← All walkthroughs

It checks dozens of sources and hands you a report

Let AI do the research for you

What it does: This is the chatbot-to-agent leap. Instead of you opening twenty tabs, Deep Research works on its own for several minutes: it searches, reads, cross-checks, then hands you a structured report with sources. You ask for the answer; it does the legwork. How long: 5–15 minutes (it works while you do something else). Does it cost money: Free tiers have limited runs per month; paid tiers more.

Effort Setup Risk
★★☆☆☆ Describe what you want ★★☆☆☆ Find the Deep Research mode ★★☆☆☆ Verify before big decisions

How to do it

Using ChatGPT Deep Research (or the equivalent in Google Gemini)

  1. Open a new chat and look for the Deep Research option (in ChatGPT it’s in the tools menu next to the message box). Regular chat won’t do multi-step browsing on its own; this mode is what makes it an agent
  2. Copy and send this prompt, swapping in your topic:

Research [topic] for me: break it into 3-5 key questions, verify each one online, then give me a report with the conclusion first, then what different sources say and where they disagree, and note how current the information is.

  1. While it works (you’ll see it searching and reading), go do something else. Come back to a structured report
  2. Ask the follow-up that matters: “How reliable are the sources for this conclusion?” It will tell you which claims rest on solid sources and which are thin

Good uses vs. bad uses

Good: comparing products across dozens of reviews, understanding a new field before a big purchase, background for a decision (“moving to Austin with two kids, what do we need to know about schools?”), sorting out what actually happened in a confusing news story

Bad: anything where being wrong is expensive. A report that reads confidently can still rest on weak sources. The follow-up question above is not optional for medical, legal, or financial topics

Watch out


Last tested: pending. This walkthrough was prepared from the products’ documented capabilities and has not yet been verified step-by-step by a real person. If you try it (whether it works or not), tell us on GitHub.

Reminder: content goes to the AI's servers. IDs, bank cards, passwords — never send those.
Didn't work?

Four causes cover most failures:


Ratings: Prompt effort = how much you have to say to get the result you want; Setup = how much you need to install; Risk = how bad it gets if things go wrong.Not yet verified on a real device