AI visibility benchmarks should show their method before their leaderboard.
This hub publishes the benchmark structure AnswerMentions uses: prompt families, source taxonomy, error patterns, privacy rules, and a seed dataset. It does not pretend a small sample is a market-wide ranking.
Seed benchmark
Four segments, one consistent evidence structure.
The seed table is intentionally modest. It is a public starting point for how audits are grouped, not a claim that one vendor or industry has been statistically ranked.
Download seed CSV| Segment | Prompt family | Prompts | Common missing source | Common error risk |
|---|---|---|---|---|
| B2B SaaS | Alternatives and best-fit | 20 | Category lists and integration marketplaces | Old plan limits or stale integration claims |
| Local services | Location plus urgency | 20 | Directories and business profiles | Hours, phone, service area, and financing conflicts |
| SEO agencies | White-label prospect audit | 20 | Prospect screenshots and vertical proof pages | Overpromising AI rankings |
| Professional services | Practice-area comparison | 20 | Provider profiles and local directories | Outdated team or credential records |
Methodology
What must be true before a benchmark earns trust?
- 01Define one buyer segment before writing prompts.
- 02Run prompts across ChatGPT, Gemini, Perplexity, and Google AI results.
- 03Record brand mentions, competitors, recommendation reason, cited sources, and wrong facts.
- 04Classify cited sources into first-party, directory, review, category list, community, media, or official profile.
- 05Publish only aggregated, opted-in, or redacted findings. Private audit URLs stay authenticated and noindexed.
Benchmark caveats
The benchmark should become more useful as real opted-in data accumulates. Until then, the page is honest about its scope.
Is this benchmark an industry ranking?
No. The current public benchmark is a seed methodology and segment table. AnswerMentions will publish aggregate rankings only when enough opted-in audits support a defensible sample.
Why publish a benchmark before broad market data exists?
Because methodology is part of the product. Buyers should understand prompt families, source taxonomy, privacy rules, and update cadence before trusting a score.
Will private audits be added to public benchmarks?
Only as aggregated or explicitly approved data. Private reports remain behind authentication and are not placed in the public sitemap.
How often will the benchmark update?
The target cadence is monthly for active segments and event-driven updates when major answer-engine behavior, source patterns, or pricing facts change.