Making a big decision? See how to use AI to dig up the facts.

Welcome to Your AI Guide — a 30-day challenge where I explore what AI can (and can't) do in everyday work. Each day, I'll introduce one AI tool or tip and break it down in simple steps to apply yourself. Subscribe to follow along, weigh in with #30DaysofAI and let's learn together.

My university career was cut short by a disability, or rather a mismatch between me and my environment: I have dyslexia and read at speaking pace. As I started exploring whether I should pursue a PhD in philosophy, reality came knocking with the enormous volumes of books and articles I’d need to read through just to find the small percentage that would be relevant to my research. Doing just the reading necessary for my study would take more time than was available to me, so in the end, I was forced to abandon my academic aspirations and shift my focus elsewhere. 

So in a way, you can blame the lack of accommodations at my alma mater for me being here teaching you about AI today! 

What I needed back then was assistance in rapidly combing through massive tomes of philosophical opinionations to find just the right parts I needed to spend my time on. Today, we’re seeing the outlines of just such a class of tools appearing in your favourite AI chat applications.

Dubbed “Deep Research,” these tools add an “agentic” layer to the now classic AI chat interfaces that can generate “plans” based on your original prompt, do internet searches based on those plans, “read” the results, make further searches, and keep looping through this process until some threshold is reached and a comprehensive summary article, complete with footnotes and direct links to the original materials is written.

Originally, Deep Research tools were confined to web searches, but now these tools have more personalized capabilities: ChatGPT Deep Research can connect to GitHub and OneDrive, Claude Research can connect to Google Drive and any MCP resource you provide, and Microsoft Copilot can connect to all Microsoft resources.

The key to success when using these extended capabilities is understanding that Deep Research isn't a research tool - it’s an enormously advanced and comprehensive search and  summary tool that generates text that looks like research. That makes them great for getting a second opinion that measures your ideas against everything available on the internet, and a "rough sort" that gives you a comprehensive starting point for deeper analysis.

The step-by-step breakdown 

Here’s an example: Last year we installed solar panels on our house in Vancouver, and I was surprised how immature the green energy market is here compared to Europe. Now I'm wondering if it's smart to invest in some sort of green energy business locally. To get a handle on the ins and outs, I crafted this prompt and sent it to three Deep Research systems: ChatGPT, Perplexity, and Claude.

Try it for yourself:

1. Pick an AI chat app with Deep Research capability. In ChatGPT and Gemini, it’s called “Deep Research”. In Claude it’s called “Research” with an added option for “Extended Thinking.” In Copilot you can choose between “Quick,” “Think Deeper,” and “Deep Research” modes. In Perplexity you can toggle between Web, Academic, and Social research.

2. Craft a clear language prompt to start the research. Here’s the one I used:

I'm considering investing in the green energy industry in Vancouver, BC, Canada, specifically solar, battery storage, and geothermal heating and cooling. The region seems well suited for these technologies, but compared to Europe projects are expensive and difficult to get off the ground. Help me understand the viability of the market, what's causing the slow adoption compared to other regions, and whether this is a worthwhile industry to invest in over the next 3-5 years.        

3. If the system asks clarifying questions, answer them in as much detail as possible

4. Take a break. This can take some time. For the above prompt, Perplexity and Claude both worked for around 15 minutes.

5. Review the results. Check the links, and ask follow-up questions.

What worked (and what didn't): While Claude and Perplexity surfaced a long list of sources and provided comprehensive breakdowns of my query, ChatGPT asked some follow-up questions and provided a shallow and vague response in seconds. It’s possible the Deep Research system didn’t kick in for some reason, but I tested numerous times with prompt variations and got the same general result.

The Claude and Perplexity responses both produced similar numbers and recommendations, varying only in structure and minor details. As a second opinion exercise this showed the value of running the same query across multiple platforms. I’m more likely to trust what I see if different systems produce similar summaries from different sources.

Go a step further: Having seen what these systems can do with data from the web, I am now curious to test them out on more custom sources, and I encourage you to do the same if you feel so inclined:

• Hook ChatGPT up to your GitHub account and ask questions about your projects
• Test Claude with a connection to your Google Drive and see if it is able to derive new meaning from your old documents
• Ask Copilot to look through your OneDrive and SharePoint data to find new insight
• Combine all of this with web search to find your own blind spots

The verdict: When it works, no matter what AI app you choose, Deep Research feels very much like having a research assistant in my pocket. I can send off a query, put my phone away, and when I come back some time later I have a huge document with lots of info to digest. And as these AI chat apps get more audio capabilities, I’ll soon be able to listen to the output which will be very helpful for my dyslexia!

Is Deep Research a replacement for a real research assistant and your own research? Absolutely not. Personally I wouldn't trust the output of these tools any more than I would a Google search. But like a Google search, they are still a huge help in finding sources I would otherwise have missed.

Here’s the mental model I use for these tools:

Deep Research is like telling someone you are curious about a topic, and then they go on a speed run through every resource they can find and hand you back a complete summary before you had a chance to give them any feedback. It may be super helpful, it will definitely give you a list of possible sources to explore further, and there’s also a good chance what you get in return isn’t what you actually needed and it’s all a very small waste of time.

Now it’s your turn: Go test out Deep Research on some tasks where you’d otherwise either do or forego research, and report back how you feel about the results. Was it useful? Did it save you time? Or was it just noise that distracted you from getting your work done? Write a post or share a video using #30DaysofAI, leave a comment below, and let’s talk about how we make AI a practical part of our research!

Thanks for reading! If this helped you make sense of AI, share it with a friend who's trying to keep up, too. Hit save so you can come back to it later – and if you're not subscribed yet, now's a good time to fix that.