(Credit: Jonathan Kitchen / DigitalVision via Getty Images)
Whenever I have to research a specific topic, I naturally turn to the web. However, I don't always have the time to spend hours reading every article out there. Instead, I can ask an AI chatbot to tackle much of the initial work for me. A variety of popular AIs offer a Deep Research mode, which performs online research on your behalf. Just submit a request, and the AI searches the web for any relevant information. It then generates a detailed report, complete with sources.
You can try Deep Research with ChatGPT, Google Gemini, Perplexity AI, and Grok. To put them to the test, I asked each one to research how GPS developed from its military origins to the commercial system we all rely on today. Here’s how each AI fared and which one was my favorite.
ChatGPT
OpenAI’s ChatGPT provides two different versions of Deep Research. One is a full version, which offers an in-depth report that can take as long as 30 minutes to complete. The other is a lightweight option that creates a shorter and less in-depth report in only a few minutes. The version the AI uses will depend on your plan.
- Free: No access to the full version of Deep Research and 15 uses of the lightweight version per month.
- Plus, Team, Edu: Users get 10 full and 15 lightweight queries per month.
- Pro: users get 125 full and 125 lightweight requests per month.
After you hit the limit for the full version, submissions automatically switch to the lightweight one.
(Credit: PCMag / OpenAI)I used the ChatGPT website, but Deep Research is also available on the desktop (Windows, macOS) or mobile apps (iOS, Android). By default, Deep Research uses the latest GPT-5.5 model. As part of my test, I used my Plus subscription in one browser and a free account in another.
I clicked the plus icon and chose Deep research, then submitted my query about GPS through the full and lightweight versions of Deep Research. Both provided a game plan with bullet points of the items it would include. I could edit the game plan, click the Start button, or wait 60 seconds for things to kick off on its own.
(Credit: PCMag / OpenAI)After the research started, the AI told me what it was doing as it checked out different websites and fashioned its research. The details here were sparse. I had to wait until the research was completed to see all of the actions that were taken.
In my testing, the full version took a whopping 49 minutes to search the web and compile the results, which was much longer than I expected. The lightweight one took only around five minutes. In both cases, I was alerted once the process finished.
The full version served up a detailed and in-depth report that felt just long enough. The information covered the progress of GPS from its early military days to its current commercial use, providing a timeline, a list of the key ways that GPS is used, and a satisfying conclusion. The lightweight one offered a shorter report but still with a hefty amount of information, more than it typically provided in the past. Both addressed the topic to my satisfaction.
(Credit: PCMag / OpenAI)Google Gemini
Gemini's Deep Research mode is available to both subscribers and free users. In the past, Google calculated your Gemini usage based on a set number of credits per day or per month. Now, the company has turned to a compute-based model that factors in the complexity of your prompt, the models and features you use, and the length of your chat. As it's a more complex tool, Deep Research will eat up more of your usage than a regular query. Google breaks down the overall use of its AI as follows:
- Free plan: Standard limits
- AI Plus: Two times the standard limits
- AI Pro: Four times the standard limits
- AI Ultra ($100): Five times higher than AI Pro limits
- AI Ultra ($200): Twenty times higher than AI Pro limits
(Credit: PCMag / Google)To start, I typed my GPS query at the prompt, clicked the Plus icon, selected Deep Research, and submitted the request. Gemini’s Deep Research generated a research plan of how it aimed to address my topic, which I could revise or approve. When I gave it my OK, Gemini took to the web to conduct its research.
(Credit: PCMag / Google)With more detail than ChatGPT provided, Gemini kept me updated along the way, showing the actions it took and the websites it visited. The entire process took around eight minutes. The generated report was long and quite in-depth, neatly organized into logical sections with a timeline at the beginning and a table of different GPS systems at the end. The AI also included charts and diagrams to illustrate the topic, resulting in an interesting, informative, and effective report.
(Credit: PCMag / Google)Perplexity AI
Perplexity’s Deep Research tool is accessible to paid subscribers and free users. In early 2025, Perplexity noted that paid Pro users got 500 queries each day, while free users received five queries per day. Those numbers may have changed over time, but subscribers can still use Deep Research much more frequently than free users can.
To dive in, I signed in to the Perplexity AI website, though apps for iOS, Android, Windows, or macOS are also available. At the prompt, I clicked the Search button, changed the mode to Deep Research, and submitted my request.
(Credit: PCMag / Perplexity)I gave Perplexity’s research mode the same GPS topic to investigate. The AI kept me abreast of its mission, showing me which specific areas it was researching and which websites it was consulting. Perplexity compiled its findings and presented its report in about 3 minutes. The results were acceptable, but they paled in comparison to those from ChatGPT and Gemini. Perplexity’s report was quite short and lacked the depth and analysis provided by the other AIs. As a quick read, it was fine, but I wanted more, and it failed to deliver in that regard.
(Credit: PCMag / Perplexity)Grok
Like ChatGPT, Elon Musk's Grok also offers two flavors for its deep research mode. DeepSearch checks out a vast number of online sources, though not all of them are useful or reliable. This mode is also quick at completing its research. DeeperSearch is a more advanced version that consults more high-quality sources and takes longer to run, but should present a more in-depth report. Whichever mode you use, the number of queries you’re able to run depends on your subscription plan as follows:
- Free: 5 to 10 Deep Search queries and 2 to 3 Deeper Search queries per day.
- X Premium: 10 to 20 Deep Search queries and 5 to 10 Deeper Search queries every two to 24 hours.
- X Premium+, SuperGrok: 20 to 30 Deep Search queries every two hours and 10 Deeper Search queries every two hours.
- SuperGrok Heavy: Nearly unlimited number of Deep Search and 10 Deeper Search queries.
You can use Grok and DeepSearch directly in X (aka Twitter), its dedicated Grok web page, as well as its mobile app (iOS, Android). To get started, you're supposed to access Deep Search from the menu in Grok, but that option didn't appear for me. Instead, I typed the following: "Use Deep Search: Describe how GPS developed from its military origins to the commercial system that we all rely on today." I also ran the same type of query using Deeper Search.
(Credit: PCMag / xAI)Using DeepSearch mode, the AI took only five seconds to complete the process. DeeperSearch took around the same amount of time. The DeepSearch report was informative but short and lacking in any real depth or analysis. The DeeperSearch version was slightly longer and more in-depth but still lacked the detail and analysis I wanted.
(Credit: PCMag / xAI)Who Came Out on Top?
Based on my specific query concerning GPS, which AI performed the best? I give the nod to Gemini. Its report was the most thorough and in-depth, as well as easy and interesting to read. ChatGPT came in a close second, but lost points for taking a surprisingly long time to generate.
That still leaves the issue of veracity. We know that AI is fallible and prone to hallucinations. How can you tell if the information you get in a Deep Research report is accurate? Here, Deep Research offers a couple of benefits. First, this type of research typically consults the most reliable and high-quality online sources, rejecting ones that are deemed of lesser quality. Second, you can easily view the original sources to double-check the results. Yes, that can be time-consuming to do with a lengthy report. But you can still check a few points in the report with the corresponding sources to see how they match up.
Of course, this is all based on just on one query. Still, I'd favor Gemini or ChatGPT the next time you need an AI to research a topic for you.