What small business websites actually score.
Every free audit we run scores a real site on five dimensions. Those scores add up to something no one else has: a first-party picture of how small business websites are actually performing. Here is the whole corpus, 79 sites deep — including the parts that make our own pitch harder.
Sites measured
79
across 6 industries
Median overall
46/100
half score below this
Weakest dimension
99%
score under 60 on AI-readiness
| Dimension | Median | Distribution | Under 60 |
|---|---|---|---|
| Speed & Core Web Vitals How fast the page becomes usable on a mid-range phone. | 50 | | 77% |
| Design & mobile experience Whether the layout holds up on the screen most visitors use. | 50 | | 94% |
| SEO fundamentals Titles, structure, schema, indexability — the basics search still runs on. | 58 | | 59% |
| AI-readiness (AEO/GEO) Whether an assistant can read the site and quote it back accurately. | 25 | | 99% |
| Lead capture Whether an interested visitor has an obvious way to start a conversation. | 40 | | 94% |
The number worth sitting with: 99% of these sites score under 60 on AI-readiness, with a median of 25. Buyers increasingly start at an assistant rather than a search box, and a site an assistant can't read accurately doesn't get quoted back — it simply isn't in the answer.
Still measuring
These industries are in the corpus above but don't yet have 25 sites of their own, so they get no page of their own. We'd rather say "not yet" than publish a median built on a dozen sites.
- Construction & Contractors 24/25
- Medical & Dental Practices 24/25
- Schools & Education 19/25
How these numbers were made
The instrument. Every site was scored by the same engine that runs our free audit. Speed and the SEO category come from Google PageSpeed Insights when that call returns a score. When it doesn't, those scores are stored as not measured — they are not estimated from the HTML. Design, AI-readiness, and lead capture are the model's judgment of the page's visible text, informed by deterministic checks (title, meta description, headings, schema, forms). They are not graded against a fixed published rubric. A manufacturing audit also records a separate 10-point buyer checklist; that checklist is not an input to these medians. The scores here are the ones stored for each site.
The sample, stated plainly. This is not a random sample of the web. Most of these sites came from cold-outreach prospect lists we assembled, plus sites whose owners requested an audit themselves. That means it skews toward small and mid-sized businesses in the industries we serve, and away from companies with in-house web teams. Treat it as "79 real SMB sites we measured", not as a census.
What we threw out. Sites are deduplicated by domain, so one site scored twice counts once. Our own test audits are excluded. Re-scores of sites we rebuilt are excluded entirely — including them would flatter the baseline with our own work. And on the two dimensions judged by a language model, any run where the model was unavailable and the engine fell back to a canned score is dropped, because a canned score would drag the median toward the middle. A score that was never measured is absent, not stored as zero.
Dates and freshness. Measurements run July 17, 2026 to September 27, 2026. Aggregates regenerated November 9, 2026. The corpus grows every week; these pages are rebuilt from it, and a median moving is the data changing, not the story.
Want your own number?
The audit is free and takes about a minute. You'll get the same five scores these medians are built from — plus what it would cost to fix what it finds.
Score my site — free