WhoCites

WhoCites AI Visibility Benchmark — First Cohort Snapshot, 2026

WhoCites publishes a first public AI visibility benchmark from real paid scans. As of 2026-05-21, production has 10 completed scans, and 2 opted-in public scans are included in the benchmark table. The public cohort covers 112 engine responses across ChatGPT, Claude, Gemini, Grok, Copilot, Perplexity, and Google AI Overviews. The sample is intentionally small; the value is reproducible methodology and concrete proof, not an industry-wide census.

Last updated 2026-05-21

Run an AI visibility scan

Why this benchmark exists

Public AI-visibility benchmarks are rare. Most AI search advice is anecdotal or vendor-pumped. WhoCites publishes an aggregate snapshot from real paid scans so operators can compare their own results to a defensible reference, not a marketing claim.

Public benchmark table

The first public cohort contains 2 opted-in scans and 112 engine responses. Median visibility score is 12.5/100. The score range is 5.36 to 19.64. Median engine coverage is 4.5 of 7 engines, or 64.3%. Median per-engine mention rate is 12.5%. Engine mention-rate standard deviation is 10.6 percentage points.

Public scan A: 5.36/100 visibility score; 3 of 7 engines mentioned the brand; 0% median per-engine mention rate; 8.11% brand share of voice.
Public scan B: 19.64/100 visibility score; 6 of 7 engines mentioned the brand; 25% median per-engine mention rate; 18.33% brand share of voice.
Completed production scans counted by /api/public/stats: 10
Opted-in public benchmark cohort: 2 scans
Engine responses in public cohort: 112
Median visibility score: 12.5/100
Median engine coverage: 4.5 of 7 engines
Median per-engine mention rate: 12.5%
Engine variance: 10.6 percentage points

Scope of the first cohort

The first public cohort is small, post-launch, AI-built apps and SaaS sites that opted into public sharing after running a paid scan with WhoCites. Each scan covers 7 engines and roughly 8 buying-intent prompts per engine, or about 56 engine responses per initial scan. Aggregate patterns are reported; private scans remain private.

Public fields in the benchmark

The benchmark tracks visibility score, engine coverage, engine mention rate, brand rank where available, share of voice, and competitor displacement. The public endpoint intentionally omits customer contact data, Stripe IDs, raw AI excerpts, private recommendations, checkout session IDs, and full citation URLs.

Engine variance

Engines do not agree with each other. In the public cohort, engine mention-rate standard deviation is 10.6 percentage points. Coverage on ChatGPT does not predict coverage on Claude, Perplexity, or Google AI Overviews. Cross-engine variance is the rule, not the exception, which is why a single-engine spot check is unreliable.

Citation source pattern

Full source URL citation rate is not published for this first public cohort because the public proof endpoint deliberately omits private citation URLs. The public proxy is engine coverage and per-engine mention rate. Future benchmark updates should publish aggregate citation-rate statistics once they can be computed without exposing private source URLs.

How the cohort changes the methodology

Each future cohort refreshes this page. The methodology stays anchored at /methodology; only the snapshot numbers move. The published snapshot is the cohort that has run scans as of 2026-05-21.

Reproducibility and limits

The benchmark is bounded by the size of the cohort and the engines' own stochastic variance. It is directional, not census-grade. WhoCites does not extrapolate beyond the cohort and does not invent industry-wide percentages. Numbers in this page are sourced from the live audits table at build time.

Use this in your own writing

Direct quotes from this page are welcome with attribution to WhoCites and a link back. The numbers are not licensed for re-publication as standalone tables, but the methodology and reasoning are open.