ChatGPT Deep Research Dives Into the Internet—But Not the Entire Ocean
OpenAI's new research tool promises depth but has its no-swim zones
Have you ever fallen asleep with your face on the keyboard while researching for work or school? I sure have. There I was at 3 AM, with a half-drunk cup of coffee that clearly failed its one job ☕.
Last month, OpenAI rolled out "Deep Research" for ChatGPT—a feature that claims to do all that tedious research work for you. The pitch sounds amazing: throw a question at it, go do literally anything else (sleep, preferably), and come back to find a complete report with sources all neatly packaged.
After playing with it myself, here's my take: Deep Research is basically a diver who promises to explore the entire ocean but actually only swims through certain areas while completely avoiding others. Useful? Definitely. But it has its limits.
I'm still getting to know this feature, but I wanted to share my early discoveries so you can decide whether to add it to your workflow now or maybe wait for it to evolve a bit more.
Let's break down what it can do, what it definitely can't do, and where it actually makes a real difference.
How Deep Research Works
When you ask a question, Deep Research starts working just like a human researcher would, but much faster.
Here's what it looks like:
You type:
"I need to compare the sustainability strategies of the main clothing brands"Meanwhile, Deep Research is thinking:
- "Okay, I first need to figure out which are the main brands"
- "Now I need to find their sustainability reports"
- "Hmm, looks like Nike takes a different approach than Adidas..."
- "Interesting—this independent source contradicts what Puma claims"
- "Let me gather some specific data on carbon reduction"What makes Deep Research really useful is that it can change direction when it discovers new information. You know how you might start looking up something simple and end up in a completely different place? Deep Research can make those same connections—but without losing sight of what you originally asked.
The system can work with text, images, and even PDFs. It can run code to analyze complex data. And while it's busy doing all this, you can be doing literally anything else (or just relaxing—no judgment here).
The best part? This whole process usually takes about as long as watching an episode of your favorite show—somewhere between 5 and 30 minutes, depending on how complex your question is.
How to Start Using Deep Research (in Under 3 Minutes)
First, the bad news: you'll need a ChatGPT Plus subscription ($20/month). The good news? You probably spend more than that on coffee each month anyway.
Here's the quick setup:
Open ChatGPT in a web browser (they haven't added it to the mobile apps yet)
Look right under your chat box for the dropdown menu
Switch it to "Deep Research" instead of the regular model
Type your question (I'll share some tips on this in a sec)
Click "Send" and... that's it! Go grab a coffee, scroll through your emails, or just stare out the window contemplating life's mysteries
You'll notice a sidebar that shows what your digital helper is up to. It's like watching a detective work their case—just faster and without the cool hat.
The real trick to getting useful results comes down to how you ask your question. Think of it as talking to someone super smart but incredibly literal:
"Tell me about investments" ← Nope, too vague
"What are the 3 best investment strategies for 30-year-old professionals with medium risk tolerance in today's Latin American economy?" ← Now we're talking
When it's done, you'll get a complete report with sources you can actually check out yourself.
What Deep Research (Still) Can't Do For You
Let's talk about this tool's limitations. You should know about a few key issues before you trust it with your next big presentation.
The "No-Go Zones" Problem
Picture a detective who's banned from certain neighborhoods in the city. That's Deep Research in a nutshell. During my testing, I found it straight-up can't access entire platforms like Medium—which is packed with valuable content.
When I specifically asked it to check out articles on these platforms, I got shutdown messages like this:
This isn't just some minor glitch—it's like trying to research music without being able to access Spotify or YouTube. Why? Many of these sites either have paywalls or they're actively blocking ChatGPT.
The Black Box of Source Selection
Here's what bugs me most: we have zero insight into how it chooses which sources to explore and which to skip over. Does it favor certain websites? Avoid others? Is there some hidden bias in its search algorithm?
It's basically your assistant saying "just trust me" without explaining how he works. NO _______ WAY.
The Hallucination Issue
While OpenAI claims Deep Research hallucinates less than regular ChatGPT, it's definitely not 100% reliable. It still sometimes makes up data or misinterprets information in subtle but meaningful ways.
Sure, it provides citations, but if you have to verify each one yourself, aren't you losing half the benefit of using the tool in the first place?
To be clear, these limitations don't mean you should write it off completely—just use it with your eyes wide open.
Who Should Use Deep Research? (And Who Shouldn't)
Deep Research isn't everyone's cup of tea, but it might be perfect for you depending on what you're after.
Deep Research is your friend if you're:
The juggler of a million things — Your day is split between countless tasks, and you need quick intel on topics outside your wheelhouse without burning hours on research.
The solo entrepreneur hustler — No fancy market research team backing you up, but you still need to track what competitors are doing and spot trends in your industry.
The "I'm already late" professional — You've got a big presentation tomorrow on something you only kinda understand, and you need enough context and data to not look like a deer in headlights.
Where this tool really shines:
When you need the big picture on a topic you're just getting into
As a launchpad for developing your own ideas
Look elsewhere if:
You need super specialized, in-the-weeds information
Your research requires subscription-only content
You're working with data that changes hourly
Being 100% accurate is make-or-break important
Here's the real deal: Deep Research is basically like having a smart but junior research assistant. They're impressive at certain tasks but definitely need supervision and fact-checking. It won't replace your judgment—it just gives it a boost.
So, Do You Really Need This Digital Diver on Your Team?
After spending a few weeks with Deep Research, here's my take: it's fascinating and useful, but definitely still a work-in-progress.
It's like having an explorer who promises to show you the entire ocean, but actually only visits certain areas while completely skipping others. Pretty impressive? Absolutely. But there are whole sections of the internet it just doesn't touch.
What I'm loving:
Having an assistant doing the heavy lifting while I focus elsewhere
Getting those neatly organized, well-structured reports
Watching it process mountains of information in minutes
Seeing it connect dots between totally different sources
What's still bugging me:
Those "no access" zones of the internet it simply can't reach
The total lack of transparency about how it picks its sources
Those occasional hallucinations that force you to double-check things
OpenAI has ambitious plans for Deep Research—they're talking data visualizations, deeper analysis capabilities, and integration with other features.
Here's my bottom line: If you're already paying for ChatGPT Plus, definitely use Deep Research for tasks where it shines. If you're considering subscribing just for this feature? Maybe wait a few months to see how it evolves (and if you're wondering when it's actually worth paying for an AI assistant at all, I've got a whole article on that topic).
At the end of the day, Deep Research is like any tool—it's all about how you use it. It won't replace your judgment or critical thinking, but it can definitely help you get to the information you need faster, which leads to better decisions.
What about you? Have you taken Deep Research for a spin yet? What have you used it for? Drop your experiences in the comments!
See you soon!
G
Hey! I'm Germán, and I write about AI in both English and Spanish. This article was first published in Spanish in my newsletter AprendiendoIA, and I've adapted it for my English-speaking friends at My AI Journey. My mission is simple: helping you understand and leverage AI, regardless of your technical background or preferred language. See you in the next one!