August 5, 2026

/google/rank reports where a domain stands for a query, scanned across the whole measured result set — typically 100–250 results. The response carries the best position, every appearance of the domain with its exact page and title, and results_scanned, so a null position always names the set it was measured against.

Subdomain handling is explicit: the default mode counts the domain and everything under it, mode=exact counts only the domain (www always counts as the domain).

Rank checks share their snapshot with /web-search — a check right after a search, or repeated checks within the snapshot window, are served from cache and cost no extra request time.

August 2, 2026

Telegram joins the catalog. /channel returns a public channel by username with a stable schema: title, description, subscriber count, verification badge and avatar.

/messages returns its posts, newest first, up to 100 per response. Each message carries its permalink, publish time, view count, plain text and a formatting-preserving HTML rendering limited to a safe subset of tags — plus whatever it attached: photos and albums, videos, stickers, files with their type, size and duration, polls with per-option shares, and the origin of a forwarded post.

Pagination walks backwards through a channel with meta.pagination.next_cursor and ends with null at the first message. A username with no public channel returns 404 NOT_FOUND on both endpoints, never an empty page.

August 2, 2026

An API is not "up" because one of its endpoints answers. Each endpoint in the catalog is now checked on its own, once a minute, and reported as its own line on the status page with its own 90-day uptime history. APIs roll up to degraded when any one of their endpoints is failing.

Each check validates the endpoint's published response contract, including that fields documented as always-present are actually present — a response that parses but has quietly lost data counts as an outage, not a success.

The /status.json feed gains group and route fields per component, and component ids now name the endpoint (for example google.web-search).

August 2, 2026

The versioning policy now states the guarantee explicitly at field level: a field is retired on the same 90-day notice as an endpoint, and never phased out by quietly returning null in place of a value it used to carry.

It also documents what null means, which is the other half of the same promise: null is always a statement about the data — the value does not exist for that resource — and never means we could not fetch it. When we cannot assemble a complete response, the request fails with a retryable error instead of returning one with fields missing.

August 2, 2026

/google/dictionary returns the full entry for an English word: syllabification, phonetic pronunciations with audio, etymology, and one entry per part of speech with its inflections, numbered senses, synonyms and usage examples.

Definitions are stable, so responses are served from a 24-hour cache and meta.as_of tells you when the entry was read. Words with no entry return a clean 404 NOT_FOUND rather than an empty result.

August 1, 2026

Instagram joins the catalog. /profile returns a public profile by username with a stable, consistent schema: identity and bio, follower/following/post counts, verification and account type, category and the highest-resolution avatar available.

/posts returns the account's most recent posts — image, video/reel and carousel — each with its permalink, caption, like and comment counts, media and thumbnail URLs, dimensions and publish time. A posts call right after a profile call is served from cache.

Responses are cached briefly and every one carries meta.as_of, so you always know how fresh the data is. Unknown usernames return a clean 404 NOT_FOUND.

July 31, 2026

Google joins the catalog with three focused endpoints. /web-search returns ranked results with absolute positions and publication dates where available; /knowledge-panel returns structured entity data for queries that resolve to one; /autocomplete returns query suggestions with entity details and a spelling correction when the input looks misspelled.

Search pagination is page-numbered and fixed-size: every page returns exactly per_page results and continues until the result set ends, signalled by meta.pagination.next_page: null. Ranking positions stay stable across pages, and meta.as_of tells you exactly when the snapshot was taken.