include_answer · POST /v1/answer
Cited AI answers
Ask a question, get a short answer whose every claim points at a numbered source from the live Google results. Add it to any search, or call the answer endpoint to read the top pages first and stream the text as it's written.
How it works
Answers written from live Google results, every claim cited.
Two ways to get one. include_answer: true on /v1/search or /v1/news adds an answer (text and citations) written over that response's results, for +2 credits (+4 when include_content read the pages). POST /v1/answer is answer-first: depth: "fast" answers from the results page for 3 credits, and depth: "standard" also reads the top 3 pages (up to 5) for 8.
Claude Haiku 5.5 writes the answer from the numbered sources only: the answer box, the knowledge panel, then the results or the pages they link to. Markers pointing past the source list are dropped, and instructions inside a page are ignored. A decision model (OpenAI's Decisions API) skips sources that don't help and sends easy questions to a faster setting.
Stream with stream: true (Server-Sent Events: results, delta…, done), or pass output_schema to get the answer as JSON. With format: "markdown", a search response leads with an ## Answer section and its sources. When the model fails, the model part is refunded with a meta.warnings entry or a 503 ai_unavailable.
curl https://api.serpkite.com/v1/answer \
-H "Authorization: Bearer $SERPKITE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"q":"what is retrieval-augmented generation","depth":"standard","num_sources":3}'{
"answer": "Retrieval-augmented generation (RAG) grounds a language model's answer in documents retrieved at query time [1], which reduces hallucinations and keeps answers current without retraining [2].",
"citations": [
{
"index": 1,
"title": "Retrieval-augmented generation",
"link": "https://en.wikipedia.org/wiki/Retrieval-augmented_generation",
"domain": "en.wikipedia.org"
},
{
"index": 2,
"title": "What is retrieval-augmented generation (RAG)?",
"link": "https://www.example.com/blog/what-is-rag",
"domain": "example.com"
}
],
"meta": {
"credits_used": 8,
"credits": {
"search": 4,
"answer": 4
},
"llm": {
"provider": "anthropic",
"model": "claude-haiku-5-5"
}
}
}include_answer or /v1/answer
| `include_answer: true` | `POST /v1/answer` | |
|---|---|---|
| Endpoints | /v1/search, /v1/news | Its own endpoint |
| Answers from | That response's results | The results page, or the top 1–5 pages |
| Response | The full search response plus answer | answer, citations, results |
| Streaming | No | stream: true (SSE) |
| JSON output | No | output_schema |
| Price | +2 credits (+4 with include_content) | 3 (fast) or 8 (standard, 3 pages) |
Why it matters
What you get
Every claim cited
[n] markers resolve to citations with title, link and domain; invalid markers are removed.
Streams as it writes
Show the sources at once, then the answer text token by token over SSE.
Free when it fails
A failed or empty answer refunds the model part; you pay only for the search that ran.
No prompt logs
Question and source text go to the model under API terms that exclude training, and SerpKite never logs them.
FAQ
Cited AI answers: common questions
Keep exploring
Related pages
Start building
Get your API key in 30 seconds
2,500 free credits, then 1,000 every month. No credit card.