The short answer. We ran our free AEO audit on 152 London small business websites across restaurants, gyms, cafes, contractors, electricians, and renovation. 123 came back with a usable result. Zero scored Strong. The average was 44%. Three quarters don't have an llms.txt file. Only 7% have a named author on the page. 97% are not blocking AI crawlers, which means the problem isn't access. It's that nobody set the site up to be cited in the first place. The whole local market is invisible to ChatGPT, Perplexity, Google AI Overviews, and Claude right now. Whoever fixes this first owns London's AI citations for the next two years.
Roughly 30% of searches now get answered without a click on a Google blue link. ChatGPT, Perplexity, Google AI Overviews, and Claude all crawl the web and pick whose site to cite when someone asks them a question. The signals that make a site citable are well-known: llms.txt, JSON-LD schema, a named author, depth on the page, AI crawler permissions. What nobody had measured was how many small businesses outside the top 100 brands actually ship those signals.
So we measured. This is the result.
We pulled 152 London small business sites by scraping 13 public listicles ("best London restaurants 2026", "best London gyms", "best general contractors in London", and so on) across six industries: restaurants, fitness, cafes, general contractors, construction, and electricians. We grabbed every external link, threw out social platforms (Facebook, Instagram, X), aggregator domains (TripAdvisor, OpenTable, Yelp), and infrastructure noise (Google, Cloudflare, jQuery), and deduped to one entry per business.
Then we ran every site through the same free AEO audit we use on clients. It pulls the homepage, llms.txt, robots.txt, and sitemap.xml, and grades 14 signals on a tiered rubric out of 88 points. 123 of the 152 came back clean. The other 29 had DNS failures, 403s, or timeouts we couldn't resolve at the time. Methodology section at the bottom has the exact sources, the audit code, and a reproducible script if you want to run the same thing.
0%
of audited London SMBs scored Strong on AEO
Of the 123 sites we audited, zero scored Strong (85% or higher on the 88-point rubric). The best one landed in Decent (65 to 84%). The average was 44%.
For context: our own site scores 85% on the same rubric. We're docked two points on purpose because we won't ship FAQ schema without matching visible Q&A on the page (a thing Google will deindex you for, and a thing the audit is supposed to flag honestly even on us). So Strong is reachable. Nobody in the London sample is reaching it.
The 123 sites broke down across the four tiers like this:
Translation: 92% of London small business sites are below Decent, meaning they're missing several core AEO signals. 35% sit in Invisible (under 40% of points), meaning AI engines won't reliably surface them even when the query is a dead match for what they sell.
The audit grades 14 individual signals. Each row below shows the percent of London sites that "passed" each check (passing = scoring at least 60% of the available points for that signal).
97% of London small business sites allow the major AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot, anthropic-ai) in their robots.txt. That's the strongest signal in the data. Most London SMB owners (and the agencies they hire) haven't blocked AI access. So the problem isn't gatekeeping. It's that nobody built the rest of the setup.
0% have SpeakableSpecification. Not one London site in the sample tells voice assistants and AI engines which parts of the page to read aloud or quote. It's a small bonus check (2 points), but absolute zero is striking. Voice search and AI-narrated answers are growing categories, and London businesses are not opting in.
3% have FAQPage schema with quality answers. FAQ schema is one of the highest-leverage signals because AI engines explicitly look at structured Q&A when answering question-format queries. 97% of London SMBs are leaving this on the table. To be fair, you can't just slap FAQ schema on a page with no visible Q&A. Google will deindex you for it. The right fix is to add a real FAQ section first, then ship the schema that matches it.
7% have a named author / E-E-A-T signal. Both Google and the AI engines weight author credentials when judging whether content is trustworthy. The fix is dirt-cheap: a <meta name="author"> tag, or a Person JSON-LD block with a short bio. 93% of London SMBs ship pages with no human attached.
24% have an llms.txt file. The single highest-leverage AEO signal a small business can ship in 2026, and 76% of London SMBs don't have one. This is the easiest, highest-ROI fix on the entire list. We have a full guide on what llms.txt is and how to write one.
31% have BreadcrumbList schema. Helps AI engines understand the structure of your site and link to the right level. Most CMS templates can add this automatically. Almost none of the audited London sites have done it.
42% have a meta description in the 120 to 160 character range. Either missing entirely, too short to fill the search snippet, or too long and getting truncated. This is a five-minute fix per page that compounds across every search appearance the site ever gets.
49% ship a page with more than one H1 (or none at all). A clean single H1 is the most basic SEO instruction on the internet, and half of London SMB sites get it wrong. Almost always a CMS template issue.
50% have shallow Organization schema. Even when a London site ships Organization JSON-LD, the average filled-in field count is low. Name and URL show up, then nothing: no logo, no address, no contactPoint, no sameAs. Average score on this check across the sample was 3.7 out of 10. AI engines use these fields to verify the brand and tell it apart from other brands with similar names.
These numbers aren't just a technical SEO scorecard. They're a measure of how visible London small businesses are on the surface where 30% of search is now happening.
When someone asks Perplexity "best plumbers in London", or asks ChatGPT "where should I get coffee in Hackney", or types "london marketing agency that knows AI" into Google's AI Overview box, the engine has to pick which sources to cite. A site with a structured llms.txt, deep Organization schema, named authors, and clean meta data is dramatically more likely to be picked than a site without any of that. The difference isn't 10%. It's 5x or 10x.
