ChatGPT’s Deep Research

In today’s episode, we'll explore the features and functionality of Chat GPT's Deep Research and compare it to Google's Gemini Deep Research. Visit AVID Open Access to learn more.

Transcript

Paul Beckermann 0:00 Welcome to Tech Talk for teachers. I'm your host, Paul Beckermann.

Transition Music with Rena's Children 0:09 Check it out. Check it out. Check it out. What's in the toolkit? Check it out.

Paul Beckermann 0:15 The topic of today's episode is ChatGPT Deep Research.

Paul Beckermann 0:22 Last week, I explored Google's Gemini Deep Research feature. I was impressed and intrigued enough to compare it with OpenAI's version of Deep Research in ChatGPT. I found both similarities and differences between the two experiences. In today's episode, I'm going to describe how Deep Research works, walk you through my experience with it, and highlight the key similarities and differences I found between Gemini and ChatGPT.

First, a quick review of what Deep Research is in both platforms. It's a more deliberate research process grounded in specific sources. These can include online content the AI finds on the web or materials you upload directly. The process is more intentional and comprehensive than a standard generative AI query. It takes longer and involves the AI adjusting its plan as it learns more, much like a human researcher who would ask better questions as they go and target new information based on emerging gaps. It's a pretty cool tool for consistency.

I used the same prompt I had tested with Gemini. The prompt was: "Act as a K-12 educational research expert. Research best practices in using generative AI in K-12 schools, focus on US schools during the past two years, generate a list of the top 10 most popular and effective strategies". Here's a summary of what I experienced in ChatGPT.

Transition Music with Rena's Children 1:47 Here's the tool for today.

Paul Beckermann 1:51 To begin, I set up a free ChatGPT account and signed in to access Deep Research. Once logged in, I located Deep Research, typically found in the mode menu next to the model selector, though the exact placement may vary depending on your device. After selecting Deep Research, I pasted in my prompt and hit ENTER.

This is where I encountered the first major difference from Gemini: before starting the search, ChatGPT asked for follow-up questions to tailor the search. It asked which grade levels to focus on, whether to limit results to classroom use or include administrative and PD uses, which subjects to include, and whether to look at public schools, private schools, or both. These were thoughtful, clarifying questions, and I felt they improved my original prompt. I selected all grades, classroom only, all subjects, and both public and private schools. Then I initiated the search. ChatGPT Deep Research opened a side panel showing live progress.

Paul Beckermann 2:56 I could toggle between the list of sources being reviewed and an activity log describing its steps. The activity view functioned like a curated research transcript. It might say, "I'm gathering a range of sources to identify popular and effective instructional strategies in K-12 education, including generative AI strategies with reported outcomes or benefits," followed by the search it was examining and a brief summary. For example, it noted strategies such as using ChatGPT as a tutor, DALL-E for content creation, and adaptive AI for personalized learning. It then moved to the next source and repeated the pattern.

Another noticeable difference emerged here as well. Gemini provided a detailed research plan before starting its work. ChatGPT did not show an upfront plan, but instead offered a play-by-play log as it progressed. I also compared the number of searches in my run. ChatGPT completed 91 searches using 26 sources, while Gemini examined over 100 sources and cited 29.

Paul Beckermann 3:55 These differences are negligible, though, and will vary by time of day, prompt wording, and site indexing. Processing time can vary as well. In my case, the ChatGPT report finished in about eight minutes, roughly half the time of my Gemini run. However, ChatGPT was tested on a Saturday and Gemini on a weekday, so the comparison isn't exact. The key takeaway is that Deep Research takes significantly longer than a normal generative AI query.

One feature I appreciated in ChatGPT was the "Update" button, which allows you to modify your query mid-task without losing context. Gemini allows revisions too, but only before starting the research. When ChatGPT finished, it presented a clear summary of the top 10 AI strategies. Citations were included as endnotes rather than inline links.

Paul Beckermann 4:56 I thought Gemini citations were stronger because they included inline citations throughout. Interestingly, when I ran the same query using my paid version of ChatGPT, the citations appeared inline, very similar to Gemini. ChatGPT notes that citation styles do vary depending on version and settings.

As for the final report itself, ChatGPT delivered exactly what I asked for: a list of the top 10 strategies. Gemini also provided a top 10 list, but it added a much longer report with additional context and background, which was impressive, but more than I had requested.

In terms of sharing, ChatGPT allowed me to post to social media, export a PDF, save the project internally, generate a shareable link, or copy and paste the text.

Paul Beckermann 5:58 Gemini offered similar options, but also included the ability to export directly to Google Docs and generate additional formats, like web pages, infographics, quizzes, flashcards, and audio overviews. In general, Gemini and ChatGPT are more alike than different. Both provide significantly deeper research than a standard chat bot query. Both require you to sign in to access Deep Research and allow you to upload your own files for analysis. Both use multi-threaded search processes, even if Gemini displays more simultaneous activity, and share progress updates as they work. They review a similar volume of sources and produce strong results even in the free versions.

There are some important differences, though. Gemini is more structured; ChatGPT is more adaptive and dynamic. ChatGPT refines your prompt through follow-up questions. Its progress updates are shorter and focused on one source at a time.

Paul Beckermann 6:55 Gemini provides a detailed search plan upfront and allows initial revisions, while ChatGPT makes mid-search changes easier. Gemini also provides more detailed thoughts or notes logs, depending on the interface version you're using.

The final reports differ as well. Gemini's outputs are longer with more background information, while ChatGPT stays tightly aligned to the end product you requested for my prompt. ChatGPT's top 10 list was cleaner and more focused, while Gemini's richer context came with stronger inline citations. In my paid version of ChatGPT, citation style aligned closely with what I got in Gemini.

I also noticed differences in datasets, accuracy, and grounding. Gemini draws heavily from Google Search, YouTube, and Scholar when available. ChatGPT uses its browsing tool, web extraction, and any uploaded materials. As a result, Gemini tends to be more explicitly source-linked, while ChatGPT writing feels more polished, but sometimes less directly tied to citations.

Paul Beckermann 7:42 Regarding pricing, Gemini Deep Research is currently free. ChatGPT offers both free and paid versions with some enhanced features available to paid users. Access may also depend on school district policies. Many Google Workspace districts have easier access to Gemini, while ChatGPT access may be more restricted.

Paul Beckermann 7:55 In the end, I liked both experiences. Each has pros and cons, and both can be powerful tools for teachers, researchers, and content creators. I'd encourage you to try both and see which one fits into your workflow, or even use both and combine the strengths of each final report.

Paul Beckermann 8:13 To learn more about today's topic and explore other free resources, visit Avid Open Access.org. Specifically, I encourage you to check out the article collection, "AI in the K-12 classroom," and, of course, be sure to join Rena Winston and me every Wednesday for our full-length podcast Unpacking Education, where we're joined by exceptional guests and explore education topics that are important to you. Thanks for listening. Take care and thanks for all you do. You make a difference.