🧠 Deep Research with AI: 9 Ways to Get Started

Practical strategies for thorough, citation-rich AI research

The AI search landscape is transforming at breakneck speed. New "Deep Research" tools from ChatGPT, Gemini and Perplexity autonomously search and gather information from dozens — even hundreds — of sites, then analyze and synthesize it to produce comprehensive reports. While a human might take days or weeks to produce these 30-page citation-backed reports, AI Deep Research reports are ready in minutes.

What’s in this post

• Examples of each report type I generated for my research, so you can form your own impressions.

• Tips on why & how to use Deep Research and how to craft effective queries.

• Comparison of key features and strengths/limitations of the top platforms.

What’s new about Deep Research? Traditional AI queries deliver isolated answers to specific questions, while Deep Research tools conduct sophisticated investigations with dozens of interconnected searches. It’s like the difference between a quick reference check and a thorough research expedition.

Al Deep Research tools use this process to deliver detailed reports after you submit a query. The steps are similar to those a human might take.

Why this matters for you: It’s now easy to generate your own reports to immediately advance work projects. ChatGPT’s Deep Research, the best I’ve tried so far, is now available on free accounts, with a limit of five reports per month.

How to get started: After reading the guide below, customize a query for one of the AI models reviewed here. Start with one of the report types suggested below. First toggle on the “Deep Research” setting:

Turn on ChatGPT’s Deep Research. Gemini & Perplexity have similar toggles.

Pick a work topic or any subject you’re curious about. Read through the resulting report and iterate on your query to get an even more useful second response. You’ll soon have a new research superpower to deploy anytime you need to dig deep into a complex subject.

Security & privacy: When using AI platforms avoid sharing sensitive data, which can leak. Adjust settings so your data isn’t retained to train future models. Take the same precautions for AI Deep Research you’d adopt for other AI queries.

Perplexity makes it easy to turn off data retention. You can also use its incognito mode for further privacy. More: Gemini’s privacy hub and OpenAI’s.

Hallucinations? I expected substantial nonsense to pollute the reports. But extensive in-line citations help with verification and I’ve found fewer errors than I expected. Still, for subjects where data is sparse, autonomous research reports occasionally include weak sourcing, so keep an eye out for that.

9 Practical Ways to Use AI for Deep Research

AI research tools shine when you need comprehensive information on complex topics. Here are specific use cases where they excel:

1. Craft Custom Itineraries ✈️

Create detailed, personalized travel plans by specifying your destination, dates, activity preferences, budget, cultural interests, and whatever else is important to you.

These AI-generated itineraries often surface unexpected gems. When planning a family trip recently, my wife and I discovered a fantastic farm stay in Pennsylvania through a Perplexity query. We wouldn't have found it otherwise. Use the results as a starting point to identify interesting possibilities, then follow up with targeted research.

• Specify dietary preferences, accessibility needs, and your taste in accommodations, restaurants, and entertainment, for more tailored recommendations. I’ve saved a block of text about this to reuse.

• Use follow-up queries to get more specifics on attractions or activities that appeal to you, or to compare and contrast potential itineraries.

Example: Deep Research itinerary for a family vacation in Brookline, MA. Compare results from ChatGPT, Gemini, Perplexity, and Copilot. Note: If you spot errors in reports I’m sharing in this post, feel free to annotate the GDocs.

2. Compile Organization Reports 📈

Get comprehensive backgrounders on companies, non-profits, or any organization in minutes rather than clicking through dozens of search results.

• Compare similar organizations or competitors

• Specify format. Maybe you’d like a case study format, a topical report, a chronological history, or an industry context analysis.

• Specify metrics of interest like funding history, revenue growth patterns, leadership changes, media coverage trends, law suits, or anything else you care about. For closely-held private organizations data may be scarce, so read results skeptically.

• Advanced tip: Copy excerpts from Deep Research reports into Claude to transform them into visual dashboards— including charts and interactive elements— using Claude Artifacts. You can share those with colleagues. Watch: 📺 Grace Leung’s helpful video illustrates how and why to try this.

