Test
10,000 searches
$7.50
$0.75/1K searches
- Try every engine on your real queries
- Same API as the biggest pack
- Credits never expire
Migration guide · facts checked on 22 October 2026
Microsoft switched off the Bing Search APIs on 11 August 2025 and pointed customers to an LLM grounding tool. If your code needs a list of results, not a generated answer, this page shows what to use and how to port it.
Quick answer
What replaced the Bing Search API?
Microsoft retired the Bing Search APIs on 11 August 2025 and recommends Grounding with Bing Search in Azure AI Agents, which feeds web data into an LLM answer at $14 per 1,000 transactions (October 2026). For a raw list of results, use a SERP API: SerpCrypto returns Bing or Google results pages as JSON from $0.30/1K searches, with no KYC and payment in crypto.
| Announcement | 15 May 2025, on Microsoft Learn's lifecycle page |
|---|---|
| Retirement | 11 August 2025: existing instances decommissioned, no new sign-ups |
| Microsoft's replacement | Grounding with Bing Search, a tool of Azure AI Agents |
| What it returns | Web data used to ground an LLM response, not a raw results list |
| Its price | $14 per 1,000 transactions (Microsoft's grounding pricing page, read 2026-10-22) |
| Old Bing Search API tiers | S1 $25, S2 $15, S3 $6 per 1,000 transactions, as reported by PPC Land |
| SerpCrypto engine | engine=bing: organic_results and related_searches as JSON, 1 credit per search |
Grounding with Bing Search is built to answer a question through an agent, not to hand you the results list. Microsoft describes it as a way for Azure AI Agents to "incorporate real-time public web data when generating responses with an LLM". It runs inside Azure's agent service, and you get the model's answer with its sources.
That suits a chatbot. It does not suit code that ranked, filtered or stored the old webPages.value array: rank trackers, research tools, deduplication jobs, link checkers or a RAG pipeline that chooses its own documents. Those need the results page itself, which is what a SERP API returns.
| Option | Returns | Price per 1,000 | Account requirements |
|---|---|---|---|
SerpCrypto (engine=bing or engine=google) | The results page as JSON | $0.75/1K searches down to $0.30/1K searches | No KYC, crypto top-up from $25 |
| Grounding with Bing Search (Azure AI Agents) | Web data for an LLM answer | $14 per 1,000 transactions | Azure subscription |
| Bing Search APIs v7 | The results page as JSON | $6 to $25 (S3 to S1) | Retired 11 August 2025 |
Prices checked on
Prices read on 2026-10-22 from the sources below. The Bing v7 tiers are those reported by PPC Land in June 2025.
Set engine=bing and most parameters translate one to one. The main change is paging: Bing v7 let you ask for up to 50 results with count, while a results page holds about 10, so you page with page instead of offset.
| Bing Web Search v7 | SerpCrypto | Note |
|---|---|---|
Ocp-Apim-Subscription-Key header | Authorization: Bearer … | Or api_key in the query string |
q | q | Operators such as site: work as typed |
mkt (en-US) | gl=us and hl=en | Country and language as two parameters |
setLang | hl | |
count (max 50) | — | About 10 results per page |
offset (0, 10, 20…) | page (1, 2, 3…) | page = offset / 10 + 1 |
safeSearch (Off, Moderate, Strict) | safe (off, active) | No middle setting |
freshness (Day, Week, Month) | period (d, w, m) | Also h and y |
responseFilter | — | Every block on the page is returned |
In the response, webPages.value[] becomes organic_results[]. The v7 fields name, url and snippet become title, link and snippet, and each result gains a position.
import os
import requests
# Before (retired 11 August 2025):
# requests.get("https://api.bing.microsoft.com/v7.0/search",
# headers={"Ocp-Apim-Subscription-Key": KEY},
# params={"q": "heat pump subsidy", "mkt": "en-GB", "count": 10, "offset": 10})
resp = requests.get(
"https://serpcrypto.com/api/v1/search",
headers={"Authorization": "Bearer " + os.environ["SERPCRYPTO_API_KEY"]},
params={"engine": "bing", "q": "heat pump subsidy", "gl": "gb", "hl": "en", "page": 2},
timeout=60,
)
resp.raise_for_status()
for r in resp.json().get("organic_results", []):
print(r["position"], r["title"], r["link"])
Switch if you chose Bing only because it had an official API. Change engine=bing to engine=google and keep the rest of the request: the same key, the same credit price and the same organic_results shape. Google also returns blocks Bing v7 never had in one response, such as People Also Ask and the AI Overview.
Keep Bing if your product depends on it specifically: some teams compare the two engines, and Bing's index feeds Microsoft Copilot answers. Many teams run both and store the two results lists side by side, at one credit each.
The old S1 tier cost $25 per 1,000 transactions and S3 $6, according to PPC Land. With SerpCrypto the same 1,000 searches cost $0.75/1K searches on the 10,000 searches pack and $0.38/1K searches on 1,000,000 searches, as of October 2026.
The billing model also changes: no Azure subscription and no monthly invoice. You buy prepaid credits from a crypto balance, each successful search uses one, and credits never expire. A request that fails is not charged.
Count one credit per results page. A job that used count=50 now needs 5 pages per query, so multiply those queries by 5. Credits work on Bing and on every Google engine.
10,000 searches
$7.50
$0.75/1K searches
100,000 searches
$56.25
$0.56/1K searches
500,000 searches
$281.25
$0.56/1K searches
1,000,000 searches
$375
$0.38/1K searches
In every plan: Every engine at 1 credit per search · Failed requests not charged · Credits never expire · Country, language, city and device targeting
On 11 August 2025. Microsoft announced it on 15 May 2025 and said existing instances would be decommissioned completely, with no new sign-ups. The retirement covered the Bing Search APIs product family, including Web Search.
No. It is a tool for Azure AI Agents that grounds an LLM's response in web data, priced at $14 per 1,000 transactions in October 2026. It does not give you the raw webPages list the v7 API returned.
Yes, through a SERP API. With SerpCrypto, send engine=bing and a query to https://serpcrypto.com/api/v1/search; the response holds organic_results and related_searches, for one credit per successful search.
No. SerpCrypto sells prepaid credits paid in BTC, USDT, USDC, ETH, SOL, LTC, with a $25 minimum top-up. It asks for no ID, no selfie, no company documents and no card.
Not as one request. A results page holds about 10 results, so 50 results take 5 requests with page=1 to page=5. Each page is one credit.
From $0.30/1K searches · no KYC
Get an API key