Five hotels take half the AI answers — while hotels pay OTAs 15–25% and AI visibility tools start at $17 a month

Asked (summary):

What evidence from 2025–2026 shows hotels — especially independent and luxury properties — are invisible or misrepresented in AI travel recommendations from ChatGPT, Google AI, Perplexity and Claude? How are travelers using AI agents to choose and book hotels? What commissions do hotels pay OTAs like Booking.com and Expedia, and what do tools for AI visibility, content distribution or direct bookings cost? List specific numbers, named studies and key data fields.

This report draws on 145 selected evidence rows scanned on 24 October 2026 and dated 2025–2026: 35 visibility and misrepresentation findings, 35 traveler-behavior data points, 30 OTA commission estimates and 45 priced tools. Most audits are commercial or vendor-produced, not standardized academic benchmarks — methods, prompt sets and dates differ, so scores are shown side by side, never combined.

AI recommendations concentrate on a handful of hotels in every audited market

Share of all AI hotel mentions captured by the top properties in each 2025–2026 audit. Each bar is a separate study with its own method and prompt set — compare the pattern, not the exact scores. Hover a bar for sample sizes and the invisibility findings behind it.

Bars: top-group share of mentions, %. Grey remainder: everything the rest of the market splits. All six audits are commercial (LuxDirect, AGR); the one academic randomized audit appears in the findings below.

Travelers use AI heavily to research and choose — almost none hand it the payment

The funnel from research to autonomous booking, from the largest 2025–2026 surveys. Colour separates what people actually did from what they say they would accept. Hover for the survey, sample and caveats.

Blue: observed use. Orange: stated preference or willingness, not behavior. Each bar is a different survey with its own denominator — they are not stages of one sample.

What hotels pay OTAs: headline rates vs effective cost

Indicative public estimates only — Booking.com and Expedia publish no universal rate, and a hotel's own extranet or contract is authoritative. Ranges shift with market, volume, collect model and merchandising programs.

Horizontal spans: commission as % of booking value. Effective rows include hotel-funded loyalty discounts, visibility surcharges and payment fees.

Tool pricing: $17 monitoring to $4,500 enterprise platforms

Named examples from the 45 priced rows, grouped by purpose. Prices are third-party observations or starting prices checked in 2026; most exclude setup, extra prompts, engines, domains, ad spend or commission.

A randomized 12-model conjoint audit found a top guest rating raised recommendation probability by 31.6 points and high price cut it by 30.0 — while list order moved choices but was omitted from model explanations, so explanations are unreliable diagnostics. link.springer.com

Misrepresentation is documented, not hypothetical: AGR captured recommendations for closed Miami properties (Mandarin Oriental Miami, Ritz-Carlton Bal Harbour), and a TravelAnywhere audit of 100 ChatGPT travel recommendations found only 75% accuracy for hotels, with rebranded names and wrong pricing tiers common. blog.palmtree.ai

Lighthouse's 4,545-prompt study across nine destinations surfaced only 2,721 unique properties in nearly 50,000 mentions — market coverage of 10% in Tokyo and 13% in Paris — and a SIGNAL NOIR audit of 17 luxury properties scored average visibility ~33/100 (US) and ~28/100 (UK). aeo.how

The fix is often basic plumbing: 14 of 15 Manchester independents lacked Hotel/LodgingBusiness schema, a Nokumo audit found only 76 of 1,337 accommodation sites (5.7%) detected in any AI-model response, and AI hotel answers route to OTAs — Booking.com appeared in 95.3% of AI hotel queries in one five-country citation study. openlens.com

All 145 evidence rows

Set	Study / program / tool	Date / category	Scope	Metric / rate / price	Source

Key data fields glossary

Study / fieldwork date — when prompts ran or respondents answered; AI answers drift within hours

Platform, version & mode — ChatGPT vs Gemini vs Perplexity; training-data vs live-web; logged-in vs logged-out

Geography / locale — market tested and locale of the session

Property universe & segment — how many hotels of which type were eligible to appear

Sample size — prompts issued, valid responses, deduplication rule

Inclusion / share of voice / rank — whether a hotel appears at all, its mention share, its position

Citation / link destination — OTA, metasearch or the hotel's own domain

Accuracy / error type — closed property, rebrand, wrong pricing tier, stale facts

Journey stage — research, comparison, confirmation, in-trip, payment

Respondent base — the denominator behind every percentage

Observed vs stated — did people do it, or say they would

Booking handoff & payment autonomy — who completes the transaction

Commission rate & charge basis — % of what: net rate, booking value, taxes in or out

Visibility / loyalty increment — Genius, Preferred, Accelerator cost on top of base

Tool price & billing unit — per month, location, domain, prompt pack

Included engines / prompts / properties — what the plan actually covers

Setup / add-ons — implementation fees, extra models, ad spend excluded from headline price

Source URL — single-sourced and vendor-reported entries are marked as such

Method: 145 selected rows from four result sets (35 AI visibility/misrepresentation findings, 35 traveler-behavior data points, 30 OTA commission estimates, 45 priced tool plans), scanned 2026-10-24, covering evidence dated 2025–2026. Visibility bars show the top-group share of AI mentions per audit; funnel bars show per-survey percentages with their own denominators; OTA spans show commission as % of booking value. Much of the evidence is commercially produced and some prices and commission estimates are single-sourced or repeated across secondary industry sites — treat them as source-reported, not independent confirmations. Long metric texts are truncated in the table for space; follow each source link for the full figure.

This report was generated automatically by Keenable SELECT at a user's request, from publicly available web sources linked herein. Keenable does not review, verify, or endorse its contents and makes no representation as to accuracy, completeness, or timeliness; AI-based extraction may contain errors. Nothing in this report is investment, legal, financial, or other professional advice. All trademarks and referenced content remain the property of their respective owners; no affiliation or endorsement is implied. To report an error, rights concern, or request removal: legal@keenable.ai.