Microsoft retired the Bing Search APIs on August 11, 2025, and its official replacement is Grounding with Bing Search in Azure AI Agents. That tool gives an agent a written answer. It gives your code no raw results. For an app that needs web results through an API, the direct replacement is a third-party search API: a SERP API that returns a full results page as JSON, Brave's independent index, or an AI search API such as Tavily, Exa or Linkup.
- Official path: Grounding with Bing Search costs $14 per 1,000 transactions. You get the agent's answer and its citations, and the terms forbid you to store the output.
- Raw results: Brave has its own index at $5 per 1,000 searches. A Google SERP API returns the whole results page, with the AI Overview and People Also Ask. SerpApi also has a Bing engine.
- Accuracy: in an open benchmark of six search APIs on 824 multi-hop questions, the two Google SERP APIs answered 80.2% and 77.4% from one search. Exa answered 64.0%, Brave 53.9% and Tavily 48.4%.
What did Microsoft retire, and when?
Microsoft published the retirement notice in May 2025. It is three sentences long:
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.
The notice names no exception, so old keys do not return results and you cannot create new ones. The paid v7 tiers cost $15 to $25 per 1,000 transactions from May 2023, per Microsoft's archived pricing page, and one call returned up to 50 results.
What does Microsoft point to instead?
The notice tells customers to "migrate to Grounding with Bing Search as part of Azure AI Agents". The agent decides when to search, Bing returns chunks of pages to the model, and the model writes the answer. Microsoft's Foundry docs now recommend a newer Web Search tool for new agents, which works the same way with a Bing resource that Microsoft manages.
Both tools fit a chat agent that runs on Azure. They do not fit an app that needs the results themselves:
- No raw results. Microsoft's Grounding docs say: "Developers and end users don't have access to raw content returned from Grounding with Bing Search." You get the model's response, the cited URLs and a link to the Bing query.
- No storage. The Grounding terms forbid you to "copy, store, cache, archive, or create a database of Output". You may use the output only in "internet search experiences", you must show the references in the form Microsoft gives, and you may not use the output to train a model.
- Azure setup. You need a paid Azure subscription, an Azure OpenAI model deployment, and the Contributor or Owner role. The search query leaves the Azure compliance boundary.
- Cost. $14 per 1,000 transactions, up to 150 a second and 1 million a day. A transaction is one tool call, and you also pay for the model tokens.
So a rank tracker, your own result list, a dataset, or a RAG pipeline that caches pages needs a different API.
Which APIs replace the Bing Search API for raw web results?
Prices are each vendor's live list price for one real-time search in the default mode, read on November 14, 2026. Pay as you go where the vendor sells it; otherwise the plan closest to $100 a month.
| Replacement | Index | Returns | Key needs | $ per 1,000 searches |
|---|---|---|---|---|
| Grounding with Bing Search | Bing | The agent's answer with citations, no raw results | Paid Azure subscription, Foundry agent, model deployment | $14, plus model tokens |
| Brave Search API | Brave's own index | Results as JSON: title, URL, description, extra snippets | API key, a card on file; storage rights only on a custom plan | $5, with $5 of free credit a month |
| SerpApi | Google, plus Bing with engine=bing and other engines |
The full results page as JSON | API key | $15 on Developer ($75 for 5,000); 250 free a month |
| cloro | The full Google results page as JSON: organic positions, AI Overview, People Also Ask, related searches | API key | $1.20 on Hobby; $2.00 with the AI Overview | |
| Tavily | Its own retrieval | Results with extracted page content for an LLM | API key | $8 pay as you go ($0.008 a credit, 1 credit a basic search); 1,000 free credits a month |
| Exa | Its own neural index | Up to 10 results; page text or highlights at $1 per 1,000 pages each | API key | $7 (auto); $10 of free credit a month |
| Linkup | Its own | Results, a sourced answer, or structured output | API key | $5 for results, $6 for a sourced answer (standard depth); up to $20 of free credit a month |
The Google SERP row is Hobby: $100 a month for 250,000 credits, so $0.40 per 1,000 credits. A Google search is 3 credits ($1.20 per 1,000), and a search with the AI Overview is 5 credits ($2.00). A synchronous request adds 2 credits, so $2.00 and $2.80.