Example: Deep Research report on Trader Joe’s

3. Research Notable People 🧓

Explore backgrounds of news figures, historical personalities, or even fictional characters. Request specific information like podcast appearances, YouTube videos, or published works to build a well-rounded understanding of the individual.

• Ask about connections between the person and influential contemporaries

• Ask for lesser-known background details or contributions often overlooked

• Specify time periods to focus on particular life phases or epochs

🔎 Example: Deep Research on Michel de Montaigne

4. Explore Complex Concepts 🧠

Learn about complex topics in any field — from botany 🌿 to venture capital 💰— with AI-structured explanations tailored to your knowledge level.

• Ask for real-world examples, analogies, anecdotes, quotes, common misconceptions, and step-by-step explanations.

• Ask for quiz or discussion questions to test your understanding.

• After reading the report, generate an AI tutor with a Custom GPT, Gemini Gem, or Poe bot to further strengthen your understanding.

🔎 Example: Applications for AI in medical diagnosis via Gemini & Perplexity

5. Discover Places in Depth 🌍

Investigate a place’s historical significance, cultural development, architecture, art, music, literature, or economic, social, or political history. I find this richer, personalized context can feel more resonant than a more generic travel guide.

• Ask about little-known local events, hidden gems, or notable personalities

• Specify your interest in fashion, architecture, history, sports, or whatever else

• Products too: Get a backgrounder on a new type of oven 🧑‍🍳 you’re considering, or pianos 🎹 that might suit your home.

🔎 Example: Help me learn about Coolidge Corner in Brookline, Massachusetts

6. Analyze Debates and Controversies 🤔

Explore complex controversies from multiple perspectives. Examine international conflicts, ethical debates, or local issues. Deep Research reports can present multiple viewpoints with examples and evidence to deepen your understanding of nuances. You can also ask for notable quotes and an annotated reading list.

• Ask how the debate has evolved over time and who has been involved

• Specify that you want evidence-based arguments from multiple disciplines

• Ask for areas of common ground between opposing viewpoints

🔎 Example: Is generative art revolutionizing creativity — or devaluing it?

7. Decode Cultural Works 🎶 🎨

Gain insights into books, paintings, music, or other creative works by exploring critical analyses, historical context, and expert interpretations. This works particularly well when you're in the middle of reading a book or have just encountered an intriguing piece of art.

At the Metropolitan 🖼️ Museum of Art last week, I saw Tang Dynasty zodiac figurines and asked AI to help me learn about them. (More info & a generated image). Caveat: as a non-expert learning about something new, I’m not always in a strong position to assess the merits of AI analysis. When in doubt, it’s helpful to have cited sources for gauging the quality of the information.

• Request info about a work's influence on subsequent artists or movements

• Ask for analysis of technical innovations or stylistic elements

• Compare interpretations of the book, play, poem, art, or piece of music from different time periods or cultural contexts

🔎 Example: Help me deepen my appreciation of Mozart’s 5th Violin Concerto

8. Explore Evolving Trends 👗

Investigate linguistic, political, fashion, sports, arts, business, or cultural trends in a particular place or time period, or compare trends across cultures.

• Ask for predictive insights about how the trend might evolve

• Include in your prompt questions about counter-trends or critics

• Request data visualizations if the trend has quantitative aspects, e.g. the Beanie Baby craze

🔎 Example: What are top AI training programs for journalists around the world?

9. Examine Historical Context 📝

Explore historical events through multiple lenses - political, economic, social, and cultural. Direct your AI research assistant to focus on specific date ranges, if relevant. Ask it draw from diverse sources across countries and perspectives.

