How does the NYT Wordle API work?
Last verified 08-07-2026
The New York Times serves each day’s Wordle from a public JSON endpoint at https://www.nytimes.com/svc/wordle/v2/YYYY-MM-DD.json. It returns the complete history back to 19 June 2021 and, at the time of writing, roughly a month of future puzzles.
{"id":1011,"solution":"posit","print_date":"2026-08-05",
"days_since_launch":1873,"editor":"Tracy Bennett"}days_since_launch is the puzzle number — id is not
This is the detail that trips people up. id looks like a puzzle number and is not one: it is an internal identifier for the word itself, and it jumps around. On 5 August 2026 the id was 1011; the next day it was 3281.
The number players see is days_since_launch, with no offset. Checked against independent contemporaneous reporting at three points spanning four years:
| Date | days_since_launch | Solution | Reported as |
|---|---|---|---|
| 241 | AROMA | #241 AROMA | |
| 926 | MURAL | #926 MURAL | |
| 1873 | POSIT | #1873 POSIT |
One edge case: the 19 June 2021 launch response omits days_since_launch entirely.
It publishes ahead
Future dates resolve. The horizon has been observed at 33 days and later at 38 or more, so it moves — NYT appears to publish in batches rather than maintaining a fixed lead. Anything past the horizon returns {"status":"ERROR","errors":["Not Found"]} rather than an HTTP error, so treat a 200 with no solution as “not published yet”.
The practical consequence for anyone building on this: fetch the whole visible window rather than just today, and you inherit weeks of resilience for free. This site does that daily, which is why it keeps serving if the endpoint goes away.
Answers have changed before
On 15 February 2022, during the acquisition, the legacy site served AGORA while the New York Times served AROMA for the same puzzle. Anything relying on this data should compare sources and alarm on disagreement rather than silently overwriting.
Etiquette
It is an undocumented endpoint on someone else’s infrastructure. Cache aggressively, fetch once per day rather than per request, identify yourself in the user agent, and do not hammer it. This site makes one pass per cron run and caches every result.