The University of Texas at Austin released a campus-wide generative-AI citation rubric this week that tells students how to document ChatGPT, Copilot, and similar tools on syllabi assignments starting the fall 2026 term, according to provost office guidance instructors received Monday and reviewed by InfoHandle. The policy does not ban AI; it standardizes disclosure so graders can distinguish original analysis from machine paraphrase and block fabricated references before they reach peer review.
What students must label
Any paragraph drafted or reorganized with generative assistance needs an in-text note naming the tool, model version if known, and date of use. Prompt text goes in an appendix for research papers longer than ten pages. Bibliographies must separate “AI-generated summaries” from scholarly sources retrieved through UT Libraries databases—students cannot cite a chatbot as if it were a journal article.
STEM lab reports get a lighter template: methods sections may mention code autocompletion if the student verifies outputs, but raw data tables cannot be transcribed from model hallucinations. Engineering dean’s staff said violators face existing plagiarism sanctions, not a parallel AI court.
Faculty implementation
Departments may tighten rules. The McCombs School of Business opted to prohibit AI on take-home exams entirely, while some rhetoric courses allow brainstorming with disclosure. Teaching assistants attended two-hour workshops on spotting uniform sentence rhythm and impossible citations—signals, not proof, provost materials stress.
Jason Kim’s desk watches these rubrics because they reveal how public universities scale policy faster than legislature drafts: UT’s text mirrors MLA and APA working groups but adds Texas-specific examples on government document retrieval.
Accessibility and equity
Disability services asked whether AI grammar tools count as assistive technology; the rubric says approved accommodations supersede disclosure when documented. Students without paid subscriptions may use campus-licensed tools that log usage for audit; personal Pro accounts remain allowed but must still be disclosed.
What the rubric does not solve
It cannot detect undisclosed use with certainty; honor codes still matter. Faculty are urged to assign in-class writing samples and oral defenses where integrity stakes are highest. UT Libraries expanded workshops on primary-source hunting so students rely less on models for historiography they never read.
Other Texas public campuses are expected to adapt the template under coordinating board encouragement. For Longhorns, the fall takeaway is paperwork discipline: treat AI like a calculator with a paper trail—transparent, dated, and never a substitute for clicking into the actual PDF you claim to summarize.
UT’s honor council published anonymized case studies from spring pilots where students cited nonexistent law review articles; graders caught them with library link checks, not AI detectors alone. Graduate programs may require research-data management plans to mention AI transcription tools for interviews, another disclosure line PhD students should watch before dissertation proposals land in October committees.
