The topic brief for this post cited a Tulane University AI cheating retraction, but the underlying incident most frequently reported in that pattern involves a UC Santa Barbara professor who publicly withdrew an accusation after a student produced Google Docs version history and handwritten notes. Because no verifiable Tulane-specific case matches the described facts, this post covers the documented UCSB-style pattern and what Tulane's own Code of Academic Conduct requires if a similar accusation lands at Tulane.
The pattern: detector flag, process evidence, public retraction
The sequence in these cases is consistent. A professor runs a paper through Turnitin, GPTZero, or a similar tool. The score comes back elevated. The professor sends an accusation. The student produces Google Docs revision history, handwritten planning notes, and browser history from the drafting window. The professor retracts.
The documented UCSB case followed this exact path. The retraction was notable because it was public, in writing, and acknowledged that the detector score had been treated as proof when it was only a probability estimate. Our coverage of the UCSB retraction walks through what the student submitted and why it worked.
Why Google Docs version history and handwritten notes are decisive
A detector produces a probability score. Process evidence shows the writing happened. The two kinds of evidence are not comparable, and faculty reviewers tend to recognize this once the process evidence is in front of them.
Google Docs version history records every substantive edit with a timestamp. A paper that was pasted in from an AI tool shows one giant insertion. A paper that was written by hand shows hundreds of small revisions, cursor movements, typos, rewrites, and reorderings over hours or days. Handwritten notes add a layer no AI tool can fabricate retroactively: paper that was photographed before the paper was submitted, with ideas that match the final draft.
Why the detector flagged the paper in the first place
AI detectors measure statistical features: sentence-length variance (burstiness) and word-choice predictability (perplexity). Careful academic writing, especially writing that follows a clear thesis structure, tends to score low on both. The 2023 Weber-Wulff study, published in the International Journal of Educational Integrity, tested fourteen detection tools and concluded none met a reliability threshold for institutional decision-making.
A separate 2023 Stanford paper (Liang et al., published in Patterns) found that detectors flag writing from non-native English speakers at sharply higher rates than writing from native speakers. Clear, structured prose looks like AI prose to a probability model, regardless of who wrote it.
What Tulane's academic conduct process requires
Tulane's Code of Academic Conduct handles alleged violations through the Office of Student Conduct and, for formal findings, the Honor Board. The code requires that a finding of responsibility rest on evidence sufficient to establish the violation, not on a single automated signal. A Turnitin AI percentage, by itself, does not meet that standard, and Turnitin's own documentation tells institutions the AI indicator is not standalone proof.
Before responding, ask in writing:
- Which detector was used, and what score was reported
- Whether any human review of the flagged passages occurred before the accusation
- What the specific allegation is: unauthorized AI use, misrepresentation of authorship, or something else
- What the standard of evidence is under Tulane's Code and what procedural steps follow
Evidence that closes these cases
The artifacts that have led to retractions in the documented pattern:
- Google Docs or Word version history showing incremental drafting across the writing window
- Photographs of handwritten planning notes, outlines, or source annotations, with EXIF timestamps
- Browser history showing library database searches and source retrieval during the drafting period
- Citation manager entries (Zotero, EndNote, Mendeley) with capture timestamps that predate submission
- Messages to classmates or writing center appointments that reference the paper in progress
Our step-by-step guide on gathering evidence after an AI accusation covers the first 48 hours in detail, including retention windows that close faster than most students expect.
If this is you at Tulane or a similar institution
Do not confess to settle the matter quickly. An informal admission can become a formal finding on your record, with consequences that extend beyond the course. Preserve your evidence first. Request the detector name and score in writing. Then write a response that pairs the process evidence with the research on detector unreliability and the specific standard your institution's code sets.
If you are preparing that response, NotBot generates a personalized defense package that names the detector, cites the applicable research, and organizes your version history and handwritten notes into a response letter tailored to your institution's conduct process. If a finding has already been issued, the appeal package covers the grounds that matter at the appeal stage.
If the proposed sanction is suspension or expulsion, or if you are an international student whose visa status depends on enrollment, consult an education law attorney before your hearing.
Build your Tulane defense package
A personalized response that documents your writing process and addresses the detector that flagged your work.
Get your defense package$49 one-time · Generated in 60 seconds