AI Deep Research

Enter a topic. The AI plans focused sub-questions, searches the live web, and writes a structured report with cited sources.

Compare the leading open-source LLMs for coding in 2026 What are the health effects of intermittent fasting? History and current state of solid-state batteries

Free. Each report costs ~800 tokens from your daily pool.

Sources

How Deep Research Works

Enter a topic

A full question works best — 'compare the top open-source vector databases for RAG in 2026' beats 'vector databases.' Broad topics are fine; the agent narrows them for you.

It plans sub-questions

The AI breaks your topic into 3-5 focused sub-questions so the research covers each angle, not just the obvious one. You'll see them appear in the research trail.

It searches + reads sources

Each sub-question runs a live web search. The AI reads the results, de-duplicates the sources, and keeps the most relevant ones — you can watch each search happen.

Get a cited report

You get a structured report with headings and key takeaways. Every claim is cited with a numbered [1][2] link — click any citation or the sources list to verify.

Use Deep Research for

Comparisons, overviews of a new topic, market or competitor scans, due diligence, and gathering cited background before you write — anything you want to actually understand, with sources.

Just need a quick answer?

For a single fast answer with citations, use AI Search — it does one search instead of a full multi-step report.

Try AI Search

Foire aux questions

Free AI Deep Research is a multi-step research agent. You give it a topic; the AI first plans several focused sub-questions, searches the live web for each, then reads all the sources and writes a structured, cited report — instead of a single quick answer.

AI Search does one search and gives you a short synthesized answer. Deep Research breaks your topic into 3-5 sub-questions, runs a separate web search for each, de-duplicates the sources, and writes a longer report with headings, comparisons, and a key-takeaways summary — better for topics you want to actually understand, not just look up.

Yes. Each research report costs about 800 tokens from your daily pool. You get 6,000 tokens/day free without an account, or 30,000/day with a free account.

A language model plans the sub-questions, our self-hosted SearXNG search engine fetches live results for each, and the model synthesizes everything into a report. You can watch each step happen in the research trail as it runs.

Yes. Every claim is cited inline with a numbered link like [1][2], and the full list of sources is shown below the report so you can verify anything or read further.

Yes. It searches the live web at query time, so it is not limited by an AI model’s knowledge cutoff.

No. You can run a full research report immediately without an account.

Usually 20-60 seconds — it runs multiple searches and writes a long answer, so it takes longer than a single quick search. You can watch the progress in the research trail.

Yes. Enter your topic in any language and the report will be written in that language, with sources from the web.

Your queries are not stored or logged. Searches run through our own self-hosted SearXNG instance, not a third-party tracker.

Comparisons ("best open-source vector databases for RAG"), literature-style overviews, market or competitor scans, due diligence, learning a new topic, or gathering cited background before writing something.

Yes. Deep research is available as a single HTTPS call to /v1/research/ with a Bearer key — see /api/ for details.

Love this tool? Share it!

Noter cette page