Median 73
The overall distribution is broad.
Scores range from 36 to 98, while the middle quartile range runs from 64 to 80.
n=3,001 sites ·
Public-web benchmark · baseline edition
The State of AI Discoverability
This inaugural authoritative cross-sectional baseline covers 3001 unique domains. The overall median is 73, with the interquartile range spanning 64 to 80. Strong access foundations coexist with much thinner survival through understanding and answerability.
2026-Q3 publication · captured · Methodology
3,001
domains analyzed
73/100
median score
64-80
middle-half range
Model-written, statistic-bound commentary
Key findings
The analysis is computed in code. GPT-5.6 Sol turns only those supplied metrics into commentary, and every numeric statement is bound to a named metric before publication.
Mean 39.6
Content Answerability is a comparatively weak scored dimension.
Its median is 40; Content Freshness & Authority is also low, with a mean of 34.7 and median of 25.
n=3,001 sites ·
Advantage 61.6 percentage points
Retrievable, self-contained sections are the clearest top-quartile separator.
The pass rate is 93.3 percent in the top quartile and 31.7 percent across the rest of the sample. This is an association, not evidence of causality.
n=3,001 sites ·
Correlation 0.454
HTML extractability and site architecture show the strongest supplied section-level association.
Their Spearman rank correlation indicates that these dimensions tend to co-occur, without establishing causality.
n=3,001 sites ·
A cumulative five-stage funnel
Where AI discoverability breaks down
The funnel represents cumulative population survival, not a score. Reachable retains 2963 domains, or 98.7 percent of the full population, after 38 leave at that stage. Permitted retains 2313, or 77.1 percent, after a further 650 leave. Readable retains 1834, or 61.1 percent, after 479 leave. Understandable retains 928, or 30.9 percent, after 906 leave. Answerable retains 538, or 17.9 percent, after 390 leave.
These percentages describe population survival through the funnel. They are not section scores and are not added to the audit score.
Reachable
98.7%
Permitted
77.1%
Readable
61.1%
Understandable
30.9%
Answerable
17.9%
Reachable
Can an automated reader retrieve the page?
2,963 remain
Permitted
Are search and answer crawlers allowed to access it?
2,313 remain; 89 unknown
Readable
Does the useful content survive different readers and rendering modes?
1,834 remain
Understandable
Is the entity presented consistently across visible and machine-readable sources?
928 remain
Answerable
Is the content divided into retrievable, self-contained answer units?
538 remain
Overall score distribution
The shape of the field
The middle half of the sample sits between 64 and 80, around a median of 73. This edition is the first authoritative baseline; earlier, much smaller samples are retained but not interpreted as change over time.
Eight comparable scored dimensions
Where sites win and lose
Content Freshness & Authority has the lowest average score; Bot Access & Control Plane has the highest. The bars show dispersion, not just the mean.
Content Freshness & Authority
avg 35
p25 25 · median 25 · p75 48 · p90 66
Freshness and authority have a mean of 34.7 and median of 25, placing this among the weakest scored dimensions in the baseline.
Content Answerability
avg 40
p25 0 · median 40 · p75 69 · p90 83
Answerability remains limited, with a mean of 39.6 and median of 40. Self-contained sections show a 61.6 percentage-point top-quartile pass-rate advantage.
Entity Clarity
avg 70
p25 56 · median 70 · p75 85 · p90 90
Entity Clarity records a mean of 69.9 and median of 70. Consistent entity naming carries a 25.9 percentage-point top-quartile pass-rate advantage.
HTML Extractability & Main Content Clarity
avg 77
p25 70 · median 82 · p75 88 · p90 92
Extractability and main-content clarity have a mean of 77 and median of 81.7. A canonical URL shows a 27.9 percentage-point top-quartile pass-rate advantage.
Trust & Security
avg 78
p25 70 · median 79 · p75 86 · p90 90
Trust and security record a mean of 78.1 and median of 78.5. About or Company links show a 28.5 percentage-point top-quartile pass-rate advantage.
Site Architecture & Coverage
avg 84
p25 80 · median 93 · p75 100 · p90 100
Site architecture is broadly strong, with a mean of 83.9 and median of 93.2. Navigation present in served and rendered HTML shows a 26.2 percentage-point top-quartile advantage.
Fetch, Render, and URL Integrity
avg 87
p25 81 · median 90 · p75 93 · p90 97
Fetch, render, and URL integrity are comparatively strong, with a mean of 86.6 and median of 89.7.
Bot Access & Control Plane
avg 94
p25 100 · median 100 · p75 100 · p90 100
Access controls are broadly strong, with a mean of 94.2 and median of 100.
Observational measure
Structured Data
Structured Data is reported for diagnostic context but carries no overall-score weight, so it is not ranked against the eight scored dimensions. It was evaluated for 44.4% of this sample; its observed median was 100.
