One row per car, every Dutch site that lists it attached

One row is one car, not one advert. Gaspedaal has already matched the same car across the Dutch sites that advertise it, and this API hands you that group: sources holds every site, its name, its id, its kind and a link that resolves to the listing on that site. Measured on 2026-10-01 over 900 rows, a car carried 1 to 12 sources, mean 3.99, and a single 100-car page touched 20 different Dutch sites. If you scraped the sites separately you would have to match the cars yourself; here the matching is the data.

The source_sites action answers that from a live sample instead of a brochure. On 2026-10-01 a 100-car sample contained 20 sites: the dealer's own website (on 100 % of cars), AutoTrack (100 %), AutoScout24 (71 %), ANWB (65 %), FinancialLease.nl (40 %), Lease.auto (30 %), DTC lease (28 %), Financial Lease For You (20 %), NationaleAutolease (19 %), Twelvelease (18 %), Regeljelease.nl (17 %), Mobilist (14 %), Vakgarage (9 %), Bedrijfsauto.nl, ActivLease, SEAT Approved Occasions, Carteam, AutoFirst, JAMES Autoservice and Mercedes-Benz Certified. A 300-car sample found 23. Worth knowing before you build on it: a good share of those are financial-lease portals rather than independent classifieds, which is why every source carries a kind field so you can keep them apart.

Inside one call, no: rows are de-duplicated by ad_id and duplicates_dropped tells you how many repeats were removed — it was 0 across 900 rows in each of two runs. Across consecutive pages of one search it does happen, because the live result set re-ranks between page fetches: three consecutive pages of a Volkswagen search returned 300 rows containing 274 distinct cars, so 26 rows (8.67 %) were a car that had already been served on an earlier page. De-duplicate by ad_id when you page. What is not happening is double-listing: in those same 300 rows, 1,581 distinct source adverts appeared and not one of them belonged to two different Gaspedaal cars, and 274 cars produced 274 distinct brand-model-trim-year-mileage-price fingerprints.

All 35 were measured against the unfiltered total in the same run, twice, and all 35 moved it. From 348,966 cars: electric 34,734, automatic 208,365, manual 139,378, convertibles 12,032, red 17,229, two-door 19,165, verified NAP mileage 142,554, VAT-deductible 155,471, margin cars 123,045, private sellers 2,146, BOVAG warranty 76,540, 300 kW and up 33,822, seven seats or more 10,438, eight cylinders or more 8,154, 500 km of electric range or more 8,950, within 10 km of Amsterdam centre 5,725, listed in the last day 11,941, Volkswagen 36,471, Volkswagen Golf 6,789. Filters the site accepts and then ignores are not offered at all, and an invalid enum value is rejected with an error instead of silently returning the whole catalogue.

Yes, and this is specific to an aggregator. source_site returns only cars listed on that site — id 2, AutoTrack, covered 213,005 of 348,966 cars. only_source_site returns cars exclusive to one site — 7,700 cars are on AutoTrack and nowhere else, and each of those rows came back with exactly one source, which is how it was verified. exclude_source_site removes a site from every car's group and drops cars that then have no source left: excluding AutoTrack left 341,253 cars and not one of the returned rows still carried it.

Each car exists twice in the same response: once in Gaspedaal's result data and once in the structured data the page publishes for search engines. The engine reads both on every row, returns price_eur and price_eur_schema, and sets price_conflict when they differ. On a 100-car page checked field by field, price, mileage and door count agreed 100 out of 100, and across 1,800 rows in two acceptance runs price_conflict was never true. The same twin is used to fill in a value Gaspedaal leaves blank in one place but prints in the other — which is how gearbox coverage went from 66 % to 99.7 %.

Counted over 900 rows per run across nine different searches, not assumed. 100 %: ad_id, title, brand, model, price, build year, mileage, listing date, photo and source count. 99.8 % trim, 99.8 % fuel, 99.7 % gearbox, 99.0 % doors, 97.7 % colour, 95.7 % power in kW and hp. Lower, and worth planning around: body style 87.6 % — Gaspedaal leaves it blank on about one ad in eight; cylinder capacity 85.8 %, mostly because electric cars have none. For the seller, town is 99.8 % and province 84.3 %, but the dealer's trading name is published on only about 31 % of ads, so where it is missing the API still returns that dealer's own Gaspedaal page instead (59 % of rows).

348,966 cars on 2026-10-01, 100 per page, and no artificial ceiling: page 3,490 returned 71 rows and page 3,491 returned none, which matches the stated total. Page size is fixed by the site at 100 and cannot be changed. Nine sort orders: relevance, distance (needs a postcode), price up or down, mileage up or down, year up or down, and newest first.

Four things that would be fair to expect and are genuinely absent from this surface: the advert text, the equipment list, the number of previous owners and the APK (Dutch MOT) date. Gaspedaal also has no detail page of its own — the full advert lives on the source sites, which is why the resolved source link matters. There is no phone number, no e-mail address and no private seller's name anywhere in the data, so none is returned. A Dutch licence-plate lookup appears to exist on the site but could not be verified against a known car, so it is not offered rather than offered untested.

Median 899 ms in one acceptance run and 1,136 ms in the other, over 12 live calls each; the slowest non-resolving call measured 2,967 ms. Resolving the source links is a separate request per link, so it is off by default in search, on by default for a single car, and capped by max_resolved_urls (10 by default, 40 maximum). A search that resolved 12 links took 6.0 s in one run and 9.4 s in the other — anything the call budget cannot finish stays null and is counted in source_urls_unresolved rather than dropped.

Gaspedaal.nl API is a ReefAPI endpoint group for the dutch car aggregator as json: one car per row, every nl site that lists it attached. It returns live JSON through POST requests under /gaspedaal/v1.

Yes. ReefAPI starts with 1,000 free credits, no card required. Gaspedaal.nl calls use the same shared credit balance as every other ReefAPI engine.

No login to Gaspedaal.nl is needed for the API response. You call ReefAPI with your x-api-key header, and the playground can run live examples before you create a production key.

The page example is captured from a live search call, and production requests fetch live data through ReefAPI rather than a static sample.