The 2026 report

We tested 2,148 small business websites. 767 of them did not load.

Independent businesses across 12 industries and 33 cities in Latin America, Spain and the United States. No chains. This is what we found, with the method and the caveats in plain sight.

The headline number, and what it really means

767 of 2,148 websites failed to return a page. That number deserves unpacking rather than repeating, so here is the breakdown of why they failed:

Read the top row with suspicion, and the middle two with confidence. "Connection error of another kind" includes sites that block automated requests and work perfectly for a human — we cannot tell those apart, so we do not claim them. But a 404 and a domain that no longer resolves are unambiguous: 382 businesses (17.8% of everyone we tested) are advertising a website that is definitively dead. That is the number we stand behind.

By industry

Hospitality is the worst affected and the most surprising, given how much of its business arrives online. Restaurants and hotels both lose four in ten.

Only industries with at least 40 audited sites are shown. Two smaller categories (travel agencies and psychology practices) were excluded for sample size, not for results.

By city

The geographic spread is wide enough to be interesting: Lima loses half of its listed websites, Barcelona one in five. Spanish cities cluster at the healthy end, Andean and Brazilian capitals at the other.

Only cities with 50 or more audited sites are listed.

Among the sites that do work

Setting aside the failures, 1,381 sites answered normally. They are in better shape than the headline suggests, with two clear weak points:

• 523 have no meta description — no control over the sentence Google shows beneath their name.
• 242 respond slower than 2 seconds, and 46 slower than 4. Against a 629 ms median, the slow tail is very slow.
• 140 are not built for mobile — no viewport declaration at all, which in 2026 means the page renders as a shrunken desktop layout.
• 100 run on a generic site builder template (Wix, GoDaddy, Weebly, Blogger, Google Sites, Canva or Jimdo). Not a fault in itself, but usually paired with the problems above.

Method

Businesses were collected from public map listings between May and July 2026, filtered to independents by removing a list of roughly 130 chains and franchises. For each business with a listed website we made a single request, following redirects, with an 8-second timeout and a standard desktop browser user agent, and recorded the status, the response time, and whether the returned HTML declared a viewport and a meta description.

What this method does not measure: visual design, whether the content is any good, Core Web Vitals as experienced by real users, or anything behind a login. A single request from a single location is a blunt instrument — it is good at finding sites that are broken, and poor at judging sites that are merely mediocre.

We also make no claim about how many businesses have no website. Absence from a map listing is not evidence of absence from the internet, and we would rather report a smaller number we can defend.