An AI cheating accusation can collapse in a single meeting when a student walks in with the right paper trail. The pattern keeps repeating: a detector flags an essay, the professor opens a case, and then the student produces handwritten notes and browser history that make the flag impossible to defend. Here is how that pattern plays out and what to do if a similar accusation lands on your desk at a school like Boston University.
The pattern: detector flag, process evidence, retraction
Cases that follow this arc share a structure. A submitted essay returns an AI-generated score from Turnitin or a similar tool. The instructor sends an email opening a conduct process. The student replies with a folder of artifacts: a spiral notebook of outlines and notes, browser history showing library database and JSTOR searches during the drafting window, and a Google Docs revision timeline that shows the paper being built over hours or days. The professor reviews the material and withdraws the allegation before it reaches a formal hearing.
This is not unique to any one campus. Similar reversals have been reported across the UC system and at private research universities. Our writeup of a UC Davis case that ended after a handwritten draft surfaced walks through the same sequence in detail.
Why detectors flag human writing in the first place
AI detectors score on statistical properties, not authorship. Low perplexity (predictable word choices) and low burstiness (consistent sentence length) push a score toward "AI-generated." Careful academic prose, formal history writing, edited drafts, and writing by non-native English speakers all tend to produce those signals.
A 2023 peer-reviewed evaluation by Weber-Wulff and colleagues in the International Journal of Educational Integrity tested fourteen detection tools and concluded that none reached a reliability level appropriate for institutional decision-making on their own. A Stanford paper by Liang et al. (2023) in Patterns found detectors misclassified non-native English writing at strikingly high rates. Neither study endorses using a detector score as sole proof of misconduct.
Why handwritten notes and browser history are decisive
A detector score is a statistical guess about a finished text. Process evidence is a record of how that text came to exist. The two operate at different levels of certainty, and process evidence generally wins the tie.
- Handwritten notes show pre-writing thinking: outlines, thesis attempts, source annotations, and margin questions. Photographs of notebook pages with visible dates and the surrounding pages establish continuity that is hard to fabricate after the fact.
- Browser history during the drafting window shows library database searches, JSTOR or Google Scholar queries, article reads, and citation lookups. Timestamps line up with the intellectual work of the paper.
- Google Docs revision history shows the document growing paragraph by paragraph, with revisions, deletions, and reordering. A paper pasted in from a chatbot appears as a single large paste with no build.
What a BU-style academic conduct process requires
Boston University's Academic Conduct Code for undergraduates in the College of Arts and Sciences, and comparable codes in other BU schools, treat plagiarism and misrepresentation as violations that must be established through a documented process. Instructors typically report suspected violations to the school's Academic Conduct Committee, which reviews evidence and gives the student an opportunity to respond in writing and in person.
A detector score is one piece of information an instructor may consider, but the code does not designate any AI detection tool as authoritative evidence. The committee weighs the record as a whole. That is exactly the point at which process evidence changes the outcome. For a fuller walkthrough of what to ask for in writing, see our procedural rights FAQ.
Questions to ask in writing
- Which specific detector produced the score, and what was the reported percentage?
- Did any human review of the flagged passages take place before the allegation was filed?
- What is the standard of evidence the committee applies, and where is it documented?
- What deadline applies for submitting a written response and supporting evidence?
Evidence that tends to close these cases
Across the reversals reported publicly, a recurring evidence bundle appears: Google Docs revision history exported and screenshotted, browser history for the drafting window saved as HTML or CSV, photographs of handwritten notes and outlines, citation manager entries (Zotero, EndNote) with import timestamps, and library database access logs where available. Version history alone can be sufficient. Combined with browser history and physical notes, the record becomes very difficult to dispute.
Preservation timing matters. Browser history retention defaults are often 90 days on Chrome and shorter on some mobile browsers. Google Docs retains detailed revision history indefinitely for most accounts, but exporting screenshots and a downloaded copy protects against later account issues. Our evidence-gathering guide covers the first 48 hours in detail.
If this is you at Boston University or a similar institution
Move in a defined order. First, read the notice carefully and note every deadline. Second, preserve everything: revision history, browser history, notes, messages with classmates about the assignment, and any research artifacts. Third, request in writing the detector name, score, and any human-review notes. Fourth, write a response that pairs your process evidence with the research on detector reliability and the specific language of your school's conduct code.
If you are drafting that written response, NotBot generates a personalized defense package that names the detector that flagged you, cites the relevant peer-reviewed research, and organizes your process evidence into a response letter and hearing prep brief. If the case has already reached a finding and you are considering next steps, the appeal package covers the procedural grounds that matter after a hearing.
If the potential sanction includes suspension, expulsion, or visa consequences for an international student, consult an education law attorney before your hearing. Research findings and process evidence carry a case a long way, but severe-consequence matters warrant professional counsel.
Build your BU 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