Which hotel does AI recommend? Rating +31.6pp, high price −30.0, list position worth $12 a night
A randomised audit quantified the signal weights behind AI hotel picks: rating and price dominate, eco-certification is overweighted and management responses are ignored.
Key numbers
- Design: a pre-specified choice-based conjoint audit across personas, prompt templates and twelve open-weight and proprietary models, asking assistants to choose among five hotels with guest rating, review volume and recency, management response, chain affiliation, price, eco-certification and list position independently randomised.
- Result: guest rating and price dominate — a top rating raises the probability of selection by 31.6 percentage points, while a high price lowers it by 30.0.
- Bias: models over-weight eco-certification and ignore management response, while list position — a content-free artefact — shifts recommendations causally, worth about $12 per night.
- Stated reasons track the revealed weights only imperfectly.
Method and sample
arXiv paper “Whose hotel does the AI recommend? An algorithm audit of reputation signals in LLM-assisted hotel selection” (2026-06-15): a pre-specified algorithm audit using a randomised choice-based conjoint to estimate the average marginal effect of each reputation signal on recommendation probability across twelve models.
What it means for brands
For accommodation, the levers in AI are now quantified: rating first, price second, eco-certification carrying a premium, management responses barely moving anything. And list position — not a content signal at all — carries a price, which means sources that rank higher keep collecting outsized recommendations.
How SourceSight handles it
For travel and hotel projects SourceSight checks how AI describes you for “nearby, how much, availability, who is it right for”, and prioritises review platforms and directories in the off-site list — ratings and review volume are exactly the strongest signals models actually use.
Limits
The audit runs in a controlled choice setting with a limited set of hotels and signals; the marginal effects are causal estimates, not booking conversion rates.
References
Run it on your own project
What the same prompts return across 10 engines is only knowable by running them.
Related research
Cross-category recommendations: +36% add-to-cart on item pages, +15% on cart pages
Recommending general merchandise to grocery shoppers — a milk frother with the milk — has long been hard; LLM retrieval plus real-time cart context moved production A/B metrics.
Industries · 2026-10-07Home services: 22% of homeowners already use AI to find a provider, and 80% of businesses don't know how to be found
The local services funnel is moving upstream: before a homeowner opens a website or calls, AI has already synthesised reviews, reputation and service descriptions into one answer.