Every used book World of Books has, by ISBN — and every copy's grade, price and warehouse

It is a first-party store. World of Books Group buys, grades and resells every book itself, so there is one seller, one grading scale and one price per copy. Nothing in a response is another trader's offer, there is no buy-box, and there is no personal data on the source — the only seller is a company.

Yes to both, and ISBN is a first-class key rather than a text search. Both ISBN-10 and ISBN-13 are indexed and both are returned on every book row. An ISBN-10 you pass in is converted arithmetically to the ISBN-13 the catalogue is keyed on, so 1408855658 and 9781408855652 resolve to the same record, and dashes and spaces are fine. Every check digit is verified before the call goes out, which means a typo comes back as a rejected input instead of as a book that does not exist. The isbn action takes up to twenty at once in a single request.

Two storefronts: the United Kingdom in GBP and the United States in USD. These are the only two the site publishes, and the country changes the catalogue, the edition, the price, the stock and the currency — not the wording. Measured on 2026-10-08 for one Harry Potter ISBN, the UK store published GBP 3.70 over 1,855 units and the US store USD 3.48 over 875, and a keyword search for the title returned the UK Philosopher's Stone on one storefront and the US Sorcerer's Stone, a different ISBN, on the other. The 66 countries World of Books ships to are all served from one of those two catalogues.

Far more than a listing suggests, because the unit is the copy. Measured across ten editions on 2026-10-08, each resolved to between six and ten separate copies with three to five available at once, each copy carrying its own grade, its own price and the warehouse holding it. The stock figure on an edition is a real unit count rather than an in-stock flag: 816, 875 and 1,855 were all returned in the same run, and filtering for a hundred copies or more left 667,365 UK editions.

Because a copy held in the other country is re-priced for that market, not converted. The same four UK-held used copies of one ISBN read GBP 3.70, 3.70, 3.70 and 3.78 on the UK storefront and a flat USD 10.00 on the US one — a ratio of 2.70 against the site's own published exchange rate of 1.32. So there is no arithmetic that gets you from one to the other, and guessing one from the other would be wrong. The API returns the storefront you asked for, names the warehouse country on every copy, and tells you whether that copy is buyable in that market. Where a copy has no price at all in the market you asked for, the price comes back null — never as 0.00, which is the literal value the source carries there.

Only the ones measured to change the result against an unfiltered call in the same run, on both storefronts and against three different control queries. Working: keyword, author, publisher, the three-level category tree, cosmetic condition, binding, language, product type, a publication-year window, a price range, a minimum stock depth and four sort orders. Measured on the UK storefront against 6,900,135 editions in one run: Hardback left 1,664,480, French 37,091, a 50 floor 65,298, a 3 ceiling 533, a hundred-copy floor 667,365, Stephen King 569. An in-stock filter is deliberately NOT offered: the storefront already drops unpriced records from every search, so the filter moved the total by exactly zero. And the site's own theme advertises thirteen sort orders while only four have a live index behind them, so the other nine are not offered rather than failing at call time.

The result window is a thousand records per query and the API says so with the exact maximum page for your page size, because the source answers an out-of-window page with a success status and an empty body that looks like the end of the results. To reach deeper stock, narrow with a filter or sort by price. The total the source reports alongside your rows is its own estimate, not a count: it answered 7,175,501, 7,173,373 and 7,191,099 for the same query minutes apart, so it is published as total_estimate with a total_is_estimate flag. A narrow query is exact and the flag says so. The rows themselves are always exact.

Search rows carry the author, publisher, publication year and date, binding, language, the three-level category path, the series where the source has one, and both ISBNs. Detail adds the page count, the description the store prints, the full breadcrumb, the publisher's list price and the complete copy table. One thing to know: the catalogue and the product page use different vocabularies for two of these facts — the page says Hardcover where the catalogue facet says Hardback, and the page often names the publishing group where the catalogue names the imprint, Simon & Schuster against Scribner. Neither is wrong, so detail returns both under separate names rather than silently picking one, and the catalogue value is the one a search filter round-trips on.

No, and the store says so itself — the image is the edition's cover, and the page carries the note that the actual binding, cover or edition may vary. There is no per-copy photograph and no per-copy defect note on the source, so none is invented here; the grade is the condition information that exists.

Almost always, and where it does not the API tells you. Across twenty-four editions checked against the product page's own published offers, the catalogue's cheapest price matched the cheapest available copy twenty-three times. In the exception the catalogue quoted USD 2.50 while the page's cheapest available copy was 14.38 — the search index lagging one reprice. Detail returns both numbers and sets a flag with a named warning when they disagree, so the gap is visible rather than quietly averaged away.

A little, and it matters for the identifiers. Measured on 2026-10-08 the catalogue held 8,222,289 books, 1,351 music releases, 785 DVDs and 168 games. On those non-book rows the catalogue puts an EAN or UPC in the ISBN-13 field and an Amazon ASIN in the ISBN-10 field, so the API publishes them as barcode_ean and asin and leaves the ISBN fields null, and it returns the director or artist where a book would have an author. Filter with product_type if you want books only.

World of Books API is a ReefAPI endpoint group for the uk's biggest second-hand bookseller as json: ~8 million used and new books searchable by isbn-10, isbn-13, author, publisher, category, binding, language, year, price and condition — then every physical copy on the shelf with its own grade, its own price and the warehouse holding it. It returns live JSON through POST requests under /worldofbooks/v1.

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

No login to World of Books 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.