What is Scholarship in the Age of Deep Research AI?
What does it mean to be a scholar when deep research reasoning AI models can produce PhD-level research in a matter of minutes?
There is a famous TV commercial from 1976 for Xerox photocopiers where a medieval monk is seen painstakingly composing a folio of documents with the tip of a feather and a saucer of ink. He proudly presents his work to the abbot who then asks him to produce another 500 sets. Brother Dominic despondently turns away to return to his scriptorium. Then, in a suspension of disbelief, the scene cuts to the interior of a contemporary business services department in an urban office building where Brother Dominic hands over his manuscript to a man who feeds it into the new Xerox 9200 copier. It has "...a computerized programmer that coordinates the entire system. Can duplicate, reduce, and assemble a virtually limitless number of complete sets, and does it all at the incredible rate of two pages per second." "It's a miracle," says the abbot, as he is handed a stack of copies produced in a matter of minutes.
The metaphor used by famed Creative Director Steven Penchina who produced this classic resonates across time because we can all identify with Brother Dominic's feelings both as the subject of being asked to produce a mountain of intensive time-consuming work as well as his revelation of using a tool that alleviates it.
But let's take this metaphor a step further. How does the intervention of the Xerox 9200 fundamentally change the character of the monastery system and the value of those who operate in it?
From the monastery’s position as an enterprise, the power to instantly produce hundreds of copies of its work expands its social and political influence.
From the abbot's position, he can conceive an entirely new strategic plan that reallocates organizational resources from copying singular works to producing a range of different literature.
From Brother Dominic's position, he can now compose multiple projects simultaneously rather than focusing on one.
As a higher education educator, I am struck by the similarity of this scenario to the historically typical college student who labors much like Brother Dominic through four years of study culminating in a 50-page Capstone paper being handed over, weary eyed, to their professor. As an institution, we have sanctified this narrative as the cost of earning a degree much like the opening of the old TV show Fame where the Director played by Debbie Allen says to her new enrollees, "You've got big dreams. You want fame. Well fame costs, and right here is where you start paying, in sweat." Scholarship, in its scarcity, costs, and has been achieved here, where the knowledge and the network happen to be.
Well, things have changed.
With proper operator skills, OpenAI, Google, and Perplexity’s similarly named Deep Research AI models are said to be capable of generating first-year PhD-level scholarly works in a matter of minutes. Ethan Mollick (2025), author of Co-Intelligence - Living and Working with AI and Distinguished Faculty at the Wharton School reports, "For the first time, an AI isn't just summarizing research, it's actively engaging with it at a level that actually approaches human scholarly work" (p. 9). College educator Craig Van Slyke states (2025, p. 1),
"As the name implies, Deep Research actually does research before preparing its response. It also thinks through its research before starting it and usually asks clarifying questions before getting to the real work. It's a bit eerie—it almost seems like you're texting with a human research assistant. The results of my early testing are beyond impressive.
"... I can say that with minimal effort on my part, Deep Research produced a 34-page report that was well-researched and comprehensive. The report will save me hours, possibly days of work on a major project. Of course, I won't pass off Deep Research's work as my own and I'll double-check the content. Fortunately, double-checking is easy since Deep Research cites its sources.
"... Although scholarship certainly isn't dying, the way we ‘do’ scholarship is about to change in a major way."
Certainly, deep research models are imperfect and are tuned to some domains more so than others. Leon Furze (2025), PhD candidate on the implications of AI in education, compared his hand-crafted PhD work to the output of OpenAI's Deep Research model and found it lacking (p. 14).
"It is designed to produce the appearance of research, without any actual research happening along the way. Will it impact the education sector if and when it becomes more accessible? Absolutely. Is it any good? Sure. Is it research? I’m not convinced."
Critique of deep research AI is well-founded. The current models are limited to openly accessible resources, and they cannot carry out a clinical trial. They are strong in summary but not necessarily in analysis. Andrew Maynard (2025), Professor of Advanced Technology Transitions at Arizona State University remarks, "...Deep Research has its flaws, and is still a few steps away from where I’d like an AI research agent to be. But I suspect that reasoning research agents like this will eventually make non-AI-augmented scholarship and research look intellectually limited and somewhat quaint" (p. 4).
It is in our best interest to assume that the jagged frontier of AI will improve continuously as long as there are markets willing to pay for it. And takeaways from preliminary experiments conclude that, if anything else, it is strong in the areas most valuable to academic work: tenaciously locating resources, encompassing vast domains of material, exploring findings, finding connections and conflicts, synthesis, summary, citation, and fluency (Furze, L., 2025; Jones, N. 2025; Maynard, A. 2025; Mollick, E., 2025). Add the developing affordances of agentic frameworks, open-source models, and local implementations, and we will soon see deep research AI become a powerful component of more complex, accessible, and private AI systems.
Like the monastery abbot in our opening scenario, institutions are now faced with reimagining the potential of our enterprise. And Brother Dominic, if he were a college student, now has much greater potential as a scholar than ever before. So, several questions emerge:
What is the nature of scholarship when PhD-Level research can be generated in a matter of minutes?
