Risk | which CRMs are SOC 2 and store data in the EU |
Twenty to fifty questions. Below twenty, a single answer swings the number. Above fifty, the tail is questions nobody asks.
Fix the set before measuring. Changing questions between runs makes movement unreadable. Add to the set, keep the old ones, and note the date each was added.
Running it
- Several engines, separately. ChatGPT, Claude, Perplexity, Gemini, Google AI Mode and Copilot answer the same question differently and draw on different sources. A blended score hides where the problem is.
- Repeat each question. These are sampled, not deterministic. The same question asked three times can name three different sets. Three runs per question per engine is the floor, and the variance is itself a finding: a brand named in one run of three is in a different position from one named in three of three.
- No personalisation. A logged-in account with history is measuring itself.
- Record the whole answer, not a yes or no. Position in the list, what it was credited with, and every source cited are the parts that tell you what to do next.
Reading the result
Four things, in this order:
- Presence. In what share of runs is the brand named at all. This is the headline.
- Position. Named first, in a list of five, or in a footnote. First is worth several times the rest.
- Who wins instead. For every question where the brand is absent, which brands are named. This list, sorted by frequency, is the real competitor set, and it is often not the one the company believes it has.
- What is cited. The URLs the engine leans on. These are almost never the brand's own site. They are listicles, comparison posts, forum threads and review pages, and they are the lever: a citation source that names competitors and not the brand is a specific, fixable gap.
What to do with a gap
A question where a competitor wins has exactly three causes, and they need different work:
- The engine cannot read the site. Crawler blocks, client-rendered content. See the
geo-auditskill. - Nothing on the site answers that question. The content gap. See the
answer-gap-contentskill. - The sources the engine trusts do not mention the brand. The hardest and usually the largest: the fix is getting into those specific pages, not writing another blog post.
Diagnose which before recommending anything. Most reporting in this space skips straight to "write more content", which fixes only the middle one.
Honest caveats
- Answers move on their own. Model updates change results with no change on the site. One run is a snapshot, and a trend needs weeks of the same questions.
- Presence is not traffic. Being named is upstream of a visit and is not measurable in analytics the way a click is. Do not promise attribution that does not exist.
- Do not report a single blended score without the question set beside it. It is unfalsifiable otherwise.
With the Peak Answer MCP connected
my_visibility returns measured history for the connected brand rather than a fresh one-off check, and my_questions with only_losing gives exactly the gap list above, with the winner named. my_brand says which domain the connection is scoped to; ask it rather than guessing. geo_audit runs ten real buying questions against a real model for any domain, but it is a background run: call it once and collect the result later, never block on it.
Free, no signup: https://peakanswer.com/tools/ai-visibility-audit
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

