# AI visibility benchmarks — how ready the web actually is

> Anonymized benchmarks from AI Syndicate's own measurement corpus: median AI Access
> scores by CMS platform and industry, per-engine citation rates, and the measured
> lift per kind of fix. Real measurements of real public pages — no estimates, no
> survey data.

Live figures: <https://www.aisyndicate.com/benchmarks/>

## Where the data comes from

Every time AI Syndicate scores a website, the resulting score is recorded as a dated
observation of that public page. That includes customers' own sites, competitors we
were asked to benchmark against, and sites scanned through our free tools. The
benchmarks published on this page are aggregates of those observations.

Nothing on the page is modelled, extrapolated, or self-reported. There is no survey
panel and no third-party data purchase behind these numbers — they are measurements
we took ourselves, kept with the date they were taken.

## What is published

- **The state of AI visibility** — median AI Access composite, page AI-readiness,
  citation share, and share of voice across every domain in the corpus, with the
  share of sites in the "strong" (75+) and "at risk" (49 and below) bands.
- **Per-engine citation rates** — the median share of tracked brands each named AI
  engine actually cites. Engines that failed to answer are excluded rather than
  counted as a non-citation.
- **Segment breakdowns** — median AI Access score by the CMS platform a site runs on
  (detected from the page itself) and by industry.
- **Measured outcomes** — the median score movement per kind of change shipped
  through AI Syndicate, measured by re-scoring after the change had time to land.

## Rules the page follows

1. **Aggregates only.** Counts, medians, and percentages. No domain, customer, or
   account is ever included.
2. **A floor before publishing.** A statistic appears only when enough distinct
   domains sit behind it to be meaningful; below that threshold the page says it is
   still gathering rather than showing a median of a handful of sites.
3. **No fabrication.** A missing measurement produces a visible gap, never an
   estimate. Measured outcomes require a real score on both sides of a change — that
   is why the outcomes section fills in most slowly.
4. **Dated and versioned.** Every observation carries the day it was taken and the
   version of the scoring method that produced it, so improving how we score never
   silently rewrites history.

## What is not published

The scoring recipes themselves, per-site results, and anything that could be used to
reverse-engineer an individual customer's position. The benchmarks are a public good;
the methodology is not.

## Frequency

The corpus grows continuously as scans run. These aggregates are recomputed daily.

## Related

- [Get your own AI visibility score](https://www.aisyndicate.com/)
- [Competitor scoreboard](https://www.aisyndicate.com/compare/)
- [What AI Access measures](https://www.aisyndicate.com/features/ai-access/)
- [Privacy policy](https://www.aisyndicate.com/privacy-policy/) — covers how
  aggregated, de-identified measurements are retained and used.