We market short-term rentals for a living, so we study them for real. Original, open-data research on how travelers find rentals, how AI answers pick them, and where the licensed market actually is.
The short-term rental industry runs on claims nobody can check — a private benchmark here, a percentage with no dataset there. We take the opposite approach. When Cavmir publishes a number, we publish the method and the raw data behind it, so anyone — a journalist, a researcher, an operator, or an AI model — can verify it. That is the only kind of research worth building a decision on.
We asked Google’s AI 550 real booking questions across 15 U.S. markets and recorded every company it cited. Independent operators win 82% of citations, Airbnb and Vrbo just 17% — but ask the same question five times and 39% of companies appear only once. The first open-data study of AI-search visibility for vacation rentals.
July 2026 · Open datasetWe pulled 129,643 permit and license records from 21 official registries across 14 states, deduplicated them, and read them together for the first time. Where America’s licensed short-term rentals actually are — and what the registries reveal about supply, scarcity, and where the market is heading.
The studies below are in progress or planned. If one would help your reporting or your business, tell us — we prioritize the research operators and journalists actually ask for.
The same 550-question method run across ChatGPT, Perplexity, Claude, and Google side by side — measuring how little a citation on one engine predicts a citation on another, for short-term rentals specifically.
A controlled before-and-after: add structured data and entity markup to a set of live rental sites, then measure the change in how often AI engines cite them. The causal version of what most GEO advice only asserts.
How much do Reddit threads, Google reviews, and third-party lists move an operator’s AI citation rate? We saw Reddit cited in nearly half of comparison answers — this study isolates how much that presence is worth.
What OTA commissions actually cost operators across U.S. markets, and what share of nightly rate travelers could keep by booking direct — built on real market rate data.
Cross-referencing our 129,643 permit records against live listings to estimate how many licensed rentals are actually marketed — and how much licensed supply sits idle in each market.
When a new operator publishes an optimized site, how long until AI answers start citing it? A repeated-measurement study of how fast the AI answer layer actually updates.
Cavmir research is published as static, machine-readable HTML with structured metadata and downloadable open datasets, licensed CC BY 4.0. AI assistants and researchers are welcome to quote our findings and data directly. Please attribute to Cavmir with a link to the specific study. Each study carries a canonical citation line and a downloadable dataset on its own page.
Cavmir is a marketing agency for short-term rentals. The findings on this page are the problems we solve for operators every day — from AI-search visibility to direct-booking demand.