Should higher education evolve from a model of scholarship to a model of meta-scholarship?
What are the competencies that matter in a new profile of academic citizenship?
Before considering these questions, we need to examine what it is that we do: What is our product? What do we actually produce? This is a discussion question often asked by institutional advancement consultant Andrew Shaindlin to his college administrator clients. The exercise is perhaps more of a thought experiment than a self-study question requiring a statement of fact. Yet, the outcome of the academic experience - whatever we or students think it is - conveys a form of legitimacy to stakeholders that view this scholarly product as a desirable asset to their organizations.
Do the questions we ask here, then, reconfigure the balance of this equation? If we are in the business of producing scholars, then deep research AI tools appear to devalue the currency upon which we leverage our purpose in society. Traditionally, a university graduate in a hiring position will view their college graduate job candidates through the same form of legitimation as themselves, given the ordeal required to successfully complete a degree program. Does this sense of comradery then change when the actual output of scholarly work no longer requires years of grinding (expensive) academia? It depends on how we define scholarly work.
In Bowen and Watson's work, Teaching with AI (2024), a significant amount of content is devoted to strategies and techniques for querying, refining, evaluating, and synthesizing the output of AI systems, both as an instructor and as a student. Readers will likely skim through the immensity of these new competencies and wonder how it would be possible to be proficient enough to attain a sense of mastery over AI rather than feeling like a vassal of it. They are in good company. Oral scholars in ancient Greece bemoaned the emerging reliance on writing and libraries; religious scholars seethed at the emergence of secular universities emphasizing dialectical reasoning and logical argument; men of high social status were outraged at empirically scientific challenges to their claims of causal phenomena; provosts and professors denied that students could achieve equivalent outcomes from fully online courses. Yet, through these transformations, scholarship did not die. The basis of its legitimation simply evolved.
So, if it is possible for a typical college student today to generate scholarly work through sophisticated machinery, shouldn't the focus of scholarship change, in part, to mastery of this machinery? The historical patterns of scholarly evolution cited previously suggest that we must. What, then, is the nature of scholarship in the Age of Deep Research AI?
In one sense, it is no different than it has ever been. Educators on the leading edge of using AI agree that students need to position it as a collaborator and treat its output as a thought partner rather than merely the dispenser of fluent information (Bauschard, S, 2024; Bauschard, S. & Quidwai, S., 2024; Eaton, L. 2024; Perkins et al., 2024). Evaluating the output of AI is nothing new as a competency, as are any others one might apply against AI output: critical thinking, creative interpretation, ethical reasoning, collaboration, and perhaps civic engagement. In practice, the subject matter in our degree programs functions mostly as a vehicle through which these competencies are practiced, which is to say that a graduate in any given degree program without proficiency in scholarly competencies does not provide much value to stakeholders. Essentially, all subject matter experts being equal, the scholarly competencies are our product.
The difference, then, between traditional scholarly work and the same competencies under the Age of Deep Research AI appears to be mostly a matter of scale. If it is conceivable today (and for the foreseeable future) that an astute AI operator can generate dozens of academically competent reports in the same amount of time than it would have taken traditional academic work to produce one, what is the scholarship we are looking for under these conditions? Now that Brother Dominic has the means to generate an infinite amount of scholarly work, what value does he bring to the community of inquiry?
The metaphor that comes to mind to describe this kind of scholar is a symphony orchestra conductor. The conductor is a second-order scholar of first-order work characterized by the following competencies:
Advanced musical knowledge to comprehend and analyze a symphonic score.
The ability to interpret musical notation according to a sense of aesthetics and historical context.
Advanced knowledge of symphonic instrumentation and acoustics.
Leadership and authoritative decision making.
Effective communication to inspire performers in the spirit of collaboration.
Creative expression as an emergent property of practice and rehearsal.
The orchestra conductor is, in effect, a scholar of scholarship capable of producing unique interpretations and meaning from the output of other scholarly work, shaping the output of human performers and machinery through dialog, providing value to otherwise inert code through iterative study, testing, feedback, and rehearsal, and then presenting it authoritatively to a discriminating audience.
How does this framework inform the design of our product, in the spirit of Shaindlin's thought experiment? A traditionalist view of this framework would argue that the human experience of demonstrating unassisted intellectual persistence punctuates the mastery of subject matter and proficiency in the scholarly competencies as a signifier of how these accomplishments are to be taken as. I do not believe we can (nor should) refute this, no matter the projection of AI, ongoing. Legitimation of one's membership among scholars can be conferred as a combination of both social codes (grit) and objective codes, such as certifications, awards, degrees, etc. (Maton et al., 2015). One does not need to march uphill through a foot of snow to get to school every morning to signify one's grit, but it doesn't hurt.
