We scored 10 of London's top luxury hotels for AI readiness. None passed.

When a guest asks ChatGPT to recommend a five-star London hotel for an anniversary, the answer doesn't come from a glossy brochure. It comes from whatever the model can read, parse, and trust on the open web. We scanned 10 of London's most prestigious hotels to see how ready their websites are for that kind of discovery. The headline finding: no hotel in the set scored above 60 out of 100.

The numbers

Ten hotels. Average score 48 out of 100. The highest score was 57 (The Ritz London). The lowest was 38 (The Goring). Four hotels scored below 50. None reached the "good" band of 70 or above.

To put that in context, our scoring rubric awards points generously for the basics: having a valid robots.txt, allowing AI retrieval bots, using clear HTML structure. A score in the high 40s means a site is doing the table-stakes work but missing most of the signals that actually drive AI citations.

Why this matters now

Luxury hospitality has spent two decades optimising for human visitors: visual design, booking flows, brand storytelling. AI search rewards a different brief. It wants crawlable HTML, structured data, clear factual content, and verifiable signals of authority. A site can be breathtakingly designed and still be effectively invisible to an LLM.

For a luxury hotel, the stakes are concrete. High-intent guests asking generative search engines for restaurant suggestions, suite recommendations, or "the best afternoon tea in Mayfair" need to be told about your property, by name, with the right facts. Right now, the model has thin material to work with.

How we scored them

Every site is run through the same 34-check rubric, grouped into five categories: crawler access, structured data, content clarity, AI-specific signals, and authority & trust. Each check is weighted by how much it actually moves the needle for AI retrieval today. The rubric is published in full on the methodology page.

The full ranking, every individual check, and per-site scorecards live on the benchmark page. What follows is the editorial cut: the patterns that surprised us.

The category breakdown

Averaging the 10 hotels across the five scoring categories:

• Crawler access: 74/100. The strongest category. Most hotels let the retrieval bots in.
• Content clarity: 60/100. Adequate. Most sites have readable HTML and sensible headings.
• Structured data: 45/100. The first real gap. Schema is patchy.
• AI-specific signals: 26/100. Almost nobody is treating AI bots differently from training crawlers, or publishing an llms.txt.
• Authority & trust: 23/100. The worst category in the set. Few sites surface review signals, awards, or press citations in machine-readable form.

Three patterns that hold across the set

1. Crawler access is mostly fine. The doors are open.

Every hotel in our set allows AI retrieval bots to read the site. Robots.txt is broadly permissive. This is the easy stuff and the sector has, by and large, got it right. The problem is not that the bots can't reach the content. The problem is what they find when they get there.

2. Organization schema is missing almost everywhere.

Nine out of ten hotels lack a proper Organization schema block. This is the structured data tag that tells a crawler "this is a hotel, here is its legal name, here is its address, here are its rooms and restaurants." Without it, the LLM is reverse-engineering your identity from prose. Adding a single JSON-LD block to the homepage is a one-day engineering task with outsized impact.

3. AI-specific signals are an open goal.

Nine out of ten sites do not differentiate between retrieval bots (which serve real-time AI answers) and training bots (which scrape content for model training). Most operators we speak to want to permit one and restrict the other. The mechanism exists. Almost nobody is using it. This is one of the highest-leverage changes on the entire rubric.

What the top performers get right

The hotels at the top of the ranking share three habits:

• Their pages answer questions. Page intros lead with facts, not poetry. The first paragraph tells you what the property is, where it is, and what it offers.
• They mark up what matters. Even where structured data is thin, the top scorers at least cover the basics: Organization, Hotel, Restaurant tags in JSON-LD.
• Their robots.txt is clean and modern. No legacy rules quietly blocking newer AI user agents.

What every hotel in this set can fix this quarter

For the entire group, three changes would move the needle materially, none of which require a redesign:

1. Add Organization and Hotel schema in JSON-LD. One-day engineering task. Single biggest leverage point on the rubric.
2. Differentiate AI retrieval bots from training bots in robots.txt. Most operators want to allow one and gate the other. Make the call explicit. We recommend allowing retrieval bots (GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot) and being deliberate about training bots like CCBot.
3. Publish an llms.txt file. A plain-text page summarising what the site is, who it serves, and where the canonical facts live. Takes an hour. Improves how LLMs cite the property.

See the full ranking

Every score, every check, every per-site finding is published on the luxury London hotels benchmark. If you operate one of these properties (or compete with them), it is worth ten minutes of your morning.