1. Home
  2. Job Skills with AI
  3. Professors
Job Skills with AI · Education

AI for Professors

Use AI to take the drafting, summarizing, and course-building hours off your week without putting student records, unpublished research, or a fabricated citation into anything with your name on it.

Reviewed September 2026. Free to use. No account needed.

Tasks covered6 workflows
Ready prompts5 to copy
Skills to build5 skills
Cautions5 role-specific
Plan5 steps, 30 days

The realistic gain is in the work that is not thinking: the syllabus revision, the fourth version of a problem set, the committee report, the response-to-reviewers letter that has to stay polite. A chat assistant drafts those in minutes from your notes. It also reads a folder of PDFs and answers questions about them, which changes how you prep a lecture or scope a literature review.

It is also the most confident liar you will ever work with. It fabricates citations with plausible authors, real journals, and working-looking DOIs. It misstates a result you know cold. It praises an argument the student did not make. Every fact, reference, and number gets verified against the source before it leaves your hands.

Then there are the rules. FERPA covers your students' grades and records whatever their age. Your institution has an academic integrity policy and probably an AI policy; journals and funders have their own, including disclosure requirements and limits on uploading material under review. Read them before you paste.

Quick wins this week

  • Paste your syllabus and ask for an assignment-level AI-use statement with three tiers (prohibited, allowed with disclosure, expected), then adapt it to your discipline.
  • Upload next week's readings into NotebookLM and ask where students most often get confused, plus discussion questions that force them back into the text.
  • Paste your own abstract and ask for the five hardest questions a hostile reviewer would ask, then answer them before you submit.
  • Give the AI a dense accreditation template and ask for a plain-language summary and a list of the evidence you still need to gather.

What AI can do for professors, task by task

Course and syllabus design

Paste your learning outcomes, course length, the prerequisite knowledge you can assume, and constraints like class size and modality, and ask for a week-by-week plan with readings, activities, and assessments aligned to each outcome. Confirm every suggested reading exists and says what the model claims. Put the AI-use rule on each assignment sheet as well as in the syllabus.

'Redesign my 14-week intro statistics course for 120 students and one TA so at least half the assessment happens in class or through staged drafts.'

Assessment design that holds up when students have AI

Describe the outcome you are assessing and ask for formats that make AI use irrelevant or visible: in-class writing, oral checks, staged drafts with version history, problems built on the student's own data, reflections tied to a specific class discussion. Ask which of your current assignments a chat assistant could finish in ten minutes, and redesign those first.

Feedback on student work

Paste an anonymized draft with your rubric and ask for two strengths, one priority revision, and marginal questions in your voice, with no grade. Read every comment; models praise arguments the student did not make and miss the real flaw. Check whether your institution allows student work to be uploaded at all, and use only approved tools if it does.

Literature review and staying current

Find candidates with a search tool, then load the actual PDFs into NotebookLM and ask what each measured, how, and where they disagree. Scholarly search tools such as Elicit, Consensus, and Semantic Scholar help you find papers; none of them, and no chat assistant, guarantees a reference is real. Verify author, year, journal, and DOI in the publisher's database before you cite.

Grant proposals and manuscript editing

Paste your specific aims and the funder's review criteria and ask for a critique against those criteria and the questions a reviewer would raise. For manuscripts, ask for tighter prose and one claim per paragraph, then disclose AI assistance as the journal requires. Check the funder's and your institution's AI rules first, keep unpublished data out of consumer tools, and never upload a manuscript or proposal you are reviewing for someone else.

Letters, committee work, and program reports

For recommendation letters, write your own bulleted specifics and ask for a draft that turns them into paragraphs; never paste transcripts or application materials into a consumer tool. For program reviews and accreditation self-studies, give it the template and your evidence and ask for a draft plus a list of missing evidence.

Prompts for professors

Replace the bracketed placeholders, paste into any chat assistant, and iterate on the result.

AI-use policy for a course

Act as an experienced professor and academic integrity expert who knows that AI-writing detectors are unreliable. Write an AI-use section for the syllabus of [COURSE NAME AND LEVEL] in [DISCIPLINE]. My stance by assignment type: [PROHIBITED, ALLOWED WITH DISCLOSURE, OR EXPECTED, FOR EACH ASSIGNMENT TYPE].

