Every AEO/AI-visibility statistic in NetWebMedia's own marketing used to be stylized — directionally true, never independently measured. We fixed that. This page reports what we actually found when we pointed our own free-audit scoring engine at 460 real small-business websites, no modeling, no rounding up.
The dead-domain finding
Before scoring anything, we had to fetch it — and 412 of the 927 website addresses we attempted (44%) failed to resolve at all, confirmed against two independent DNS resolvers. Not slow, not broken layout — the domain itself is gone. These businesses are still operating (they're listed in current business directories); their website simply isn't there anymore. We excluded every dead domain from the scoring average entirely — it is not counted as a 0, it is not counted at all. The 36/100 average below is measured only across the 460 sites that were actually live and reachable on 2026-09-22.
Where the 460 live sites actually stand
Score distribution across the 460 reachable, scored sites (0-100 scale, higher = more AI-readable):
Not one of the 460 scored 80 or above. The highest score in the entire sample was 70/100.
What's actually missing
Break the score into its five components and the weak spots are consistent and specific — not a vague "needs improvement," a concrete list of what's absent:
The pattern: small-business sites generally have enough real content (topical depth is the strongest score), but almost none of it is marked up in a way a machine can parse — no JSON-LD, no FAQ schema, no Organization entity, nothing pointing outward to a source a model could verify. 45% of the 460 live sites carry zero structured data of any kind. That is not a ranking problem. It's an invisibility problem: an AI assistant reading the raw HTML of most of these pages has almost nothing structured to extract.
Methodology
- Sample source: a corpus of 1,080 real small businesses catalogued across 17 Chilean cities, spanning NetWebMedia's 14 target industry niches (tourism, restaurants, health, beauty, legal, real estate, automotive, education, events, financial services, home services, wine/agriculture, local specialists, general SMB).
- Attempted: 927 of those businesses had a parseable website address on file.
- Scored: 460 addresses returned a live page and were run through the same heuristic AEO-readiness engine behind netwebmedia.com/free-audit.html — no AI model in the loop, pure structural analysis of the served HTML (JSON-LD presence and type, entity/brand signals, word count and heading structure, outbound trust links, dateline/provenance markers).
- Excluded, not zeroed: 467 addresses were left out of every average — 412 for failing DNS resolution on two independent resolvers, the rest for fetch failures, SSL errors, or HTTP error responses. None were scored as 0.
- Date: scan run 2026-09-22. Small-business sites change; this is a snapshot, not a permanent ranking.
- What this does NOT measure: whether ChatGPT, Claude, or Perplexity would actually recommend a given business in a live conversation. That requires running real buyer-intent queries through a model with live web search — a separate, deeper tier of NetWebMedia's free audit that was not run across this full sample (each run costs real API spend). The scores above are AI-readiness, not AI-recommendation testing.
Cite this
NetWebMedia. "The State of SMB AI Visibility, 2026." netwebmedia.com/state-of-smb-ai-visibility-2026.html, September 22, 2026. n=460 scored (927 attempted, 1,080-business corpus).
Where does your business score?
Same engine, your own site, free, no login, results in about a minute.
Run My Free Audit →