What you can build
Answers from AI engines
ChatGPT, Gemini, Perplexity, Google AI Overviews and AI Mode, Bing Copilot, Naver, Meta AI, Qwen and Doubao: answer text and citations in one shape.
Search and web data
Google, Bing and Naver result pages, Reddit, single-page scrapes, site maps and crawls, and keyword volume.
Prompt research
The questions people ask AI assistants about a topic, ranked by search demand.
GEO monitoring
Track how often AI answers mention and cite your brand against competitors, on a schedule.
How it works
Every task runs asynchronously. A query to an AI engine takes seconds to minutes, so the API accepts the task, runs it in the background, and hands you the result when it is done.1
Submit a task
POST /v1/async/task with a taskType (the engine) and a payload (the prompt and options). The response returns a task id with status QUEUED.2
The task runs
A worker queries the engine and extracts the answer text, the cited sources, and any engine-specific fields.
3
Collect the result
If you set
webhook.url, the result is POSTed to your endpoint. Otherwise poll GET /v1/async/task/{id} until task.status is COMPLETED or FAILED.What a result contains
Every answer engine returns the same core inresponse:
Engines add their own fields on top. Google engines return a full results-page envelope, ChatGPT and Perplexity can return the search queries they ran, and Naver AI Tab returns its shopping panel. Each engine page documents its shape.
Base URL
Authentication
Send your API key as a bearer token:Authorization: Bearer $QUERYING_API_KEY. Every endpoint needs a key except the public ones: /health, /ready, /v1/capabilities (the engine roster and prices), and answer evidence files under /v1/evidence/. See Authentication.
Product pages
Next steps
Quickstart
Submit your first task and collect the result.
Authentication
Create a key and learn which endpoints need it.
MCP server
Let an AI agent call querying.ai as tools.
Credits and limits
What each task costs and how much you can run at once.