I tested ChatGPT's Deep Research against Gemini, Perplexity, and Grok AI to see which is best

More AI chatbots now offer a deep research option, through which they can investigate a topic for you. Acting as autonomous AI agents, the bots will surf the web on your behalf, find the right online sources, and then present you with a detailed report based on their findings. The goal is to save you the time of checking out dozens or hundreds of websites yourself.

Deep Research is quickly becoming a powerful feature among a host of AIs. You’ll find it with OpenAI’s ChatGPT, Google Gemini, Perplexity AI, and even xAI’s Grok (which calls it DeepSearch). Microsoft introduced a type of deep research with two AI agents, Researcher and Analyst; however, they require a Microsoft 365 Copilot license with an Enterprise or Business subscription, so they’re not yet available to the average Copilot user.

Also: The telltale sign that you used ChatGPT – and a trick to avoid it

This certainly sounds like a useful and helpful skill. But how do different AI services fare when presented with this challenge? To answer that question, I tried Deep Research across ChatGPT, Gemini, Perplexity AI, and Grok. I submitted the same query at each one, asking it to “explore how time travel is portrayed in film and television and what it says about our values, fears, and desires.”

Here’s how each AI’s research mode works and how each handled my topic.

ChatGPT

OpenAI’s ChatGPT offers two different flavors of Deep Research: the full version and a lightweight version. The full version delivers a detailed, in-depth report but can take up to 30 minutes to find the best sources and present its findings. The lightweight version provides a much shorter, less in-depth report, but typically takes only a few minutes to finish. Which one is available to you, and how many queries you can submit, depends on your plan.

ChatGPT Plus, Team, and Edu users get 25 queries per month (10 full and 15 lightweight), Enterprise users get 10 (all full), Pro users get 250 (125 full and 125 lightweight), and free users get 5 (all lightweight). Once you reach your limit for the full version of Deep Research, your queries automatically default to the lightweight version.

Also: How to use ChatGPT: A beginner’s guide to the most popular AI chatbot

Whether you’re using the full or lightweight version, the process is the same. Head to the ChatGPT website or open the app for iOS or Android. You can also use the desktop app for Windows or macOS. Type or speak your query at the prompt, select the button for Deep Research, and then submit your request. If the full version is in effect, prepare to wait a while for the response. If the lightweight edition is in play, you won’t have to wait as long.

I submitted my query about time travel in film and television to both the full Deep Research version and the lightweight version — using my Plus subscription in the first instance and a free account in the second. Both used the GPT-4o model. Both also asked me to clarify the kind of analysis I wanted, such as deciding between a thematic take or a more historical one and whether to include only classic films and TV shows or modern ones as well.

Also: 5 reasons I turn to ChatGPT every day – from faster research to replacing Siri

The full version took about 17 minutes to search the web and compile the results, but it delivered a detailed, in-depth report with several examples and a helpful chart of TV shows and movies. The lightweight version took only about eight minutes from start to finish, but it provided a shorter, much less in-depth report — a Cliff’s Notes version of the full report. Both reports addressed my topic and were interesting to read, though the full Deep Research version gets the nod for thoroughness.

Google Gemini

Gemini’s Deep Research mode is available to both subscribers and free users. Subscribers typically get 20 queries per day, though that number could vary. Free users are limited to 5 queries per month. To try this, go to the Gemini website or set up the mobile app for iOS or Android.

Also: Why Gemini Deep Research is my new favorite rabbit hole finder – and it’s free

To enable Deep Research, click the dropdown menu at the upper left that lists the current model. Subscribers can choose among 2.0 Flash, 2.5 Pro (experimental), and Deep Research with 2.5 Pro. Free users can select either 2.0 Flash or Deep Research. After choosing the model you want, the button for Deep Research should appear under the prompt. Type your question at the prompt, select the button for Deep Research if it’s not already highlighted, and then submit your request.

After I submitted my query about time travel, Gemini’s Deep Research quickly generated an outline of how it planned to tackle my topic, which I could tweak or approve as is. I gave it a thumbs up, and Gemini took to the web to conduct its research.

The AI kept me informed at each step of the way, indicating the actions it was performing, the websites it was consulting, and the ways the report was progressing. The entire process took around 10 minutes.

Also: 5 easy Gemini settings tweaks to protect your privacy from AI

The resulting report was quite in-depth, thorough, and long. I liked the included table with examples of the films discussed. Gemini’s writing style was more scholarly than ChatGPT’s, which was less formal and more interesting to read. But Gemini still proved worthy of the task.

Perplexity AI

Perplexity’s Deep Research mode is available for paid subscribers and free users. Pro subscribers get a whopping 500 queries per day, while free users receive five queries each day. To get started, head to the Perplexity AI website or use the app for iOS, Android, Windows, or macOS. At the prompt, type your question, select the Research button, and submit your request.

I gave Perplexity’s research mode the same time travel topic to chew on. Here, the AI kept me abreast of its progress, telling me which specific subtopics it was researching and which websites it was analyzing. Perplexity took only around five minutes to compile its findings and present its report. But the results were disappointing.

Also: Perplexity is the AI tool Gemini wishes it could be

The report itself was much shorter than those generated by the other AIs. Each theme or element garnered only a couple of paragraphs, most of which were lacking in any in-depth analysis. The report wasn’t bad as a quick read. But it reminded me of a homework assignment turned in by a kid who just wanted to get it done without expending too much time or effort.

Grok AI

xAI’s Grok 3 offers two variations of a deep research mode: DeepSearch and DeeperSearch. DeepSearch looks at a large number of online sources, though not all of them are useful or reliable. This mode also gets the job done very quickly. DeeperSearch is an upgraded version of DeepSearch that taps into more high-quality sources and takes longer to run but typically delivers a more in-depth report.

Also: X’s Grok did surprisingly well in my AI coding tests

Whichever mode you choose, X Premium+ subscribers enjoy an unlimited number of queries, while free and Basic users are limited to 10 DeepSearch queries every 24 hours. X also offers a $30-per-month SuperGrok plan with extended access to DeepSearch and DeeperSearch. Beyond trying Grok 3 and DeepSearch directly in X (aka Twitter), you can also use it through the dedicated Grok web page, the iOS app, and the Android app.

To try this, type your query at the Grok prompt, click the download arrow for DeepSearch, and choose either DeepSearch or DeeperSearch. When ready, submit your request.

I tossed the same time travel query to Grok. With DeepSearch mode, the AI took only a minute and a half for the entire process, a speed record. DeeperSearch took slightly longer, clocking in at two and a half minutes. Given the speed, I was expecting a poor report in response. But the results surprised me. In both modes, Grok delivered reports that were interesting and informative, albeit brief. The research listed a variety of examples, a helpful table of TV shows and films, and some clever analysis. Not bad at all.

OK, so which AI fared the best? I’d have to declare ChatGPT the winner. Though it took the longest to finish the job, its report was the most thorough, in-depth, well-written, and interesting to read. Otherwise, if you’re in a hurry, Grok is certainly worth trying. Of course, all of this is based just on one query. With other topics, I might name a different champion. But the results are still worth considering the next time you need an AI to handle your own type of deep research.

Get the morning’s top stories in your inbox each day with our Tech Today newsletter.