Claude Releases Deep Research
Claude Research (beta)
Anthropic recently made the beta version of “Research” available to Pro members (the $20 per month plan). It’s been available to Max, Team and Enterprise users since mid-April but has received surprisingly little attention. I remember seeing an announcement about it, but little else. That’s unfortunate because Claude Research is a nice supplement to existing “deep research” offerings from OpenAI, Google, and Perplexity. I’m a HUGE deep research fan, despite it’s limitations. It has saved me more time than I can fathom.
Overall, I’m reasonably impressed with Claude Research, although I think I still prefer ChatGPT Deep Research. The bottom line is that Claude Research is a useful new tool for Claude users. I’ll use it from time to time, but it won’t replace ChatGPT for me.
Process
The process for creating a Research report is similar to that for ChatGPT, Gemini, and Perplexity. You provide a prompt describing the research task, Claude asks some clarifying questions and goes to work. Normally, I use meta-prompting to get AI to write a highly detailed prompt for research tasks. This time, however, I wanted to see what Claude would do with a relatively simple prompt.
Please create a detailed report on how AI will affect the level of employment for knowledge workers.
Claude responded with clarifying questions about the type of knowledge workers to consider (information systems primarily but all knowledge workers should be considered), time horizon (2 to 10 years), and whether I was interested in job displacement, job creation, other aspects (go broad). (My responses are in parentheses).
One major difference between ChatGPT Deep Research and Claude Research is that you need to select the underlying model for Claude. (As far as I can tell, ChatGPT chooses which model to use automatically.) I used Claude Sonnet 4 for the original report. The one I created using the more powerful model, Claude Opus 4 was quite different. It was half the length and seemed to focus much more on skill shifts and new human-AI hybrid roles. Which one do you prefer? Let me know in the comments, please.
Results
Claude got to work and work it did, consulting 709 (!) sources over almost 20 minutes. At 8 pages, the final report was brief by deep research standards but it hit the most important points. In fact, the conciseness could be an asset in some cases.
Here are some highlights:
There is some hope that AI will create more jobs (170 million) than it will displace (92 million) by 2030, but this transformation will require tremendous adaptation. Up to 375 million workers may need to change occupations.
75% of knowledge workers are already using AI tools, but only 6% of organizations are engaged in serious upskilling.
Information technology/systems professions face the biggest changes due to AI, but other roles may change significantly.
The future may be in hybrid human-AI roles that focus on responsible AI deployment, AI system interaction optimization, and human-AI workflow design.
Many professions are already seeing noteworthy productivity gains from AI.
There’s a lot more to the report, so I encourage you to read it.
Sidebar: AI Assignment Repository Survey
If you haven’t already done so, I would appreciate it if you would complete this short survey gauging interest in a free, open-source repository of learning activities that discourage inappropriate AI use and/or leverage AI to enhance student learning.
Digging in
Out of curiosity, I asked Claude to describe the process it followed to draw the conclusions in its report. Keeping in mind that AI is often pretty terrible at self-awareness, I found the results interesting. Here are the main takeaways from Claude’s response.
Claude planned before executing any actions. One interesting aspect of the planning is that Claude recognized that this is a rapidly evolving area and its knowledge cutoff is January 2025. So, it determined that it would need to engage in extensive web searching.
Claude conducted 15-20 distinct web searches, “deliberately varying my search terms to capture different perspectives and data sources. These included broad foundational, sector-specific, data-focused, and policy and response searches. It also specifically prioritized more authoritative sources such as major consulting firms and research institutions. Peer-reviewed sources were NOT prioritized, which was interesting.
For each major claim in the report, Claude tried to cross-validate across different sources. Claude gave specific examples, like cross-referencing job displacement and creation claims.
Then Claude organized the report along analytical dimensions: temporal analysis, sector and geographic breakdowns, skills transformations, and economic limitations.
Claude also performed a critical assessment that identified limitations, established implicit confidence levels based on evidence quality, and planned out a narrative structure.
Claude included numerous quality control checks such as source diversity, claim verification, and logical consistency.
Overall, the process seemed comprehensive and reasonable. In fact, if a graduate assistant laid out this approach to me, I would be more than a little impressed.
Bottom Line
The bottom line for me is that AI research tools keep improving. Competition will lead to further development, including the emergence of more specialized tools. All that being said, for academic researchers, deep research tools are still useful primarily for quickly coming up to speed on a topic. AI deep research is still no substitute for expert human deep research. That being said, I’m still a fan and will continue to use ChatGPT, Gemini and now Claude’s deep research tools. (The jury’s still out on Perplexity’s version.)
Want to continue this conversation? I'd love to hear your thoughts on how you're using AI to develop critical thinking skills in your courses. Drop me a line at Craig@AIGoesToCollege.com. Be sure to check out the AI Goes to College podcast, which I co-host with Dr. Robert E. Crossler. It's available at https://www.aigoestocollege.com/follow.
Looking for practical guidance on AI in higher education? I offer engaging workshops and talks—both remotely and in person—on using AI to enhance learning while preserving academic integrity. Email me to discuss bringing these insights to your institution, or feel free to share my contact information with your professional development team.