Include a plain rule for each major assignment type, how students disclose AI use (a short note plus the prompts they used), why the rule exists in terms of what they need to learn, what happens if it is broken (a conversation first, then the institutional process), and one sentence stating that detector scores alone are not evidence. Under 400 words, direct, not preachy. Then list three assignments where my stance may be unrealistic.

Tip: Put the rule on each assignment sheet too. A paragraph in the syllabus is not something students reread in week nine.

Anonymized feedback in your voice

You are my teaching assistant with a doctorate in [DISCIPLINE]. Here is my rubric: [PASTE RUBRIC]. Here is an anonymized student draft: [PASTE DRAFT WITH ALL IDENTIFYING DETAILS REMOVED]. Here is feedback I have written before so you can match my voice: [PASTE ONE EXAMPLE].

Give two specific strengths quoting the student's words, one priority revision explained in two sentences, and three marginal questions that push the argument, each tied to a specific sentence. Do not assign a grade. Do not praise anything that is not actually in the draft. If the draft misreads a source, name the misreading.

Tip: Keep 'do not praise anything that is not in the draft' in the prompt. Empty praise is the default failure.

Reviewer simulation for a proposal

Act as a senior reviewer on a [FUNDER OR PROGRAM] panel scoring against these criteria: [PASTE REVIEW CRITERIA]. Here is my draft of the specific aims and approach, with preliminary data described in general terms only: [PASTE DRAFT].

Score each criterion on the funder's scale and explain each score in three sentences. Then list the ten questions you would raise in panel discussion, in order of how much they would hurt me. Identify every claim that needs a citation and every gap in the logic from aim to method to outcome. Do not rewrite the proposal.

Tip: Check the funder's current AI-use notice before you paste anything; some programs require disclosure and restrict what applicants and reviewers may do.

Citation audit of your own draft

You are a meticulous research librarian. Below is a draft with its reference list. For each in-text citation, tell me whether the reference list entry supports the specific claim as I have stated it, whether the claim overstates the finding, and whether the entry is complete (authors, year, title, journal, volume, pages, DOI). Mark every reference 'must verify in publisher database' because you cannot confirm a reference exists. Then list the claims that have no citation and probably need one.

Draft: [PASTE DRAFT]
References: [PASTE REFERENCE LIST]

Tip: The model cannot confirm a reference is real, no matter how confident it sounds. Open every one in the publisher's database.

Recommendation letter from your own notes

Draft a letter of recommendation for a student applying to [PROGRAM OR POSITION TYPE]. I am a [YOUR TITLE] in [DEPARTMENT] and I know the student from [CONTEXT AND DURATION]. Here are my own specific observations, with no grades or transcript details: [YOUR BULLETED NOTES]. Use [STUDENT'S CHOSEN NAME AND PRONOUNS].

One page, concrete, no generic praise, one memorable specific example in the opening paragraph, a plain statement of how the student compares to others I have taught, and a closing that states my recommendation directly. Flag any sentence that sounds like a template so I can rewrite it.

Tip: Your notes are the letter. If the model has to invent the specifics, do not send it.

Want a prompt for something else? Use the Prompt Builder or browse education and training prompts.

Skills to build

Verifying citations before anything leaves your desk

Why: Fabricated references have already turned up in published papers and court filings. In your field they end careers. The model cannot tell a real reference from a plausible one.

How: Open every reference in the publisher database or your library's discovery tool and confirm the paper says what you cite it for. Never paste a generated reference list into a manuscript.

Anonymizing student work and records by reflex

Why: Grades, IDs, accommodation letters, and identifiable writing are education records under FERPA, and your institution may bar uploading student work entirely.

How: Strip names, IDs, and identifying details before any paste. Check your institution's policy on student work and use only approved tools for it.

Designing assessment for a world with AI in it

Why: Any take-home essay or routine problem set can be produced by a chat assistant in minutes, and detectors will not tell you which ones were. Assessment design is the only durable answer.

How: Audit each assignment for how well AI could do it. Move weight toward in-class work, oral checks, staged drafts, and tasks tied to your specific class discussion or data.

