Airbnb public search snapshots

On November 6, 2026, a cloud run returned 18 public London search cards for December 6–7, two adults, in 3.592 seconds. One actual row:

{
  "id": "1785620906517933133",
  "name": "Central Room Near King\u2019s Cross & Euston",
  "title": "Room in Camden",
  "url": "https://www.airbnb.com/rooms/1785620906517933133",
  "displayPrice": "$92",
  "priceQualifier": "for 1 night",
  "checkIn": "2026-12-06",
  "checkOut": "2026-12-07",
  "rating": 5,
  "reviewCount": 7,
  "searchComplete": false,
  "scrapedAt": "2026-11-06T08:31:44.752709+00:00"
}

Export a dated first-page observation for rental-market spreadsheet review. This is a bounded search snapshot, not a full market inventory. Paid demand is unproven. The example is one owner test, not a speed guarantee or a customer sale.

Input and output

{"location":"London, United Kingdom","adults":2,"maxListings":18}

Omitting both dates selects one night 30 days from UTC today. Supply both dates for a reproducible comparison. Every run resolves and saves its dates. USD and English public cards only;1–16 adults,1–28 nights, ASCII destination names. If the source rewrites the destination or stay in a way we cannot verify, the run fails explicitly. Nearby recommended listings may appear; this is not a geographic inventory.

Download JSON/CSV. Filter type=listing for cards and type=search_report for the one-page observation. The report includes source count/IDs, requested and resolved destination, stay and pagination evidence. searchComplete=false always: this actor retrieves one page, at most 18 cards, without following pagination. OUTPUT.snapshotDelivered means the requested bounded snapshot was delivered, not that all available properties were scraped.

Prices are the source's rounded display strings with their qualifiers, not exact nightly quotes or totals at checkout. displayAmount parses only a dollar-prefixed numeric display string; exactCheckoutPrice is always null. Per-card date overrides are explicit. New/unrecognized ratings and missing prices remain null, never zero. Identity and pagination drift fail the whole page before its first billable delivery. No host profiles, personal contacts, precise coordinates, reviews, amenities, property detail pages or booking actions.

Pricing

Event	Base and all paid tiers
Public listing card	$0.99/1,000
First-page search report, including explicit empty source results	$0.0025/report
Actor start	$0.00005/start at 512 MB

One 18-card snapshot: $0.02037. One-card snapshot: $0.00354. Explicit empty-page report: $0.00255. These include start; the maximum charge is a ceiling, not the invoice. No charge for an undelivered report/card; source errors still incur the start event. A cap can deliver the report plus an affordable prefix of cards. Inspect OUTPUT and row count before using a snapshot. Repeated independent runs bill unchanged rows again; no watcher is offered. Platform compute, storage and transfers are included; no separate proxy or model fees. Set a ceiling of at least $0.00255; allow $0.02037 for all 18 cards.

Recovery and limits

The immutable first source observation is saved before delivery. Cancellation/resurrection uses that observation and the actual dataset ledger, preventing a changed live page from silently replacing the observation. Only delivered rows count as complete. Changed input, corrupt saved state, duplicates or unexpected dataset contents fail recovery; start a separate run for a new query. Independent concurrent runs have separate storage and bill separately. Manual edits to a run's storage are unsupported.

HTTP refusals/redirects are not bypassed or retried. One bounded retry is allowed for transient transport/server errors, with a five-second deadline per attempt and a 3 MB response bound. No login, model inference, cookies supplied by the customer or paid proxy service. Public-source availability can change.

Spreadsheet recipe

1. Save the exact input and run ID. Check OUTPUT.snapshotDelivered and source coverage before using rows; a successful capped run can still be partial.
2. Export JSON or CSV and keep both row types. The report carries scope, counts and pagination evidence; listing cards carry public URLs and capture times.
3. For a later independent run, keep the same destination, dates, adults and limit. Compare matching listing IDs and price qualifiers. Repeated cards are billed again; missing cards reflect bounded source coverage, not a delisting.
4. Resurrection resumes the saved observation, with another start event. It does not refresh prices. Start a new run for fresh data.

FAQ

Is this a full Airbnb market dataset? No. It is a small dated snapshot of one public results page, with source URLs and explicit coverage.

Can I compare prices across dates? Save separate inputs and snapshots. Match listing IDs and stay/guest scope; inspect price qualifiers and overrides. Rankings and rounding can vary; no checkout equivalence is promised.

Is it cheaper than every alternative? No. The advantage under test is the total base price for the same bounded first page versus tri_angle/airbnb-scraper. That actor offers broader coverage and lower per-listing prices at some tiers; curious_coder is also cheaper on paid tiers. On identical Stanley, Falkland Islands / December 1–2 / two-adult / USD input, both actors returned listing ID 1618374567122337756. Our configured one-card whole-run base charge is $0.00354; the leader charged $0.00410 and supplied broader property details. This is one scoped identity/price comparison, not a universal accuracy or coverage claim.

Changelog

0.1.2 — first-page parser, strict location/stay scope, dated report, cap/recovery ledger, explicit sparse/empty results and bounded retry. Initial private tests covered 1/10/18 cards, genuine 1-card/empty sources, report-only/final-row caps and cached resurrection; final release qualification is recorded separately.