The AI answer API with verified citations Answer API

An answer API takes a natural-language question, searches the live web, and returns a written answer with citations instead of a list of links. It bundles retrieval, passage selection, synthesis and citation checking into one request. Developers use one when a chatbot, copilot or internal tool needs a grounded, sourced reply without building and tuning that pipeline themselves.

Three things. answer is a Markdown answer with numbered inline citations such as [[1, 2]]. citations is the ordered list of sources, each with a source URL and the verbatim excerpts the model used to build the answer. results.web is every web result considered during synthesis, cited or not, with URL, title, snippets and, where available, description, thumbnail and page_age.

The Answer API reports 93.48% accuracy on SimpleQA, a 4,326-question benchmark of short fact-seeking questions, with a p50 latency of 2.67 seconds. Accuracy comes from the stages around the model: live retrieval on You.com's own index, passage selection, and a verification step that confirms each citation exists in the source text and supports the claim before the answer is returned.

Before a response returns, the API checks that every citation actually appears in the source text and supports the statement it is attached to. The excerpts field in each citation contains the exact passages used, so your application or your users can confirm a claim without trusting the model alone. For legal, financial or medical use, you can add a step that follows each source URL and checks the excerpts.

Use the Answer API when you want a cited answer from one search pass with flat cost and latency, and nothing to configure. Use the Research API when a question needs multiple searches and reasoning steps, when you want to pick an effort level from lite to frontier, or when you need source_control or structured JSON through output_schema. Many products use both: Answer for interactive questions, Research for reports.

Yes. include_domains restricts answers to an allowlist, exclude_domains removes sources you don't want, and boost_domains ranks preferred sites higher without excluding others. Each accepts up to 500 domains. Combine them with freshness, country, language and safesearch to shape the evidence behind every answer.

Answers are built from real-time web results on You.com's continuously recrawled index. Use the freshness parameter to require recent sources: day, week, month, year, or an exact date range such as 2026-01-01to2026-03-31. Results include page_age so you can show users how current each source is.

Most grounded answer APIs return text with links. The You.com Answer API also returns the verbatim excerpts behind each citation and verifies them before responding, runs on an independent index rather than resold results, and exposes the same domain, freshness and locale controls as the Web Search API. Pricing is a flat $5 per 1,000 calls with no token-based charges or effort tiers.

Yes. It is built for question answering in AI applications: chatbots, copilots, support assistants and research tools that need grounded, cited responses. Because the output is Markdown with citations and excerpts, you can render it directly or pass it to your own model as context. It also works as a research step in content pipelines and internal knowledge tools.

A query can be up to 400 characters and cannot be blank. For complex topics, include the angle or constraints you care about, such as the workload size or the time period. If your question is longer or needs a long brief, use the Research API, which accepts inputs up to 40,000 characters.

The Answer API is REST-only. The You.com MCP server exposes search, contents, research and finance tools, but there is no you-answer tool, so agents that need a one-pass cited answer call POST https://api.you.com/v1/answer directly with the X-API-Key header. Any HTTP client or the youdotcom Python SDK works.

New accounts get $100 in free credits with no credit card required, which covers roughly 20,000 answers. That is enough to evaluate accuracy on your own questions, run a pilot, and ship a proof of concept before paying anything.

The Answer API costs $5.00 per 1,000 calls. One request is one billable call, and citations and the full web result list are included. There are no effort tiers and no token charges, so cost and latency stay flat across queries. Volume discounts and annual pricing are available for high-throughput applications.

We publish our numbers: 93.48% accuracy and 2.67 seconds p50 latency on SimpleQA. Because every call includes citations with verbatim excerpts, you can also run your own faithfulness checks on your own questions. Your free credits cover evaluating the Answer API on your workload instead of relying on vendor claims.

No. Queries, answers and results are never used for model training. Zero Data Retention (ZDR) can be added to an enterprise agreement and restricts retention of Answer API request and response content account-wide with no changes to your integration. The platform is SOC 2 certified and DPA-ready.