How to Use Deep Research Like a Pro

Image by Author | Canva

In the age of AI and increasing data, finding the correct information and answers is not easy. This is where deep research can come in handy, which will help you find accurate information faster. But what exactly is deep research?

Unlike regular search in AI tools like ChatGPT, deep research enhances the process of collecting and combining information from multiple web pages at the same time. It delivers a user-friendly and detailed report from minimal inputs by creating an action plan. It does this by referencing or even sometimes rephrasing your prompt. This process can involve manual browsing or note-taking.

Deep Research Tools

 
There are several deep research tools available.

• ChatGPT Deep Research is good at combining information across multiple sources to create a user-friendly report with minimal input
• Perplexity AI creates citation-backed answers more quickly
• Scite.ai and Consensus focus on heavily scientific literature
• Gemini stands out with interactive canvas-based research and multimodal summaries like voice notes

Let’s try using deep research to answer a question.

Using Deep Research to Answer a Real Question

 
“What are the machine learning models that win the most competitions?”

Image source: Kaggle

Kaggle offers competitions for creating the best machine learning models. But what are the best machine learning models that tend to win competitions? Let’s use deep research to find out which models have an advantage in these competitions involving machine learning.

In this walkthrough, we will explore how to use ChatGPT’s Deep Research feature step by step to find an answer to our question.

Step 1: Enable Deep Research Mode

 
First, click on the “Deep research” setting.

Step 2: Writing the Prompt

 
As I mentioned, we want to determine which machine learning models are winning the Kaggle competitions. Here is the prompt I use:

I want to analyze the machine learning models and techniques most commonly used by top Kaggle competition winners in the last 5 years. What trends can be identified in their approaches, including feature engineering, model choice, ensemble methods, and data preprocessing strategies?

Step 3: Clarifications/Requirements

 
In the Deep Research functionality, ChatGPT typically asks you a few questions to clarify your requirements. It’s a smart move because you don’t want to do a deep search into a topic you didn’t explain clearly.

Here are my answers.

1. I’m interested in competition categories involving predictions.

2. The analysis should focus on all winning models from competitions with prizes above $ 10 K.

3. I’d like to see the results in a report with tables and dedicated sections for the top 3 methods.

4. No, only competitions with prizes above $ 10 K.

Now, ChatGPT starts the research. You can see the progress bar.

Step 4: Let the AI Do the Research

Behind the scenes: How Deep Research Works

But what is happening in the backend? It writes what is happening in the sidebar in the “activity” panel. It first creates tasks and then visits websites based on these tasks. For instance, the first task was mapping out Kaggle competitions from 2020-2025 with prize pools over $10,000, focusing on winning solutions involving prediction tasks like tabular regression/classification and time series.

Track Research Progress in the Activity Panel

To do that, it reads from Kaggle, and the progress is displayed in the activity panel on the right.

You can also explore the websites ChatGPT visits to perform its research.

Step 5: Reviewing the Final Report

 
It’ll take a few minutes, but once the research has finished, a full report with headers and sections will be generated.

Time, Sources, and Search Summary

 
13 minutes using 68 sources and 44 searches.

It also includes a section that shows the references.

Also, if you click the arrow button, the preview pops up, and here you can share the link of this research or download it as a PDF.

Conclusion

 
While general web search functionality of large language models will give you a proper explanation about the topic, deep research is excellent for the topic you want to add an extra layer of depth, structure, and credibility, by giving features like tasks, sources, and activities.

If you're a data scientist, student, or anyone working on research — heavy tasks, deep research tools can save you hours. You get citations, task logs, and a final report that’s easy to read and share. And the best part? Some of the best tools are free.
 
 

Nate Rosidi is a data scientist and in product strategy. He's also an adjunct professor teaching analytics, and is the founder of StrataScratch, a platform helping data scientists prepare for their interviews with real interview questions from top companies. Nate writes on the latest trends in the career market, gives interview advice, shares data science projects, and covers everything SQL.