92% of London small businesses are below Decent. The gap between the top of the London AEO market and the average London business isn't narrow. It's several years of compounding. The brands that fix this in the next twelve months will own AI citations in their category for the next two to three years.
Three concrete moves, in order of leverage.
Sample. 152 unique London small business website origins, scraped from 13 public London industry listicles published in 2025 and 2026. Industries covered: restaurants (Avenue London's 2026 Best 13, The Best London's New Restaurants), gyms and fitness (The Best London's Gym Memberships, Rank London's 2026 Top Gyms), cafes and coffee (The Best London's Cafes and Coffee), general contractors (The Best London's General Contractors, Construction Companies, London Contractors, Renovation Contractors), electricians (The Best London's Electricians), and renovation specialists (Clever British, Elevate London). Of 152 origins, 123 returned a successful audit. The other 29 failed at the network layer (DNS, timeout, 403). The 19% failure rate is normal for public scraping at scale, and there's no reason to think it skews the sample one way or the other.
Audit tool. Every site was processed through the free TigerYellow AEO Audit (the same one available at tgylw.com/tools/aeo-audit). It fetches the homepage, llms.txt, robots.txt, and sitemap.xml from the same domain, then runs 14 graded checks on a tiered rubric out of 88 points. The audit source code lives in our public codebase. The bulk-audit script that ran this study is at tools/bulk-audit.js in the same repo, and the URL list at data/london-urls.txt.
Tier definitions. Strong = 85% or higher. Decent = 65 to 84%. Behind = 40 to 64%. Invisible = under 40%. These match the live audit's tier output.
Limitations. This is a single-URL audit per site (the homepage), not a full-site crawl. Domain-level signals (llms.txt, sitemap, robots.txt) are checked once per site. Page-level signals are only evaluated on the homepage, so a site with weaker AEO on the homepage and stronger AEO on inner pages would score worse here than it deserves. The sample is also biased toward businesses the listicle authors found worth recommending, so the long tail of London SMBs without strong reputational signals is almost certainly worse, not better, than what we measured.
Replication. If you want to reproduce this on another market, the audit tool is free and the bulk script is open source. Publish your analysis and send us the URL via the contact page. We'll link back from this methodology section.
We'll repeat this audit every year to track the trend, and we're expanding it to the other markets we work in (London Canary Wharf, Plovdiv) next. If you're a London marketer, journalist, or business owner who wants to dig into the data yourself, we're open-sourcing the per-site results. Ping us via the contact page and we'll send you the JSON.
If you want help moving your London business out of Behind and into the empty-but-claimable Strong tier, our SEO and AEO retainer covers exactly the four high-leverage fixes above: llms.txt buildout, Organization schema, Person schema with E-E-A-T, AI crawler permissions audit. Plus the longer-term content production that compounds those wins.
We scraped 13 public London listicles across restaurants, fitness, cafes, contractors, construction, electricians, and renovation. From each listicle we pulled every external link, filtered out social platforms and infrastructure domains (Google, Facebook, Instagram, TripAdvisor, Yelp, etc.), deduped by origin, and validated the resulting list. 152 origins went in, 123 came back with a usable audit. The other 29 had connectivity issues we couldn't resolve at the time.
The audit grades each site on 14 AI search readiness signals (llms.txt, JSON-LD schema, AI crawler permissions, named author / E-E-A-T, single H1, meta description length, Open Graph, sitemap, content depth, plus bonus signals) on a tiered rubric out of 88 points. Strong = 85% or higher. None of the 123 London small business sites we audited cleared that bar. The best one landed in Decent (65 to 84%). The average was 44%.
llms.txt is a guide file at a website's root that tells AI engines (ChatGPT, Perplexity, Google AI Overviews, Claude) what the site is about and which pages to prioritize. It's the highest-leverage AEO signal a site can ship in 2026. 76% of the London sites we audited don't have one. The audit grades llms.txt across six sub-criteria, so even sites that do have it often score barebones because they wrote one good paragraph and called it done.
Yes. 97% of the London sites we audited allow major AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot, anthropic-ai) in their robots.txt. That's the strongest baseline in the data. Most London SMBs aren't blocking AI engines. The problem isn't access. It's that nobody set the site up to be cited.
There isn't a good baseline yet. Most existing AEO research focuses on enterprise brands or Fortune 500 companies, where the structural setup is much further along. Small business AEO data is rare. Our hunch is that London numbers track with most North American secondary markets where AEO awareness hasn't hit SMB owners yet. We'll repeat this audit every year to track it.
Three things. First, run your own site through the free AEO audit at tgylw.com/tools/aeo-audit and see where you land against the London average of 44%. Second, prioritize the highest-leverage fixes: ship a structured llms.txt, add Organization JSON-LD with all the fields filled in, add a named author with Person schema on key pages, and make sure robots.txt allows AI crawlers. Third, treat this as a head start. With zero London sites in Strong today, even reaching Decent puts you ahead of 92% of the local market.
Where do you stand?
See how your site compares to the 44% London average. No email gate. The same audit we ran on the 152 sites in this study.
Run the Audit