Knowing the disclosure rules that apply to you

Why: Journals, funders, and your institution each have AI-use policies, and they differ: disclosure requirements, bans on AI authorship, restrictions on uploading material under review.

How: Keep a one-page note per venue with its current AI policy and the date you checked. Re-check before each submission; these policies change often.

Editing generated prose into your own voice

Why: Reviewers, students, and colleagues can tell when a paragraph is generic. Your credibility rests on sounding like someone who has thought hard about the problem.

How: Ask for a draft, then rewrite the first and last paragraph yourself. Cut hedges, generic transitions, and any sentence you could not defend in a seminar.

Tools worth knowing

Cautions for professors

FERPA applies to college students too

Grades, student IDs, disability accommodation letters, advising notes, and identifiable student writing are education records under FERPA regardless of the student's age. Never paste confidential student data into a consumer AI tool unless your institution has approved that tool for that use. Check whether your institution allows student work to be uploaded at all.

Academic integrity policy and AI detectors

AI-writing detectors are not reliable enough to support an integrity charge on their own, and they flag multilingual students and formal writers at higher rates. Set the AI rule at the assignment level, rely on process evidence and a conversation with the student, and follow your institution's process rather than a detector score.

Research integrity: fabricated citations and misstated findings

Models generate references that do not exist, attribute real papers to the wrong authors, and misstate results. Verify every citation in the publisher's database and confirm the paper supports the specific claim. Treat a generated literature summary as a lead list, never as a source.

Journal, funder, and institutional AI policies

Most publishers now require disclosure of AI use in manuscripts and prohibit listing an AI as an author. Funders including NIH and NSF have published notices on AI use in applications and peer review, and uploading a proposal or manuscript you are reviewing to an AI tool can breach confidentiality. Read the current policy for each venue before you submit or review.

Unpublished data, IRB-protected information, and your own IP

Preliminary results, participant data, and unpublished manuscripts may be covered by IRB protocols, data use agreements, and confidentiality terms, and a consumer tool's terms may allow training on what you paste. Use an institutional or enterprise tool with no-training terms for anything you would not post publicly.

Your 30-day plan

  1. Week 1: Read your institution's AI policy and the current AI notices from your main journal and funder. Write the AI-use statement for one course. Use AI for one low-stakes task a day: an email, a discussion question set, a meeting summary.
  2. Week 2: Load next week's readings into NotebookLM and prep one lecture from it. Audit your assignments for how easily a chat assistant could complete each, and redesign the one that worries you most.
  3. Week 3: Run the feedback prompt on three anonymized drafts in an approved tool and compare with your own comments. Run the citation audit on one of your working drafts and see what it catches.
  4. Week 4: Put one proposal or manuscript section through the reviewer simulation and answer the questions it raises. Draft one committee report from your evidence and time it.
  5. End of month: Decide which three tasks you will keep using AI for, write a one-page note on the AI errors you caught, and share your assignment-level AI policy with your department.

Frequently asked questions

Can professors use ChatGPT for research?
For brainstorming, editing your own prose, summarizing papers you have loaded, and critiquing your arguments, yes, subject to your institution's policy. Not as a source: every citation and fact it produces must be verified in the primary literature, and unpublished or participant data stays out of consumer tools.
Is it a FERPA violation to put student essays into AI?
It can be, if the work is identifiable and the tool is not covered by an institutional agreement. Anonymize first, use only approved tools, and check whether your institution allows uploading student work at all. Many treat student writing as an education record.
Do AI detectors work for catching students using ChatGPT?
Not reliably. They produce false positives, cannot be audited, and flag multilingual students at higher rates. Use assignment-level rules, process evidence, and a conversation with the student, then follow your institution's integrity process.
Do I have to disclose AI use in a journal article?
Usually yes. Most major publishers require disclosure of generative AI use in the methods or acknowledgments and prohibit listing AI as an author. Rules differ by publisher and change often, so check the journal's current author guidelines before submitting.
What is the best AI tool for professors?
NotebookLM is the best fit for questioning your own readings and papers. Claude and ChatGPT are strong for drafting, feedback, and critique; Perplexity for finding candidate sources with links you then verify. If your institution provides Copilot or Gemini under an enterprise agreement, use that for anything involving student or research data.

Terms used on this page

Related roles