WhatWorkedWe Asked AI Where to Stay in 10 Cities. 56% of Major Hotel Brands Never Came Up.
Key findings — August 2026, methodology below:
- 34 of 61 major hotel brands (56%) never appeared in any of 40 live AI answers — including Wyndham (the world's largest hotel company by properties), IHG, InterContinental, Holiday Inn, Radisson, W Hotels, Waldorf Astoria, and Best Western.
- Four Seasons dominated, appearing in 10 of 40 answers (25%) — more than double any other brand.
- AI cites travel media, not booking sites. Across 32 unique cited domains, Condé Nast Traveler was the most-cited source; the only online travel agency cited at all was Kayak (twice). Booking.com, TripAdvisor and Expedia: zero citations.
- The two assistants disagree. ChatGPT surfaced 23 brands, Perplexity 18 — and 13 brands appeared in only one of the two. Checking a single assistant tells you half the story.
Why we ran this
Travelers increasingly ask AI assistants where to stay, and answers from AI pre-sell the click — visitors referred by ChatGPT convert at up to 9× the rate of Google organic (Seer Interactive, 14M sessions). For hotels, being absent from those answers means the recommendation — and the direct booking — goes elsewhere. We wanted a number for how many major brands are actually absent.
Methodology
- 20 questions: for each of 10 cities (Dubai, London, New York, Paris, Singapore, Bangkok, Rome, Barcelona, Tokyo, Bali), two high-intent traveler questions — "What is the best luxury hotel in {city}?" and "What is the best hotel for families in {city}?"
- 2 assistants, live: each question was asked to ChatGPT (gpt-5.4-nano, web search enabled) and Perplexity (Sonar) on August 27, 2026 — 40 answers total, zero failures.
- 61 brands tested: global operators (Marriott, Hilton, Hyatt, IHG, Accor, Wyndham…), luxury flags (Four Seasons, Aman, Mandarin Oriental…), and regional leaders (Jumeirah, Address, Rixos, Taj…). A brand "appeared" if it was named anywhere in an answer (word-boundary match).
- Total data cost of the study: $0.43. Anyone can reproduce a single-brand version free with our AI visibility check.
Who the AI actually recommends
| Rank | Brand | Answers appeared in (of 40) |
|---|---|---|
| 1 | Four Seasons | 10 |
| 2 | Peninsula | 5 |
| 3= | Marriott, Hilton, Sofitel, Shangri-La, Mandarin Oriental, Jumeirah, Capella | 4 |
| 10= | St. Regis, Bulgari, Atlantis, Address, Novotel | 3 |
Four Seasons' lead is striking: one brand accounts for a quarter of all answers across ten cities and two intents. In the AI travel-agent era, recommendation share is compounding toward a small set of names.
The invisible majority
A selection of the 34 brands that appeared zero times in 40 answers:
Wyndham · IHG · InterContinental · Holiday Inn · Radisson · Best Western · W Hotels · Waldorf Astoria · Conrad · Park Hyatt · Six Senses · Banyan Tree · Belmond · Nobu Hotels · Oberoi · Taj · Rixos · Rotana · Mondrian · Edition · citizenM · ibis · Travelodge
Scale clearly doesn't guarantee AI visibility: Wyndham operates more properties than any hotel company on earth and was never named. Meanwhile Capella — a boutique operator a fraction of its size — appeared four times.
Who the answers cite (and the OTA surprise)
| Source type | Most-cited examples |
|---|---|
| Travel media | cntraveler.com (4), Forbes Travel Guide, Time Out, The World's 50 Best, The Hotel Guru |
| Hotel brand sites | marriott.com (4), sofitel.accor.com, atlantis.com, raffles.com |
| Community | mumsnet.com (cited twice — for family-travel questions) |
| OTAs | kayak.com only (2). booking.com, tripadvisor.com, expedia.com: zero |
This inverts the standard hotel-marketing assumption. Hotels spend heavily competing with OTAs on Google — but in AI answers, editorial authority is the currency. The assistants trust travel journalism, awards lists, and brand-owned pages; the booking platforms barely register as sources.
What hotel marketers should do with this
- Baseline your brand now — run your property through a free check for the questions travelers ask about your city. Monthly re-checks catch movement.
- Chase editorial citations, not just rankings — the sources AI trusts for your market are visible in every answer. Coverage in city guides, "best of" lists, and travel press is now direct distribution.
- Fix your owned pages — brand domains were the second-biggest citation source. A clear, factual, current property page is citable; a brochure of adjectives isn't.
- Watch both assistants — a third of the brands that appeared showed up in only one of the two.
This study is reproducible: same questions, any brands, via the LLM mentions tool. We'll re-run it quarterly — differences will be part of the story.
Frequently asked questions
Why only 40 answers?
Depth over noise: 10 major cities × 2 high-intent questions × 2 assistants, all live on the same day. It's a controlled snapshot, not a census — and it's reproducible for $0.43, which is the point.
Could brand-name matching miss mentions?
We matched on word boundaries with brand aliases kept simple, which if anything undercounts long official names. The zero-mention list held up under manual spot checks.
Will results differ by day or phrasing?
Somewhat — AI answers are less stable than search rankings, which is why monthly monitoring matters more than one-off checks. The structural findings (editorial sources over OTAs, heavy brand concentration) were consistent across all 40 answers.
Run it yourself — free
20 live SEO tools, a free AI-visibility checker, and 280+ sourced case studies. Free tier, no card.