It was switched off on 11 August 2025, along with the rest of the Bing Search family. This page is what Microsoft actually said, what they offer instead and why it is not the same shape, and a field-by-field mapping if you are moving to us.
From the Microsoft Lifecycle announcement, dated 15 May 2025:
"Bing Search APIs will be retired on August 11, 2025. Any existing instances of Bing Search APIs will be decommissioned completely, and the product will no longer be available to be used or new customer signup."
"Customers are encouraged to migrate to Grounding with Bing Search as part of Azure AI Agents. Grounding with Bing Search allows Azure AI Agents to incorporate real-time public web data when generating responses with an LLM."
It covered the whole family — Web, Image, Video, Entity, Custom and News — across the free F1 tier and the paid S1 to S9 tiers. Not a deprecation with a long tail: decommissioned.
The documentation is still readable and it is worth noticing how. Microsoft's own pages for the News Search API now carry is_retired: true, is_archived: true and ROBOTS: NOINDEX,NOFOLLOW, and live under /previous-versions/. That is the primary source confirming it, not a blog post about it.
Grounding with Bing Search is a real product and a good one — for its purpose. That purpose is letting an Azure AI Agent consult the web while an LLM writes an answer.
| Bing News Search API | Grounding with Bing Search | |
|---|---|---|
| You call | An HTTP endpoint | An agent, inside Azure AI Foundry |
| You get back | A JSON array of articles | Text an LLM wrote, with citations |
| You can | Page, sort, filter by category | Ask a question |
| Fits | A pipeline, an index, a database | A chat experience |
If what you had was a call that returned news as structured JSON — to fill a table, build an index, feed a retrieval pipeline — then being handed an agent that writes prose is not a migration. It is a different product answering a different question, and rewriting your pipeline around it is a larger job than replacing the endpoint.
The old NewsArticle object, next to what we return. Taken from Microsoft's archived v7 reference, so the left column is exact.
Bing NewsArticle | hawkcrawl | Notes |
|---|---|---|
name | title | Same thing |
url | resolved_url | Needs resolve=true; costs 4 credits |
description | snippet | Both short; neither is the article |
provider[].name | source | A string, not an array |
datePublished | published_at | Not the same date — see below |
clusteredArticles[] | duplicates[] | With dedupe=true, and the copies are not billed |
| — | content | Bing never returned article text. We do, on 77% |
| — | content_status | Says why, when we could not |
category | — | We do not have it |
image, video | — | We do not have them |
mentions[] | — | No entity extraction |
totalEstimatedMatches, offset | — | No paging. About 100 results is the ceiling |
sort=date|relevance | — | No sort parameter; there is relevance scoring |
contractualRules | — | Not applicable |
Microsoft documented datePublished as "the date and time that Bing discovered the article". Ours is the date the publisher put on it, from the feed. For fresh news they are close; for anything indexed late, or for archive queries, they are not, and any date arithmetic you carried over will quietly shift.
Read that table honestly: four things you had are gone — categories, images, entity mentions and paging. If your product leaned on any of them, we are not a drop-in and you should know that before you write the adapter, not after.
What you gain is the article body, which Bing never returned, and an archive that reaches 1995.
Before:
GET https://api.bing.microsoft.com/v7.0/news/search?q=tesla&count=20
Ocp-Apim-Subscription-Key: YOUR_KEYAfter:
GET https://api.hawkcrawl.com/v1/news?q=tesla&limit=20&resolve=true&content=true
Authorization: Bearer YOUR_KEYOne endpoint, a bearer token, JSON out. The API reference has every parameter and the full response shape.
You are not choosing between Microsoft and us, and pretending otherwise would be silly.
| Option | Worth knowing |
|---|---|
| Google News RSS, direct | Free, and for one subject in one language it is the right answer. We wrote the reference for it, including the three walls that stop it being enough. |
| NewsAPI.org | $449/month to start, and content is truncated to 200 characters by its own documentation. |
| GNews | €49.99/month, full content, archive back to 2020. |
| NewsData.io | $199.99/month; ten years of archive costs $1,299.99. |
| hawkcrawl | $10 once, no renewal. Archive to 1995, article text on 77%, and nothing charged for what we cannot deliver. |
Measured on 18 September 2026 over 636 articles — thirteen subjects, sampled this week and again in 2010, 2016 and 2021.
Full text runs from 40% to 100% depending on the subject. Three publishers never open: reuters.com answers 401 because it wants a subscription, nytimes.com answers 403, and finance.yahoo.com refuses the connection. The full breakdown is here.
Start free Read the API reference See pricing
Retirement dates and the wording quoted above come from Microsoft's own Lifecycle announcement and archived v7 reference, read on 18 September 2026. Our own figures come from a measured run the same day. If either stops matching what you find, tell us and we will correct the page.
hawkcrawl · News API · Google News API · Google News RSS · Crypto news API · Docs