Google News API. One JSON shape for the whole edition.

Eleven routes: search, top stories, topic streams, topic resolution, Full Coverage stories, publisher articles, navigation sections, the home briefing, Headlines by section, topic and publisher suggestions, and article-link resolution. Every route returns the { data, meta } envelope.

Send a supported language and country pair, for example en-US and US, en-GB and GB, de and DE, or ja and JP. You can send the language family with a country, so en with US selects the en-US edition. An unsupported pair is rejected with a 400 that lists every supported pair. The resolved edition is returned in data.edition.

The first search call returns the ranked Google News results, about 100 articles. When it pages, the first page.nextCursor covers today's Pacific day, then each later cursor one older day, with an after: plus before: bound, up to 100 items per window. With an after and before range, the first cursor covers the last day of that range, then earlier days down to after. A busier window is cut short; page.windowFull is true when that happened, so you can tell a complete window from a truncated one. A first call with after, before or when already has a date bound, and it says so too. The walk stops after seven empty days or at your own after/when floor. Keep the effective query; a cursor cannot be combined with when, after, or before.

Some entity topics have no stories for a period. The answer is still lookupStatus: found with articles: [], because the topic exists. Do not retry it as a miss; a genuinely unknown topic returns not_found with topic: null.

Read source.publicationId and pass it to GET /v1/google-news/publishers/{publicationId} with the outlet's own language and country. Read storyId and pass it to GET /v1/google-news/stories/{storyId}. Both fields are null when the article record does not carry them.

Call GET /v1/google-news/topics/resolve with exactly one of place, section, mid or query. The answer carries resolved plus candidates from the suggestion box. resolved.matchedInput is false when a place name fell back to the U.S. topic, for example an unknown place resolving to U.S.; an unknown place can also return not_found. A localised name for a real place, such as München for Munich, is still a match.

Call GET /v1/google-news/articles/resolve?ids=… with 1 to 10 article ids, comma-separated. Each entry in results is { id, url } in request order. url is null when it is not available, which is a normal answer rather than a failure.

url is the publisher article URL on topic, story, publisher and Headlines results. It is null on search and top-stories results, where googleUrl is the Google News link; call articles/resolve to turn those ids into publisher URLs. image is the publisher's image with its width and height. Authors, labels and favicons come through when the record has them.

Each route charges one credit per completed request, and each search page is a separate request. Check meta.creditsCharged on every response. New accounts start with 100 free credits.

No. HTTP 200 means the request finished. Check data.lookupStatus: not_found means the id does not exist or the story has expired, and the object is null. For a publisher, also check streamStatus: an outlet requested in the wrong language and country is found with an empty list and streamStatus: unavailable.