Rather, I propose that the argument towards a new framework of scholarship is not a conflict between the legitimacy of tradition against the ability to extract value from the Leviathan of AI models. One cannot operate as a second-order meta-scholar without comprehension and mastery of first-order research methods. Instead, the exemplary AI-augmented meta-scholar would be proficient in all of the traditional competencies plus the following:
Responding to instructional challenges by crafting language that produces optimal output from deep research AI.
Refining the output of deep research AI to improve the focus, orientation, narrative, citation, visual representations, and findings according to the intended use case.
Validating the output of AI systems according to scholarly standards for reliability and quality.
Organizing collections of AI-generated research into AI-optimized systems for the purpose of extracting trends, patterns, alignments, conflicts, and other useful information according to larger research needs.
Accounting for multiple perspectives of research findings according to a variety of roles or personae staged and implemented within AI systems.
Developing communication to convey the meaning and value of meta-research to various audiences.
Documenting the degree to which AI was involved in the research process and the final artifact.
As innovative as these competencies might be, however, we have not yet answered the question as to why we should adopt this model. It would be easy enough to just say that our stakeholders will demand it, though they too are mired in a scramble to transform their organizations as much as we are.
Instead, the motive for this proposition is centered on creating a model that aligns with what ought to be embodied in our product: How does a scholar go about the task of locating, eliciting, organizing, synthesizing, conveying, and attributing knowledge in an age where such work has been streamlined to the point of zero resistance? Potkalitsky (2025) might tag the approach presented here as a Re-negotiation - a redefinition of scholarly skill entirely - or Synergistic - a "... new [scholarly] literacy [characterized by] fluency in orchestrating human-machine creativity" (p. 3). Yet, it is important to remember that even if a medieval monastery had a Xerox 9200, it would not stop them from producing the same thing they had always produced in the past. It would simply be a matter of managing the empowerment of scale. A call to adopt a new model of scholarship is not a reproach, relegation, or dismissal of what we do. It is a call for a second-order transformation of it.
Let's admit that it is easy for us, in our present position in time, to predict what might happen if we were to drop a Xerox 9200 photocopier into a medieval monastery. We can easily see that it would transform more than just the organization and its people - it would affect the entire environment in which it was relevant. Instead, the more compelling observations would be: How long do you suppose it would it take for the abbot to move past the obvious implications of this tool and realize the breadth of opportunity that stood before him, and the extent to which the stature of his organization would be elevated? How long would it take for him to develop a new organizational philosophy and mission? How long would it take for the abbot to help Brother Dominic realize what he is now capable of orchestrating for his audiences, and to what degree?
Higher education, in its most desperately disrupted and socially diminished state, must recognize this moment, the Age of Deep Research AI, as its finest hour.
References
Bauschard, S. (2024, August 25). How to Help Students Learn to Think Using AI. Education Disrupted: Teaching and Learning in An AI World
Bauschard, Stefan and Quidwai, Sabba, Humanity Amplified: The Fusion of Deep Learning and Human Insight to Shape the Future of Innovation (November 2, 2023). Retrieved from SSRN: https://ssrn.com/abstract=4621210 or http://dx.doi.org/10.2139/ssrn.4621210
Bowen, J. A., & Watson, C. E. (2024). Teaching with AI: A practical guide to a new era of human learning. JHU Press.
Eaton, L. (2024, August 6). AI Syllabi Policies - A Look at the Collection. AI + Education = Simplified. https://aiedusimplified.substack.com/p/ai-syllabi-policies-a-look-at-the
Furze, L. (2025, February 15). Hands on with Deep Research. Leon Furze. https://leonfurze.com/2025/02/15/hands-on-with-deep-research/
Jones, N. (2025, February 6). OpenAI’s ‘deep research’ tool: is it useful for scientists? Nature. https://doi-org.unh.idm.oclc.org/10.1038/d41586-025-00377-9
Maton, K., Hood, S., Shay, S. (Eds.) (2015). Knowledge-building: educational studies in legitimation code theory. Routledge.
Maynard, A. (2025, February 4). Does OpenAI's Deep Research signal the end of human-only scholarship? The Future of Being Human. https://futureofbeinghuman.com/p/openai-deep-research-ai-scholarship
Mollick, E. (2025, February 3). The End of Search, The Beginning of Research. One Useful Thing. https://www.oneusefulthing.org/p/the-end-of-search-the-beginning-of
Perkins, M., Furze, L., Roe, J., MacVaugh, J. (2024). The Artificial Intelligence Assessment Scale (AIAS): A Framework for Ethical Integration of Generative AI in Educational Assessment. Journal of University Teaching and Learning Practice, 21(6). https://doi.org/10.53761/q3azde36
Potkalitsky, N. (2025, February 27). The Five Faces of Education in the Age of AI: A Spectrum of Survival, Skepticism, and Symbiosis. Educating AI. https://nickpotkalitsky.substack.com/p/the-five-faces-of-education-in-the
Van Slyke, Craig (2025, February 26). Breaking News: ChatGPT Deep Research Available to the Masses. AI Goes to College. https://aigoestocollege.substack.com/p/breaking-news-chatgpt-deep-research