Pick by workload. Classic result lists: Brave Search API ($5/1k, independent index). SERP JSON at volume: Serper (~$1/1k) or SerpAPI (verticals, SLAs). AI features — RAG, copilots, agents: return answers instead of links. SERPdive gives extracted, answer-ready content in one call at $5/1k credits — 60.7% of decided blind duels won vs Tavily's default search on a public benchmark, 20.2% fewer tokens — with 1,000 free searches a month to migrate on.
What happened, briefly
The Bing Web/News/Image Search v7 APIs were the web's workhorse search backend for over a decade — Google's official offering had been capped and half-deprecated for years, so “search API” in production usually meant Bing. Microsoft announced the retirement in May 2025 with ~three months' notice; existing keys stopped working August 11, 2025. The blessed successor, Grounding with Bing Search, lives inside Azure AI Foundry's agent framework: a tool an agent calls, priced per tool call, returning grounding material — not a search endpoint returning a result list you control.
So every migration starts with the same question: what did your code do with the results?
Route 1 — you showed users search results
You need another honest result feed. Brave Search API is the closest in spirit: its own independent index (not a Google/Bing proxy), $5 per 1,000 requests on the AI tier, web/news/image endpoints. The free plan carries $5 of monthly credit but requires a credit card. Result shape is familiar: titles, URLs, snippets.
Route 2 — you mined SERP data
Rank tracking, monitoring, dataset building: go to the Google-SERP-as-JSON providers. Serper at roughly $1/1k queries is the budget default; SerpAPI at $15–25/1k buys verticals (Shopping, Maps, Scholar), throughput guarantees and its legal-protection program. Details in our SerpAPI alternatives breakdown.
Route 3 — Bing fed an AI feature (most 2025-era usage)
Here, replacing Bing like-for-like re-creates the worst part of the old stack: the fetch-parse-truncate pipeline between the search call and your model. That pipeline exists only because the API returned links. SERPdive removes the reason: one POST /v1/search returns content already extracted from the live pages that answer, sized for a context window — 1–2 s with Mako, full-page depth with Moby, hosted MCP server included for Claude/Cursor stacks.
- SafeSearch-filtered consumer results pages → Brave (Route 1).
- Bing News API feeds → Brave news endpoint, or a news-specialist API if coverage breadth is the product.
- Autosuggest / spellcheck APIs → these have no true successor; most teams absorb them into their LLM layer.
- “Search + read the pages” for RAG/copilots → extraction-native search (Route 3).
The candidates, priced
| SERPdive | Brave | Serper | SerpAPI | Grounding w/ Bing | |
|---|---|---|---|---|---|
| Shape | Answer-ready content | Classic results, own index | Google SERP JSON | SERP JSON + verticals | Agent-framework tool |
| Per 1k | $5 credits (launch) | $5 requests | ~$1 | $15–25 | per-tool-call, Azure-tied |
| Free tier | 1,000/mo, no card | $5/mo credit, card required | 2,500 one-time | 250/mo | Azure trial |
| Drop-in REST search | Yes (POST) | Yes | Yes | Yes | No |
| Page content included | Yes | Snippets | No | No | Grounding material |
1,000 free searches a month, no card — port one Bing call path and compare what your model receives.
Get your free API keyFrequently asked questions
Why was the Bing Search API discontinued?
Microsoft announced in May 2025 that the Bing Search v7 APIs would retire on August 11, 2025, steering developers toward AI-mediated access — “Grounding with Bing Search” inside Azure AI Foundry Agents — rather than raw result feeds. The subtext most observers read: raw SERP access at Bing's old prices undercut the AI products built on the same index.
Is Grounding with Bing Search a drop-in replacement?
No. It is not a REST search endpoint you call with a query string; it is a grounding tool wired into the Azure AI Agent framework, priced per 1,000 tool calls, and it returns material for an agent's answer rather than a result list you control. Teams that just wanted GET ?q=… back have to migrate off Microsoft either way.
What is the closest like-for-like replacement for Bing Web Search?
For classic results from an independent index, Brave Search API ($5/1k requests, own index) is the closest in spirit. For Google-flavored SERP JSON, Serper (~$1/1k) or SerpAPI ($15–25/1k with verticals and SLAs). If the Bing API was feeding an AI feature, skip the like-for-like step entirely and return answers, not links.
What should RAG and copilot builders migrate to?
To a search layer that returns content, not references. Most Bing API code in AI products was half search call, half scraping-and-cleaning pipeline. Replacing both halves with one extraction-native call (SERPdive: $5/1k credits, 1–2 s, public quality benchmark) usually deletes code and lowers the token bill at the same time.
Is there a free Bing Search API alternative?
Free tiers exist across the board (July 2026): SERPdive 1,000 credits/month with no card, Tavily 1,000 credits/month, Serper 2,500 queries one-time, SerpAPI 250/month, Brave $5 monthly credit with a card on file. For hobby volumes you may never pay at all — see our free web search API breakdown.