Where does AirDNA get its data?

The blocked-versus-booked classification problem

This is the load-bearing technical challenge of all aggregator-style STR data. Airbnb does not publicly disclose actual bookings; from outside, all you see is "this date is blocked." A blocked night could mean a real paying guest reserved it, the host pulled it for personal use, or the host paused the listing. AirDNA's models try to solve this from review timestamps (a new review near a blocked period implies a stay), price changes (rate hikes near a date often signal demand response), and historical seasonality patterns. PMS-integrated providers like Key Data sidestep this entirely because the PMS knows whether each night is sold or owner-blocked.

What changed in 2024-2026

Airbnb has, in the last two years, made calendar scraping harder. Several listing-page elements have been moved behind authenticated requests or rate-limited. AirDNA has not publicly described how this has affected its model accuracy, and the company continues to claim "97%" accuracy on its modeled metrics. That figure should be read as the company's own self-reported benchmark, not an independently audited one. For high-stakes citations, prefer language like "AirDNA estimates" or "according to AirDNA modeling."

Where the data is reliable, where to use caution

Generally trustworthy: listing counts, asking prices, location concentration, year-over-year supply changes. These are observed directly and don't require modeling.

Estimated: occupancy, revenue per available rental, average daily rate realised (versus listed). These are the outputs of the blocked-versus-booked classifier and inherit its uncertainty.

Modelled with extra layers: Rentalizer property-level revenue projections, which are estimates layered on estimates (model the comparable set, model their occupancy, then model what your specific listing would do).

How journalists should phrase AirDNA citations

The accurate inline form is "according to AirDNA, which tracks short-term rental performance by scraping public Airbnb and Vrbo listings." For a quick reference: "AirDNA estimates" or "AirDNA's MarketMinder shows" both work. Avoid presenting AirDNA numbers as observed bookings; they are modeled estimates from public data.