The xd Crossword Corpus

Contributing puzzles

For editors, constructors and publishers. Draft, November 2026.

If you edit, construct or publish crosswords, we would like to keep yours. This page says what we do with puzzles we are given, who can see them, and what you can ask of us. The short version: your originals are kept safe and unchanged, the puzzles are never offered to solvers, and you stay in control of how your work is represented.

Why we are asking

Crosswords disappear. Newspapers fold, websites go dark, file formats stop opening, and a run of puzzles that was solved by thousands of people survives only in a box or on one person’s hard drive. The xd Crossword Corpus exists so that published puzzles stay readable and findable for the long term, a century or more, in a plain-text form that needs no particular software. The corpus already holds more than 160,000 puzzles from over 160 publications, back to 1913, and it is used to study how crosswords and their language have changed.

What happens to what you send

• The originals are kept exactly as received. Whatever you send, in whatever format, is stored unchanged in a checksummed archive. We never edit originals.
• Each puzzle is converted to one plain-text file holding its grid, clues, answers, title, constructor, editor, date and copyright line, as published. Obvious misprints are corrected and the correction noted; nothing you did on purpose is changed. The conventions are in EDITORIAL.md.
• Every puzzle carries its provenance: where it came from, when, and from whom, so your contribution is always identifiable as yours.
• The corpus is private. It lives in a repository that is not public, with more than one copy in more than one place. How it is kept is described in PRESERVATION.md.

Who can see it

The puzzles are not sold and not offered to solvers. Researchers can ask for access for a stated purpose and must agree to terms that forbid passing puzzles on or using them for anything that delivers puzzles to solvers. Most research needs only derived data, such as clue and answer pairs or statistics, and that is what most people get. The forms of access and the terms are in ACCESS.md.

Aggregate statistics and analyses are published at xd.saul.pw. They describe the corpus as a whole and do not reproduce puzzles.

The Cross(word) Reference. Puzzles in the corpus appear on The Cross(word) Reference, a site for looking up clues, answers and constructors across crossword history. A puzzle is shown there as a reference, with its answers filled in, not as something to solve. If you would rather your puzzles, or your whole series, did not appear there, say so when you contribute, or at any time, and they will be left out.

What you can ask of us

At any time:

• what we hold from your publication;
• corrections to anything we got wrong;
• that a constructor be named as they wish to be named;
• that specific puzzles, or all of yours, be removed. Removals are honored and recorded, and anyone who has a copy is asked to delete it.

What to send

Anything. Puzzle files in any format (.puz, .ipuz, .jpz, Crossword Compiler, Across Lite, spreadsheets), PDFs, page scans, photographs of clippings, or a link to an online archive. Complete runs are ideal, but partial ones are welcome; gaps can be filled later. If you are not sure what you have, send a sample and we will tell you what we can do with it.

Send it, or ask us anything, at xd@centuryarcade.org. Large archives can be shared however is convenient for you.

What you get back

A clean, checked copy of your own archive in a durable plain-text format, with the errors we found listed, which for most publications is a better copy than the one they started with. If it is useful to you, we can also provide statistics about your run: constructors, word counts, grid sizes, how your puzzles compare with the rest of the corpus.

Doing the conversion yourself

Some contributors prefer to convert their puzzles and submit them directly. That is welcome. The converters are in xdformat, and the import guide, with the rules for naming, filing and provenance, lives in the private corpus repository; ask for access (see ACCESS.md) and we will set you up.