• Request primary source recommendations

• Ask for comparisons across countries, regions or time periods

• Specify your prior knowledge so the report is tailored for your context

🔎 Example: Help me learn more about the history of the Dozier School for Boys

When Deep Research isn't your best option ⚠️

Use other tools when you’re not looking for a comprehensive research report, but instead want something quick, or for:

• Simple factual questions like award winners or sports results are better addressed with basic Google or Perplexity searches

• Breaking news where online info is limited

• Multimedia searches may work better with specialized search engines, like Listen Notes for finding someone’s podcast appearances

• Paywalls If the open Web lacks relevant info, don’t expect miracles

Further Caveats

• Keep an eye out for errors. Verify info in these reports. The presence of citations doesn’t guarantee accuracy. For example, some sources may publish estimates that get treated by an AI search agent as definitive data.

• Research is only as good as its sources. Some subjects lack extensive source material. AI research reports may, in such cases, rely heavily on publishers with flimsy fact-checking or an axe to grind.

• The bottom line: Check the source list before diving into a report. When you know of high quality sources, reference them in your prompt. This Claude thread helped me include high-quality sources for my Trader Joe’s inquiry.

How to strengthen Deep Research queries 💡

The quality of your prompt significantly impacts your results:

Be detailed about your topic, reasons for interest, and how you'll use the information

Unlike standard Google searches where you only provide keywords, deep research queries benefit from detailed direction. Guide your AI research assistant on specific areas of focus, recommended sourcing, prior context, and formatting: how best to present its findings.

Specify your preferred tone & format —tables, lists, pros/cons, bullets

Request tables for comparing options, pro/con lists for debates, or categorized lists for resources like podcasts, videos, and books. You can even customize language complexity—graduate-level analysis vs beginner-friendly simplicity.

Provide context about your existing knowledge and audience

Mention what you already know. If you’ll be sharing a report with colleagues, clarify that specific audience’s context. If you want something brief, say so.

Be patient. Quality research isn’t instant.

While Gemini and Perplexity typically deliver results within a few minutes, ChatGPT's deeper analysis can take a half hour. The thoroughness of these results justifies the wait vs. instant but shallower search results.

Research Rage: our need for speed can lead us to resent waiting. Image generated with ChatGPT 4o based on my cartoon idea

Top Tools for AI Deep Research 🧐

ChatGPT, Gemini and Perplexity are the three most useful tools I’ve tested for deep research. All work on desktop and mobile and can be used for free.

Paid vs free tiers: Paid Gemini and ChatGPT users ($20/month) get access to higher-quality models than free users. Perplexity Pro users ($20/month) avoid the free account limit of three research queries per day.

ChatGPT Deep Research

• Strengths: Most thorough results of any platform I tested. it can generate 30 to 50-page comprehensive reports customized for your needs.

• Limitations: Results can sometimes take 30 minutes.

• Cost: Free ChatGPT users get 5 queries per month, as of April 24, 2025. Plus, Team, Enterprise, and Edu users get 25 queries per month. Pro users get 250.

• More guidance: Read ChatGPT’s Deep Research introduction and watch Andy Stapleton’s YouTube video on how to use it to save research time.

Gemini Deep Research

• Strengths: Available even on free accounts; you can generate instant, free, shareable audio overviews summarizing your findings, akin to what NotebookLM produces; handy button lets you immediately export results into a Google Doc; you can edit Gemini’s research plan after it responds to your initial prompt, giving you additional agency after the initial prompt.

• Limitations: Reports aren’t quite as thorough as those from ChatGPT

• Cost: Free to use, though Pro users get access to a stronger 2.5 model. Pro access is temporarily free for students with an edu address. Apply here.

• More guidance: Google’s Gemini Deep Research overview, and 6 tips for getting the most of it

Perplexity Deep Research

• Strengths: Fast results in under 3 minutes, detailed citations that you can copy/paste with the content into a doc for editing or further research. Helpful follow-up questions are provided or you can ask your own follow-ups.

• Limitations: Less thorough than ChatGPT's deep dives

• Cost: Free for 3 uses a day, or unlimited for Pro

• More guidance: Perplexity’s Deep Research overview

---

What insights or challenges have you encountered with Deep Research? Leave a comment below to share your experience.👇