The best web search API for an AI agent returns fresh, structured results the agent can cite, and Desearch does this with live web and X search through one API, and its web search API returns ranked titles, URLs and snippets as JSON. An AI agent that answers questions about the present needs a web search API: send a query over HTTP, get structured results back, and ground the model in sources it can cite. The market has split into APIs that return ranked links, APIs that return page content or excerpts sized for a context window, and APIs that return a generated answer with citations. Two long-standing options have also left the field: one big-engine search API was retired, and one big-engine site search API is closed to new customers.
This web search API comparison groups the options into four categories and shows what a result contains and which job a web search API for AI agents fits.
Disclosure: Desearch publishes this guide and builds the live social API in it.
How to compare search APIs on predictable pricing for agent workloads
Agent workloads repeat calls in loops, so compare three things before you pick a web search API:
- Price per 1,000 calls. Use the provider's own pricing page and convert every rate to the same unit.
- Rate limits. Check which windows apply (per minute, per day, per month), whether the numbers are published, and how you raise them.
- Billing model. Pay as you go or a plan, and whether each response reports its own cost so you can meter a single agent run.
Desearch values (checked October 24, 2026):
| What to compare | Desearch value | Source |
|---|---|---|
| Web Search price | $1.00 per 1,000 searches (Console shows $0.10 per 100 searches, the same rate) | Pricing |
| AI Search price (web and X) | $0.40 per 1,000 items | Pricing |
| Extract price | $0.50 per 1,000 pages | Pricing |
| Rate limits | Enforced per API key on three windows: per minute, per day and per month. The first limit reached applies. Numeric limits are not published on the public docs; your key's limits are shown in Console. Exceeding a limit returns 429. | Rate limits and errors |
| Billing model | Pay as you go, billed per usage unit per API. New accounts get $5 in API credits after attaching a card in Console, then pay as they go. | Pricing |
| Cost per request | Billable responses report cost_usd and usage_count, and the same values arrive as X-Desearch-Cost-Usd and X-Desearch-Usage-Count headers | Pricing and billing |
Rates are reference rates on the pricing page; confirm current rates in Console.
Get your API key and log cost_usd on your first request.
Quick comparison
Each row is a category. Cells describe what that kind of API returns and which agent job it fits. Starting credits and allowances differ by provider and are not repeated here.
| Category | What a search returns | Fit for agents |
|---|---|---|
| General web search | Independent web index with titles, URLs, descriptions, and extra snippets. Also a web index with optional full-page content and a big-engine site search API that is closed to new customers. A big-engine API in this category was retired. | Ranked links and snippets for agents that cite sources |
| AI answer and grounding | Agent-focused results with a content snippet, score, and optional generated answer. Also semantic search returning text, highlights, or summaries; markdown excerpts for LLM context; answer-engine search with ranked results and dates; content results with an optional sourced answer or structured output. | Excerpts or a generated answer inside the search response |
| SERP scraping | Parsed search-engine results, including SERP features. Also search plus optional full-page markdown per result. | Result layout and SERP features, or page markdown with each hit |
| Social and live search | AI web search API returns web results. AI Search adds cited sources across the web and X with an optional summary. Rates are on the pricing page. | Web results and live public posts in one integration |
How we compared
- What a search returns. Links and snippets, extracted text, a parsed result page, or a generated answer. That decides how much extraction, reranking, or summarization an integration still has to build.
- Fit for agents. Ranked citation, model grounding, SERP features, or live social posts.
- Limits. Whether a self-serve rate limit is published. Read the current number on the provider's site. This guide does not copy those numbers into a category.
- Price. Search API pricing comes from each provider's own pricing page for a standard search.
We did not run a relevance benchmark for this guide. For quality, test the same set of queries from your own application against two or three shortlisted APIs.
How to compare web search APIs for developers (2026)
The best web search API for developers in 2026 is the one that matches your result format, freshness needs, and pricing model; compare them on the seven criteria below. If you need ranked results as structured JSON with per-request cost reporting, the Desearch Web Search API returns titles, URLs, and snippets from one GET /web endpoint.
Before you pick a web search API, decide what your code will do with the results. These criteria cover most developer builds.
| Criterion | What to check | Why it matters |
|---|---|---|
| Freshness | How quickly new pages appear in results, and whether you can filter by date | Agents answering questions about the present need recent sources |
| Result format | Ranked links and snippets, extracted page text, or a generated answer | Decides how much parsing and prompting you do yourself |
| Page extraction | Whether full page content is returned or you need a separate crawl step | Affects context quality and the number of calls per task |
| Pricing model | Per search, per credit, or per token, and whether cost is reported per request | Lets you forecast spend and reconcile bills |
| Rate limits | Requests per second or minute on each plan, and how limits are raised | Sets your ceiling for concurrent agents and batch jobs |
| Paging and controls | Offsets, result counts, and query parameters | Needed for deeper result sets and repeatable queries |
| Docs and SDKs | API reference quality, code samples, client libraries, MCP support | Shortens time to first working request |
Where Desearch fits. The Desearch Web Search API returns ranked titles, URLs, snippets, and usage metadata as structured JSON from a single GET /web endpoint. You can page with the start offset, and each billable response reports cost_usd and usage_count (also sent as response headers), so spend is easy to track per request. Parameters and fields are listed in the generated API reference. If you want cited answers instead of ranked links, use the Desearch AI search API, which covers the web and X. Get your API key and make your first request.
