A clearer way to measure whether AI finds your brand.
100 Questions turns a vague concern—“do AI systems recommend us?”—into a reviewable benchmark with frozen questions, web-grounded answers, source evidence, and explicit limits.
Why it exists
AI visibility should be inspectable, not mystical.
AI answer systems can mention a brand, omit it, cite its website, or favor a competitor. A single unexplained score cannot show which of those events occurred or whether the underlying provider calls were usable.
The benchmark keeps visibility, prominence, selected-competitor share of voice, claimed-domain citations, and provider coverage separate. Reports link those measures back to the buyer questions, answers, and sources that produced them.
Measurement principles
Trust comes from showing the work.
These principles shape the product, the public methodology, and the way every result is described.
Evidence before conclusions
Reports retain normalized answers and source URLs so metrics and recommendations can be reviewed against the underlying evidence.
Comparable tests
Each provider receives the same frozen question set. Failed or ungrounded answers remain visible in coverage instead of silently changing a score.
Limits in plain sight
A benchmark is a time-stamped directional snapshot, not proof of causation, factual accuracy, or parity with consumer chat products.
Private by default
Runs belong to the authenticated account owner. The default answer-retention window is 30 days, and reports are not public profile pages.
Editorial accountability
Meet the person accountable for the product and guides.
Kyle Kane
Product maintainer and editor
Kyle Kane builds and maintains 100 Questions and reviews its public measurement methodology, product documentation, and evidence-led AI visibility guides.
View the public GitHub profileAreas of responsibility
- AI visibility measurement
- Benchmark design and interpretation
- Product documentation
- Technical search accessibility
Editorial and testing method
- 01Separate observed product behavior, first-party documentation, and interpretation.
- 02Prefer primary sources and link material claims to evidence readers can inspect.
- 03Keep study-specific model labels and dates frozen while current product facts update centrally.
- 04State limitations, denominators, and coverage beside conclusions.
- 05Correct material errors when identified and update the visible review date only after review.
No unverified degrees, certifications, independent-review claims, customer ratings, or professional credentials are published. Corrections can be sent through the contact page.
Featured on
Find 100 Questions around the web.
These launch, product, and search-performance profiles provide independent places to verify the product. Inclusion is not a rating, customer review, or endorsement.
What it is—and is not
Directional evidence for better decisions.
- The benchmark is
- A frozen, source-backed comparison across four model providers.
- The benchmark is not
- A guarantee of future mentions or a ranking factor sold as certainty.
- Reports include
- Coverage, citations, competitor evidence, retained answers, and prioritized actions.
- Reports do not claim
- Statistical representativeness, model causation, or consumer-chat parity.
Verify the details
Read the definitions before trusting the score.
The methodology documents question construction, grounding eligibility, denominators, coverage rules, retention, and known limitations.