UCSB Professor Retracts AI Accusation After Browser History Evidence

A UC Santa Barbara student who was accused of using AI on a written assignment had the allegation withdrawn after producing browser history and timestamped drafts that traced the paper from research to final submission. The case follows a pattern that has emerged across UC campuses: a detector flag prompts an accusation, and process evidence closes it. Here is what the student produced and why it worked.

What happened

The accusation followed a familiar sequence. An assignment was submitted, an AI detector returned a high probability score, and the instructor raised the score with the student. The student was given an opportunity to respond before the case was referred to the campus conduct office.

Rather than argue the detector was wrong on its own terms, the student produced a package of process evidence: browser history covering the drafting window, Google Docs version history showing the paper built up over multiple sessions, and file timestamps matching the search and reading activity. After reviewing the material, the instructor withdrew the allegation.

The withdrawal is consistent with the broader UC pattern documented in earlier cases, including a separate UCSB retraction driven by version history and a UC Davis case dropped after handwritten drafts.

Why browser history and draft timestamps worked

AI detectors produce a probability estimate based on statistical properties of the submitted text. They do not observe how the document was written. Browser history and draft timestamps do exactly the opposite: they show a sequence of authorship activity that a language model cannot produce.

Three things made the evidence persuasive:

• Chronological coherence. Search queries in the browser history matched sources cited in the paper, and both preceded the paragraphs where those sources appeared in the draft.
• Multiple drafting sessions. Version history showed the document evolving over days, not appearing fully formed in a single paste.
• Independent timestamps. Browser history is stored by the browser, version history by Google, and file metadata by the operating system. Fabricating all three consistently is not practical.

How to prove you did not use AI with browser history

Browser history is one of the strongest pieces of authorship evidence available, but it has a short shelf life. Chrome and Firefox default to a 90-day retention window on some systems, and clearing browsing data wipes it permanently. Preserve it before you do anything else.

1. Export your history immediately. In Chrome, use a history export extension or open chrome://history and screenshot the relevant date range. In Firefox, use the Library window to export as HTML. In Safari, screenshots of the History window with timestamps visible are the practical option.
2. Filter to the drafting window. Pull the entries from the day you started research through the submission date. Highlight searches and page visits that match your citations.
3. Cross-reference with your bibliography. For every source in your paper, identify the corresponding search query or page visit and the date it occurred.
4. Pair it with draft timestamps. Open Google Docs version history (File → Version history → See version history) and screenshot the timeline. Match specific revision times to research sessions in the browser log.
5. Email a copy to yourself. The received timestamp acts as a third-party record that the evidence existed on that date.

For a fuller walkthrough of the Google Docs side, see the guide on using version history to fight an AI accusation.

What UCSB policy requires

UC Santa Barbara handles alleged academic integrity violations through the campus Student Conduct process, governed by the UC systemwide Policies Applying to Campus Activities, Organizations, and Students (PACAOS) and the UCSB Student Conduct Code. Instructors are expected to consult with the student before referring a matter, and findings must be supported by a preponderance of the evidence.

A detector score is not, on its own, a preponderance. UC teaching centers cautioned faculty in 2023 that AI detectors should not be treated as reliable evidence of misconduct. When the detector is the only evidence and the student produces process artifacts that are inconsistent with AI generation, instructors have latitude to withdraw the referral before it reaches a formal hearing.

If this is you at UCSB

The steps that closed this case are available to almost any student, but only if the evidence is preserved quickly. Take these actions in order:

• Stop editing the flagged document. New saves overwrite version history metadata that supports your case.
• Export browser history for the full drafting window today, not next week.
• Download the Google Docs version history and screenshot the timeline.
• Ask the instructor, in writing, which detector was used and what score triggered the referral. The procedural rights FAQ covers what you can request before responding.
• Write a short narrative that ties your browser history, drafts, and citations together into a single timeline.

If you are drafting a written response, NotBot generates a personalized defense package that names the specific detector, integrates your process evidence, and matches the procedural expectations of your campus conduct office. If your case has already gone through a hearing and you are considering next steps, the appeal package covers the grounds that actually get reviewed.