How we ran the UK Small Business AI Readiness Study 2026
The full method, our checks, the limits and the data. For anyone who wants to check our work.
We sampled 5,040 sites at random and visited politely
- Source: OpenStreetMap, the free map anyone can edit.
- Our copy is from 17 October 2026.
- Six groups: trades, health, food and hospitality, personal care and leisure, professional services, retail.
- Independents only. No chains or franchises (a brand tag, or one website shared by several premises).
- No branch pages, public bodies, or social media and booking pages.
- That left 73,567 eligible websites.
- Bigger firms out. After scanning, we dropped homepages saying they were part of a larger group, listed on a stock market or had thousands of staff.
- 420 per region, in all 12 UK regions. Even across the six groups. Max 10 per postcode district.
- Polite visits. Our research bot read each homepage plus robots.txt, llms.txt and, if needed, sitemap.xml.
- It obeyed robots.txt. 27 sites asked bots to stay away, so we skipped them.
- 3,671 homepages loaded properly. Dead, insecure, parked, spam or bot-blocking sites were left out.
- Same checks as our free AI readiness checker.
- “Business details” means the standard business details code on the homepage, as described on schema.org and in Google’s guide to it.
- “Answers questions” means FAQ code, or at least three question-style headings.
- Weighted so UK figures match how many eligible businesses each region and group has.
- 95% ranges (Wilson intervals) on every figure.
We checked our own work, and the figures held
- Every site that failed first time was re-tested.
- A second, separate checker agreed with ours on 99% of sites.
- 100 homepages were loaded in a real browser with JavaScript on. Not one had details that only appeared later.
- We viewed 20 screenshots ourselves. 19 were real, working sites. One showed a server error.
- Removing the bigger firms found on the second pass moved every figure by under 1 point.
Most llms.txt files were made by the website platform
llms.txt is a newer, optional file that summarises a site for AI tools.
23.4% of sites had one (95% range 21.5–25.4%).
Of those, 84.1% were made automatically by the platform or a plugin (95% range 80.2–87.3%).
We sorted files by each builder’s and plugin’s tell-tale wording.
If unsure, we didn’t count it as platform-made.
Bars show our best estimate. The thin lines show the 95% range.
See the numbers
| Group | Share | 95% range | Sites |
|---|---|---|---|
| Made by the platform or a plugin | 84.1% | 80.2–87.3% | 852 |
| Not clearly platform-made | 15.9% | 12.7–19.8% | 852 |
Hardly any sites block AI search crawlers
Only 0.4% (95% range 0.3–0.8%) had robots.txt rules turning away the crawlers AI assistants use.
Some figures didn’t hold up, so we left them out
Homepages hidden from search, content that needs JavaScript, and missing contact routes.
Our checks on these weren’t reliable enough to publish.
It’s a homepage snapshot, not the full picture
- Only businesses on OpenStreetMap with a website.
- Homepages only. Some sites put details on other pages.
- October 2026. Websites change all the time.
- These details help. They don’t guarantee Google or AI will recommend anyone.
Download the data
Download the aggregate data (CSV). Totals only, under CC BY 4.0.
Questions, or want your site removed? Email hello@aiseoagents.co.uk.
Data credits
Map data © OpenStreetMap contributors, available under the Open Database Licence.
Region boundaries: Office for National Statistics, Open Government Licence v3.0.
Contains OS data © Crown copyright and database right 2025.