Ai housing recommendations often miss better rental options despite following preferences
Following the Preference, Missing the Optimum: Compliance Without Optimization in AI Housing Recommendation
Computers and SocietyInformation Retrieval
Summary
When people use AI to find apartments, the AI usually respects what they say they want, like budget or number of bedrooms. But the researchers found that the AI still often suggests places that are not the best deals—more expensive and with longer commutes—even when cheaper and better options exist. The AI follows the user's rules but doesn't always pick the optimal choice. This was tested on many made-up renter cases in New York City using real listings. The authors suggest a way to measure this problem so future AI systems can improve.
What this means in practice
- •For real estate platform developers: Detect when AI recommendation systems suggest suboptimal housing by measuring dominance rates against known rental options and improve user outcomes.
- •For ai service operators: Monitor and audit AI responses to ensure users receive optimized rental suggestions rather than only preference-compliant but inferior options.
Authors
Hsuan Lo
Abstract
Large language models are becoming the first point of contact for consumer search in domains where the stakes are material and the law is explicit. Existing audits show that models steer housing seekers by perceived identity, but none can say what a user loses when a recommender overlooks a suitable option, for want of an enumerated inventory to score omissions against. We audit AI housing recommendation against a verifiable ground truth. For each of 150 synthetic renter scenarios in New York City we build a pool of 120 real listings with known rent, bedrooms and GTFS-computed transit commute, compute the exact set satisfying the renter's stated constraints, and derive its Pareto frontier. The primary outcome assumes no utility function: a recommendation is strictly dominated if the same pool holds a listing cheaper, faster to commute from and no smaller in bedrooms. Across 9,945 calls to three models from two vendors, compliance is near-perfect (1.8% violation against a 66.6% random floor), yet 39.0% of recommendations are strictly dominated, and the dominating listing is a median 900 USD/month cheaper and 3.5 minutes closer. A within-scenario manipulation separates two capabilities usually conflated: changing one sentence moves median recommended rent by 646 USD/month in the correct direction, so preferences are honored, yet recommendations still sit 606 USD/month above the five cheapest qualifying listings on the same screen, and an unambiguous lexicographic instruction gives no improvement under equivalence testing against a pre-specified 50 USD/month bound. The gap widens with candidate-set size and replicates across OpenAI and Anthropic models to within 3 USD. We characterize the failure as compliance without optimization, propose dominance-rate instrumentation as a deployable diagnostic, and release all code, prompts and per-call results.