Research has always been my passion. Some people scroll through social media before going to bed, I do archaeological studies on color usage and obsessively blog about the evolution of old television shows. OpenAI’s Deep Research feature sounded like the perfect companion. It could conduct real-time web searches and compile detailed reports. Deep Research was originally an exclusive feature of ChatGPT’s $200 per month subscription, but it is now available for those who pay $20 monthly for ChatGPT. However, you can only run 10 Deep Research queries each week.
Deep Research is a tool that does what I and many others do when we are faced with a question that is too large for a quick Google search. Researching something usually involves clicking through multiple sources to separate marketing fluff and useful insights. It also means resisting the temptation to fall down unrelated Wikipedia rabbit hole. Deep Research promises to do all the work for you, and give you a nicely formatted report in return. ChatGPT isn’t the only AI that’s trying to solve this problem. Google Gemini, DeepSeek and Perplexity all have features with the same name. ChatGPT’s Deep Research is a system with its own quirks. But, in theory, it aims to produce something more structured and thoughtful than a handful search results.
To test it, I chose three research challenges that I thought would be fun. The reports were impressive but sometimes a little wandering. Imagine a brilliant librarian who is also a bit absent-minded. She can find an obscure 18th century manuscript in less than five minutes, but will occasionally hand you a 20 volume dissertation when you request some beach reading.
Espresso
My initial request was for help in choosing an espresso machine. ChatGPT asked for more information about the price and other details. This led to the following prompt: “Provide a guide to setting up an at-home espresso station for beginners, including recommendations for budget-friendly espresso machines, grinders, and accessories, along with maintenance tips and common beginner mistakes.”
While a normal ChatGPT response can be almost instantaneous, Deep Research may take up to 30 minutes depending on how complex your request is. Deep Research provided a very detailed espresso setup guide in ten minutes.
The guide covered everything, from machine recommendations (the Breville Bambino and Gaggia Classic Pro among others) to grinder choices, the importance fresh beans and even a short lesson on coffee extraction. It also covered common beginner mistakes, such as using pre-ground beans, not weighing shots correctly, or ignoring an important milk frother. There were quirks. Some of the product recommendations were skewed towards expensive options when there were cheaper alternatives. It was a very thorough, helpful guide. I liked that.
Star search
I asked for a local hobby I’ve been considering, and I got this response: “Provide an overview of beginner-friendly astronomy, including necessary equipment, recommended resources for learning, and local astronomy clubs or events in the Nyack, New York area.”
Deep Research provided an introductory guide on amateur astronomy, covering telescopes, eyeglasses, and binoculars. It provided some great recommendations for equipment and locations to visit, as well as websites and apps that could help me plan my evenings out and celestial events. It also listed nearby astronomy clubs and groups I could join.
Although the report was neutral, it had a lot of enthusiasm that I found charming. The suggestions for equipment were reasonable and did not always go with the most expensive options. It suggested some outdated event information, but it was more the fault for the websites that weren’t updated.
Monster mash
I chose a less factual report to see how AI would handle a report that was based mainly on rumors. “Investigate the origins and history of the ‘Lake George Monster’ legend, analyzing its first known mentions, how it evolved over time, and whether there is any real historical basis behind it.”
It took the least amount of time, about five minutes. Maybe a fictional character that is only a local legend requires less time to research. Deep Research provided a surprising breakdown of the Lake George Monster – a local legend from upstate New York. It traced back the first major sightings to the late nineteenth century, citing newspaper clippings describing a mysterious snake-like creature lurking below the lake. It explained that the legend was fuelled by hoaxes. One of these included a 20th century prank in which a local prankster built a mechanical sea monster.
The report also attempted to analyze whether a real creature could live in the lake by referencing known aquatic animals and scientific skepticism around such myths. This was the most entertaining report I’ve ever read. It was written like a local historian, complete with citations of sources and anecdotes. It even referred to other lake monsters like Champ, from Lake Champlain. This was done in order to draw comparisons between local folklore. It wasn’t perfect. It did a good job of recounting past sightings as well as debunking hoaxes. However, it had a hard time separating firsthand accounts from modern retellings. It was a fascinating read about a quirky part of local culture.
Deep thoughts
Deep Research was one of the most ambitious AI programs I’ve ever tested. I must admit, I love it. It’s like getting a report from someone who loves to dig through the weeds and find hidden gems just as much as I enjoy doing. I’ll say that compared to the regular ChatGPT responses, Deep Research seemed to make a real effort in finding fresh, relevant information. It’s not perfect, but it does a great job of creating easy-to read reports that save you time and effort. You don’t have to click through endless links and articles, or wonder if a recommendation was helpful or just a cleverly-disguised ad. Instead, you can get a report which at least tries distilling everything for you.
It’s not something I would trust to choose a car seat, but it could be a good place to start my own research. All AI research tools are like that, of course. While a librarian, search engine or AI report cannot replace the hard work of finding and organizing information, they can help to streamline the process.
With the understanding that deep researching is not actually deep research, this can be a great tool to get an early start.
ChatGPT Tasks replaced my to-do-list and completely changed how I plan my day
Eric Hal Schwartz has been a freelance writer at TechRadar for more than 15 years. He has covered the intersection of technology and the world. He was the head writer of Voicebot.ai for five years and he was at the forefront of reporting on large language models and generative AI. Since then, he has become an expert in the products of generative AI, including OpenAI’s ChatGPT and Anthropic’s Claude. He also knows Google Gemini and all other synthetic media tools. His experience spans print, digital and broadcast media as well as live events. He’s now continuing to tell stories that people want to hear and need to know about the rapidly changing AI space and the impact it has on their lives. Eric is based out of New York City.
Latest in Artificial Intelligence (19659037)
More on artificial intelligence