Is a SERP API a good replacement for Bing's organic results?
Yes, when your app needed web results and did not need Bing's own ranking. cloro's Google SERP API is a good replacement for that case: one call returns the full results page as JSON, with the organic positions, the AI Overview and its sources, People Also Ask and related searches. The results are Google's. cloro does not return Bing's organic results. On Bing it returns only the AI answer (Bing AI Mode, which Microsoft calls Copilot Search in Bing).
Most v7 code reads three fields per result, and each one has a direct match in a Google SERP response:
| Bing v7 field | Google SERP field |
|---|---|
webPages.value[].name |
result.organicResults[].title |
webPages.value[].url |
result.organicResults[].link |
webPages.value[].snippet |
result.organicResults[].snippet |
Order in webPages.value
|
result.organicResults[].position |
relatedSearches.value[].text |
result.relatedSearches[].query |
| No equivalent |
result.aioverview.text and result.aioverview.sources[]
|
| No equivalent | result.peopleAlsoAsk[].question |
Three cases where a Google SERP API is the wrong pick:
-
You need Bing's ranking. For Bing rank tracking or Bing market data, use a SERP API with a Bing engine, such as SerpApi's
engine=bing. Google and Bing are different result sets: in a five-query pilot they shared a mean of 4 of their 10 first-page domains. - You need an answer in under 3 seconds. The median search took 7.96 s and 8.20 s on the two Google SERP APIs in the benchmark below, against 1.21 s on Brave and 0.74 s on Exa.
- You wanted Bing because it was not Google. Brave is the closest match: an independent index at $5 per 1,000 searches. AI engines name it most often for this question today.
Which Bing Search API replacement is most accurate?
Six web search APIs ran in an open benchmark in November 2026. Each API got the question as its query, one search and 10 results. One model answered from those results alone, and a second model graded each answer.
| API | Short factual (SimpleQA Verified, 1,000) | Recent events (FreshQA, 358) | Multi-hop (FRAMES, 824) |
|---|---|---|---|
| Parallel | 97.8% | 82.7% | 65.3% |
| Exa | 97.3% | 83.0% | 64.0% |
| cloro | 96.9% | 88.3% | 80.2% |
| SerpApi | 94.7% | 88.5% | 77.4% |
| Brave Search API | 93.1% | 84.9% | 53.9% |
| Tavily | 91.8% | 81.8% | 48.4% |
On short factual questions, the top three are within the margin of error. On recent events, the two Google SERP APIs tie. On multi-hop questions they lead, and the four independent indexes answered "not found" on 151 to 295 of 824 questions. Linkup and Grounding with Bing Search were not tested. On the benchmark's cost basis (the plan that covers 100,000 searches a month), the top Google SERP API cost $1.92 per 1,000, Brave and Tavily $5, Exa $7 and SerpApi $7.25.
Which one should you pick?
| Your app needs | Use |
|---|---|
| An answer inside an Azure agent, with nothing stored | Grounding with Bing Search, or the Foundry Web Search tool |
| Bing's own ranking | A SERP API with a Bing engine, such as SerpApi |
| Fast results from an index that is not Google | Brave |
| The full results page with the AI Overview, or the best multi-hop accuracy from one search | A Google SERP API |
| Page text ready for a model, at low latency | Exa, Tavily or Linkup |
How this was checked
This comparison is published by cloro, which is one of the APIs in it. Microsoft's notice, docs, terms and Grounding price, and every other vendor's price, were read on their own pages on November 14, 2026. The v7 tier prices come from Microsoft's archived pricing page. The Google SERP field names come from one live call on the same day. The accuracy figures come from the benchmark linked above, which publishes its questions, code and graded answers. The five-query Google and Bing overlap pilot, the Bing-native SERP APIs, and code to migrate a v7 call are in the full Bing Search API migration guide.
Top comments (0)