How to Use Deep Research in ChatGPT, Gemini and Perplexity

Deep research is an AI mode that searches dozens or hundreds of web pages on its own, reads them, and writes a long, cited report, usually in 5 to 30 minutes. To use it, open ChatGPT, Gemini or Perplexity, choose the Deep Research option in the tools menu or mode switcher, describe your question in detail, approve or edit the research plan, then review the finished report and its sources.

This guide shows how to use deep research in ChatGPT, Gemini and Perplexity step by step, what each one is best at, plan limits, prompts that get better reports, and how to check the results before you rely on them.

What is deep research?

A normal AI chat answers from the model’s knowledge plus a quick web search. Deep research is an agent: it makes a research plan, runs many searches, opens and reads pages, follows leads, compares sources and then writes a structured report with citations. It trades speed for depth. A regular answer takes seconds; a deep research report takes minutes but covers far more ground.

Because it acts on its own across many steps, deep research is a good everyday example of agentic AI. It is best for questions where the answer is spread across many places, not for quick facts.

Deep research in ChatGPT, Gemini and Perplexity compared

OpenAI does not publish exact deep research counts per plan, and Google and Perplexity change limits often, so check the current numbers in each app. Microsoft removed Deep Research from the consumer Copilot app in August 2026, so for Copilot users the options above are the main choices. If you are choosing a research tool overall, see our best AI research tools and Perplexity vs ChatGPT.

How to use deep research in ChatGPT

1. Open a new chat. Go to chatgpt.com or the app and sign in.
2. Turn on deep research. Click the “+” or tools button in the prompt box and choose deep research.
3. Write a detailed brief. Include your goal, audience, region, time frame and output format (see the prompt templates below).
4. Answer clarifying questions. ChatGPT often asks a few questions first. Answer them carefully; they shape the whole report.
5. Let it run. You can leave the tab; you will see progress and the sources it is reading.
6. Review the report. Check the citations for the most important claims, then ask follow-up questions or request a shorter summary.

On paid plans you can connect plugins and company knowledge so deep research can include sources beyond the open web. Our ChatGPT beginner guide covers the rest of the interface.

How to use deep research in Gemini

1. Open Gemini. Go to gemini.google.com or the Gemini app.
2. Choose Deep Research. Select it from the tools in the prompt box.
3. Enter your question. Gemini proposes a research plan with the steps it will take.
4. Edit the plan. Remove steps you do not need or add angles you care about, then start the research.
5. Read and export. When it finishes, open the report and export it to Google Docs if you want to edit or share it.

Note that Gemini’s new Skills feature does not yet work with Deep Research, so keep your research instructions in the prompt itself; see Gemini Skills explained. For more Gemini tips, read how to use Gemini.

How to use deep research in Perplexity

1. Open Perplexity. Go to perplexity.ai or the app.
2. Switch to Research mode. Choose the research option in the search box mode selector.
3. Ask a focused question. Perplexity works best with a clear question and a defined scope.
4. Review sources. Every claim links to sources; check the ones that matter most.
5. Export or continue. Save the answer to a Space, export it, or ask follow-ups in the same thread.

Perplexity Pro, at $20 a month, includes deep research along with premium data sources such as PitchBook and Statista. Our Perplexity guide and Perplexity pricing breakdown cover the details.

Deep research prompts that get better reports

The single biggest factor in report quality is your brief. Compare a weak prompt with a strong one.

Templates you can copy:

1. Competitor scan: “Compare [competitor A, B, C] on pricing, features, target customers, recent launches and reviews. Use official sites and reputable reviews from the last 12 months. Output a comparison table and a one-page summary of gaps we could exploit.”
2. Buying decision: “I need [product type] for [use case] under [budget]. Find the top 5 options available in India, with current prices, pros and cons, and long-term reliability. Recommend one.”
3. Topic primer: “Explain [topic] for a smart beginner. Cover history, how it works, current state in 2026, key players, open debates and five sources to read next.”
4. Content research: “Find the most common questions people ask about [topic], the best existing answers, and gaps no one covers well. Output a content outline.”

Best uses for deep research (and when to skip it)

Deep research uses a lot of compute, which is why plans limit it. Save your runs for questions where breadth really matters, and use normal chat or web search for everything else. Students can use it to map a topic before reading primary sources, freelancers to prepare client proposals, and small business owners to compare suppliers or software before buying. Marketers often pair it with SEO work; our guide to ChatGPT for SEO shows how to turn research into briefs.

How to check a deep research report

Deep research reports look authoritative, which makes errors easy to miss. Before you use one for a decision or publish anything from it:

• Open the citations behind the three to five claims your decision depends on.
• Check dates. Prices, rules and statistics change; make sure sources are recent.
• Watch for weak sources such as anonymous blogs, press releases presented as facts or outdated pages.
• Look for missing perspectives. Ask “What evidence contradicts this?” as a follow-up.
• Do not copy reports directly into published content; rewrite in your own words and verify facts.

Deep research for your own documents

Web deep research is not ideal when the knowledge lives in your own files. For research across PDFs, notes and transcripts, Google NotebookLM is a strong free option: its free tier includes 10 Deep Research runs a month and up to 50 sources per notebook. ChatGPT and Claude projects also let you upload files and ask questions across them. See how to use NotebookLM and AI tools to chat with PDFs.

Frequently asked questions

Start your first deep research report

Pick one question you would normally spend an afternoon researching, write a detailed brief using the templates above, and run it in the tool you already use. Then check the key sources and compare the result with a second tool if the decision matters. Plan limits and features change often, so confirm details on OpenAI’s ChatGPT pricing page and Perplexity’s pricing page.