Check-level pass-rate lift
What separates the top quartile
These are the checks with the largest pass-rate gaps between sites at or above the overall seventy-fifth percentile and the rest of the sample. The comparison is descriptive, not causal.
| Check | Top quartile | Others | Lift |
|---|---|---|---|
| Content is divided into retrievable, self-contained sectionsContent Answerability | 93.3% | 31.7% | +61.6 pp |
| Title, primary heading, and opening content share a clear topicContent Answerability | 51.4% | 9.7% | +41.7 pp |
| Social profile links presentEntity Clarity | 84% | 48.2% | +35.8 pp |
| Exactly one H1HTML Extractability & Main Content Clarity | 74.6% | 45.6% | +29 pp |
| About/Company link presentTrust & Security | 76.8% | 48.3% | +28.5 pp |
| Canonical URL presentHTML Extractability & Main Content Clarity | 81.6% | 53.6% | +27.9 pp |
| Important navigation is present in served and rendered HTMLSite Architecture & Coverage | 92.1% | 65.9% | +26.2 pp |
| Entity name is consistent across visible and machine-readable sourcesEntity Clarity | 65.4% | 39.5% | +25.9 pp |
Relationships inside the audit
Which signals move together
Spearman rho compares ranked section scores; phi compares pass versus non-pass outcomes for pairs of checks. Strong relationships can reveal shared implementation patterns or overlapping measurement, but they do not prove that one signal causes another.
Section-score correlations
| Pair | rho | n |
|---|---|---|
| HTML Extractability & Main Content Clarity + Site Architecture & Coverage | 0.454 | 3001 |
| Site Architecture & Coverage + Trust & Security | 0.443 | 3001 |
| HTML Extractability & Main Content Clarity + Trust & Security | 0.431 | 3001 |
| Entity Clarity + HTML Extractability & Main Content Clarity | 0.386 | 3001 |
| Entity Clarity + Trust & Security | 0.351 | 3001 |
| Entity Clarity + Site Architecture & Coverage | 0.332 | 3001 |
| Content Answerability + HTML Extractability & Main Content Clarity | 0.329 | 3001 |
| Fetch, Render, and URL Integrity + Site Architecture & Coverage | 0.323 | 3001 |
Check-pass associations
| Pair | phi | n |
|---|---|---|
| Viewport configured for mobile devices + Viewport meta present | 1.000 | 3001 |
| Sitemap available + Sitemap looks like XML (not HTML) | 0.999 | 2306 |
| Last-modified date present (schema or meta) + Publication date present (schema or meta) | 0.879 | 3001 |
| Canonical URL matches the current page URL + Canonical URL matches site origin | 0.625 | 1838 |
| Privacy policy link present + Terms link present | 0.600 | 3001 |
| Main content requires JavaScript that fetch-only AI crawlers do not run + Sufficient on-page text | 0.574 | 3001 |
| Canonical URL present + Meta description missing | 0.566 | 3001 |
| About/Company link present + Privacy policy link present | 0.515 | 3001 |
Only groups with at least 30 domains
Differences by inferred industry
Industry labels are inferred from public site signals and should be read as directional. Small groups are suppressed rather than over-interpreted.
| Inferred industry | Domains | Median | Middle half |
|---|---|---|---|
| SaaS & CloudThe inferred SaaS & Cloud group contains 318 domains and has a median overall score of 80. | 318 | 80 | 72-83 |
| EducationThe inferred Education group contains 193 domains and has a median overall score of 77. | 193 | 77 | 72-82 |
| Marketing & AdvertisingThe inferred Marketing & Advertising group contains 93 domains and has a median overall score of 77. | 93 | 77 | 67-82 |
| Healthcare | 40 | 76 | 68-81 |
| Finance & Banking | 97 | 75 | 66-80 |
| Non-Profit | 55 | 75 | 70-81 |
| Government | 91 | 74 | 64-80 |
| TechnologyThe inferred Technology group contains 870 domains and has a median overall score of 73. | 870 | 73 | 60-81 |
| Travel & Hospitality | 52 | 73 | 63-81 |
| Media & Entertainment | 500 | 72 | 66-77 |
| Telecommunications | 38 | 72 | 64-77 |
| Other | 69 | 71 | 66-79 |
| E-CommerceThe inferred E-Commerce group contains 118 domains and has a median overall score of 68. | 118 | 68 | 57-75 |
| GamingThe inferred Gaming group contains 116 domains and has a median overall score of 68. | 116 | 68 | 59-78 |
A clean quarterly publication
Methodology
The sample includes active, public-index-eligible domains from the automated benchmark crawler. Each domain contributes its latest completed root-page baseline audit using the current scoring system inside a 365-day window. Nyman Media domains are excluded. This is not a self-selected user-audit sample.
The funnel is deterministic and cumulative. A domain only remains at a stage if it clears that stage and every earlier one. The benchmark crawl collects audit evidence only; it does not pull customer outcomes, product usage, or payment data.
Statistics are deterministic and generated before commentary. The commentary model cannot change the sample, scores, correlations, or tables. See the full scoring and population methodology.
- This is a cross-sectional baseline and does not support claims about change over time.
- Each domain was evaluated in a single automated run.
- The population is limited to active, public-index-eligible domains in the benchmark crawler.
- Industry labels are inferred; observations are limited to supplied credible industries meeting the stated sample threshold.
- No outcome data are included, so associations and separator gaps do not establish causality.
- Structured Data is observational and is not ranked as a scored dimension.