General web search
These APIs return ranked web results rather than a generated answer.
- Independent web index. Titles, URLs, descriptions, and extra snippets. See compare with an independent-index search API.
- Web index with optional full-page content. Ranked results, with the page body available on request.
- Big-engine site search API, closed to new customers. Existing customers have until January 1, 2027 to move to an alternative. For an integration that still calls it, see the site search API alternative.
A big-engine API in this category was retired on August 11, 2025. It no longer returns results to your own code. Customers were pointed to a grounding feature that supplies web data to an agent rather than returning results for application code to parse. The retired search API alternative page maps that style of call to a replacement.
Best for: agents that cite ranked links and snippets, including teams leaving a retired or closed big-engine API.
AI answer and grounding
These APIs return material a model can ground on: a snippet with a score, an excerpt, or an answer with sources.
- Agent-focused results with a content snippet, a score, and an optional generated answer. See compare with an agent-focused search API.
- Neural or semantic search returning text, highlights, or summaries. See compare with a semantic search API. Desearch AI Search returns web and X results with titles, links, snippets and highlights in one call, at $0.40 per 1,000 items.
- Markdown excerpts for LLM context.
- Answer-engine search with ranked results and dates. See compare with an answer-engine search API. Desearch AI Search searches the web and X in one call and can add a written answer that cites its sources, at $0.40 per 1,000 items.
- Content results with an optional sourced answer or structured output.
Best for: workflows that want excerpts or a cited answer from the search call.
SERP scraping
These APIs return the result set a search engine shows, or the page body with each result.
- Parsed search-engine results, including SERP features. See compare with a SERP scraping API.
- Search plus optional full-page markdown per result. See compare with a search-and-crawl API.
Best for: applications that need result layout and SERP features, or the full content of each page in the same workflow.
Social and live search
Desearch is the API in this category. It offers a plain Web Search endpoint and an AI Search endpoint behind one API key, alongside dedicated X API endpoints for public posts, profiles, replies, and trends.
- Web Search:
GET /webreturns ranked titles, URLs, and snippets as JSON. See pricing. - AI Search:
POST /desearch/ai/searchanswers a prompt across the web and X, with results you can cite and an optional final summary. Reddit threads use GET /web with site:reddit.com in the query. See pricing. - X data: public posts, profiles, replies, and trends are on the X API. See pricing.
- Billing visibility: billable responses report
cost_usd,usage_count, and the meteredservice, and the same values arrive asX-Desearch-*headers.
Best for: agents that need web results and public X data from one integration, and teams that meter usage per request.
How to choose
- Web results and public X data together: Desearch covers both from one key. Rates are on the pricing page.
- An index independent of the major search engines: compare with an independent-index search API.
- Full page content with each result: compare with a search-and-crawl API. A general web index can also return full-page content, and an agent-focused API can return raw content. See compare with an agent-focused search API.
- Search-engine result layout and SERP features: compare with a SERP scraping API.
- A cited answer or structured output from the search call: content results with an optional sourced answer or structured output, or Desearch AI Search. For text, highlights, or summaries, compare with a semantic search API. For answer-engine results, compare with an answer-engine search API.
- Replacing a retired or closed big-engine API: start with the retired search API alternative or the site search API alternative, then run existing queries against two candidates.
Whatever you shortlist, measure a complete task rather than a single call: the searches, any extracted pages, and any generated answer your application needs.
Frequently asked questions
Does a web search API come with starting credits?
Starting offers differ by provider and change over time, so check each provider's current terms. New Desearch accounts get $5 in API credits after attaching a card in Console, then pay as they go. Current rates are on the pricing page.
What is the cheapest web search API?
Billing units, result counts, and the amount of page content differ, so compare the cost of a full task against each provider's current rates. Desearch lists its rates on the pricing page.
Are big-engine search APIs still available?
One big-engine site search API is closed to new customers. Existing customers have until January 1, 2027 to move to an alternative. A separate big-engine search API was retired on August 11, 2025 and no longer returns results to application code. See the site search API alternative and the retired search API alternative.
What replaced the retired big-engine search API?
The retired big-engine search API no longer returns results to calling code. Customers were pointed to a grounding feature that supplies web data to an agent, rather than a response an application parses itself. Teams that need structured results typically move to another web search API. The retired search API alternative page maps that style of call to a replacement.
Sources
Desearch product pages linked here were part of the September 16, 2026 check.
- Desearch: Web Search API, AI search API, Search API guide, X API, Pricing and billing, Pricing
- Migration maps on this site: retired search API alternative, site search API alternative