These are the aggregate results of 2,528 real small-business website scans, published in full in our AI-visibility directory. Each site was fetched once, as an AI crawler fetches it — raw HTML, no JavaScript — and each statistic below counts what that fetch actually contained. Groups smaller than 30 sites are never reported, and no individual business is ever named.
Across the whole corpus the average health score is 53 out of 100, and 33% of sites score under 50 — the range where sites measurably lose visitors and visibility. The building blocks AI answers are made from were present at these rates:
| Measured on the homepage or domain | Share of 2,528 sites |
|---|---|
| Structured (schema.org) business data present | 46% |
| SPF record on the domain | 53% |
| DMARC record on the domain | 35% |
| Mobile viewport declared | 74% |
| No meta description for answers to quote | 29% |
| Page title missing entirely | 4% |
Put the other way: a majority of small-business sites give an AI assistant no structured data to quote, and about one in three gives it no description either.
Averages hide the range. In this corpus, 71% of travel businesses had schema markup — the highest of any sector — while uncategorised local businesses sat at 11% and florists at 16%. Pharmacies had the worst average health score, 46 out of 100, with 60% of them under 50. Movers had the lowest DMARC adoption at 13% and the lowest average speed score at 23. Every sector and region cut, with its own table, is published in the directory.
Three published third-party numbers frame why this matters. BrightLocal's 2026 research found 45% of consumers now use AI to find local businesses — up from 6% a year earlier. SOCi's study of 350,000 locations found just 1.2% of local businesses are ever recommended by ChatGPT, against 35.9% visible in Google's local pack. And Vercel's 2026 crawler analysis found 52% of all crawler traffic is now AI crawlers — and the largest of them don't render JavaScript. These are their measurements, not ours; ours are the scan aggregates above.
One scan is one HTTP fetch of the homepage plus public DNS lookups — no JavaScript executed, no login, no forms. A check that could not run reports nothing rather than a guess, a metric appears in a group only when it was actually measured on at least 30 sites there, and no third-party benchmark is ever blended into our numbers. The full method is on the methodology page.
In our corpus of 2,528 scanned small-business websites, 46% had structured (schema.org) business data on the homepage. The spread by sector runs from 71% for travel businesses down to 11% for uncategorised local businesses.
From 2,528 real website scans we ran and publish as aggregates in our AI-visibility directory. Each site was fetched once, as an AI crawler fetches it, and every statistic counts what that fetch contained. No individual business is named, and groups under 30 sites are never reported.
BrightLocal's 2026 research puts it at 45% of consumers, up from 6% a year earlier. That is their measurement, cited on our homepage; our own numbers are scan aggregates, and we never blend the two.
Yes — the AI-visibility directory publishes every cut with at least 30 scanned sites: 24 sectors from restaurants to accounting, and regions including the East Village, West Village, Chelsea and Edison, NJ. Each cut page carries its own table.
Where does your site stand? The scan is free, takes about 60 seconds, and needs no account. It fetches your homepage exactly the way GPTBot, ClaudeBot and PerplexityBot do and tells you what they receive.