How AI decides which hotel wins the answer
Original research on how ChatGPT, Gemini, Claude and others select hotels — the biases they carry, the features they weigh, and what independent hotels can do about it.
95.9% of hotels have no AI crawler rules
104,214 hotels parsed. 95.9% have zero AI-specific rules. Most hotels are open to AI crawlers by default — but not strategically.
Only 5.8% of hotel websites have llms.txt
104,214 hotels scanned. 5.8% adoption. Zero AI crawler requests in server logs. The honest state of llms.txt in hospitality.
Geographic context changes which hotel wins
A driving-times table prepended to a plain query reshapes ChatGPT’s recommendations — a case study on how context alters AI’s pick.
Agents inherit positional bias too
Claude, Gemini, and GPT agents tested via OpenCode and browser UIs. Even with tools, the top of the list wins.
Fixing positional bias: what actually works
15 mitigation strategies tested head-to-head across 15,616 trials. Only one category of intervention actually worked.
Would a human pick the same hotel?
We compared human travelers (conjoint literature) and AI models on the same decisions. The gaps reveal where AI helps — and where it quietly fails guests.
Each AI has a different taste
ChatGPT, Claude, Gemini — compared side-by-side. Which model favors brands, which favors independents, and why it matters for visibility strategy.
Does the order of results matter?
Testing positional bias: whether AI models favor the first hotel they see — and how much of the effect is real vs. API noise.
AI recommends branded hotels 60% of the time
A 20-point lift from the brand name alone. We hid brand names in a paired control — the bias collapsed. Here’s what it means for independents.
What AI actually weighs when picking a hotel
Location, price, reviews, amenities — ranked by how much each one moves the AI’s choice. Hint: it’s not what you’d guess.
Subscribe to Huxo Research
One study a month. No marketing fluff — just the data, the method, and the playbook for independent hotels.