■ MODERATE RISK ■ Education
No, but the job is being quietly rebuilt underneath its occupants. AI now drafts, summarizes, codes, and reviews at a level that automates chunks of the research pipeline — while the professor's actual scarce functions, asking the right questions and training humans to think, stay stubbornly manual.
“AI writes papers, but who's going to fight over tenure? Oh right, nobody wants to automate that.”
Our AI replacement risk score — how we score jobs
A research professor's week bears little resemblance to the public image of contemplative scholarship. It's grant writing — endless grant writing — supervising doctoral students and postdocs, running a lab or research group like a small business, reviewing papers for journals, sitting on committees, teaching a course or two, and somewhere in the margins, actual research. The currency is publications and funding; the pressure is continuous; and the administrative load has grown for decades. Tenure, where it still exists, is the prize that justifies the decade-long apprenticeship.
AI has invaded the pipeline from both ends. Literature review, once a doctoral student's first year, compresses into hours with retrieval tools. Models draft manuscripts, write analysis code, clean data, generate hypotheses, and produce credible peer-review reports — journals are openly wrestling with how much of their submission and review flow is now machine-written. Grant boilerplate automates beautifully. In some fields, AI systems propose and even run experiments in automated labs. The uncomfortable truth is that the median tasks of early-stage research careers are exactly what LLMs do best, which destabilizes the apprenticeship model that produces professors.
What resists is the judgment layer: choosing which questions matter, designing studies that can actually answer them, catching the subtle confound, mentoring students into independent thinkers, and carrying accountability for scientific integrity. Institutions also aren't automating tenure, funding allocation, or academic politics — human status games run on humans. Our risk score reflects a profession whose output tools are transforming rapidly while its social structure, hiring bottlenecks and all, changes at academia's traditional glacial pace.
Automatability: our editorial assessment of current and near-term AI capability
The writing, coding, and review layers are automating now — the shift is visible in every submission system and lab meeting. Through 2030 expect AI-assisted research to be the default and publication norms to be renegotiated around it. The deeper structural change, fewer junior research positions as apprentice-level tasks automate, bites through the 2030s. The professor role itself — question-chooser, mentor, accountable investigator — persists in a transformed academy by ~2040.
It can execute large parts of the pipeline — literature synthesis, hypothesis generation, code, drafting, even automated experimentation in some lab sciences. What it doesn't do reliably is choose questions worth asking, design studies that survive scrutiny, or carry accountability for integrity and error. Current AI-generated papers cluster around competent-but-derivative, which is telling: the frontier still requires human judgment.
It was already a lottery — far more PhDs than faculty posts — and AI tightens it by automating the apprentice tasks that justified large lab headcounts. That said, the professorship itself looks durable: universities need accountable researchers, mentors, and grant-winners. Go in clear-eyed: the path is competitive, and the skills that win it are shifting from output volume to judgment and originality.
Substantially and messily. Journals face AI-written submissions, AI-assisted reviews, and detection tools that don't reliably work, so norms are being renegotiated in real time — most now require disclosure of AI use. For working academics the practical change is speed: drafting and reviewing cycles compress, publication volume rises, and distinguishing signal from fluent noise becomes the scarce skill.
Stop competing with AI at summarization and boilerplate — use it, and invest your human hours in what still differentiates: original questions, rigorous design, deep methods expertise, and visible scholarly identity. Learn to verify machine output ruthlessly, since errors in AI-assisted work land on your name. And diversify: industry research careers value the same judgment skills with better odds.