# Flagged for AI on something I wrote myself. Here is what helped **URL:** \ **Category:** Education & Academic Writing\ **Created:** [September 14, 2026, 11:08pm UTC](https://humanizeaiforum.com/t/flagged-for-ai-on-something-i-wrote-myself-here-is-what-helped/203 "2026-09-14T23:08:04Z")\ **Posts on this page:** 7\ **Page:** 1
**Author:** ![codewrath58](https://avatars.discourse-cdn.com/v4/letter/c/ecccb3/32.png) [@codewrath58](https://humanizeaiforum.com/u/codewrath58)\ **Post date:** [September 14, 2026, 11:08pm UTC](https://humanizeaiforum.com/t/flagged-for-ai-on-something-i-wrote-myself-here-is-what-helped/203/1 "2026-09-14T23:08:04Z")
flagged for AI on something I wrote myself. here is what helped. posting this because I searched for this when it happened to me and couldn’t find anything useful. got a 68% AI score on a paper I wrote from scratch. no AI tools used at all. the professor said I needed to prove I wrote it. spent a week figuring out what to do. ---
**Author:** ![rough.Matrix](https://yyz1.discourse-cdn.com/flex033/user_avatar/humanizeaiforum.com/rough.matrix/32/35_2.png) [@rough.Matrix](https://humanizeaiforum.com/u/rough.Matrix)\ **Post date:** [September 14, 2026, 11:21pm UTC](https://humanizeaiforum.com/t/flagged-for-ai-on-something-i-wrote-myself-here-is-what-helped/203/2 "2026-09-14T23:21:02Z")
Worth testing a few on your actual use case rather than the demo examples they show. In my experience the gap between the best-case demo and what you get on a generic brief is significant. Your mileage will vary based on what you’re actually trying to produce. ---
**Author:** ![stormhaze71](https://avatars.discourse-cdn.com/v4/letter/s/e47774/32.png) [@stormhaze71](https://humanizeaiforum.com/u/stormhaze71)\ **Post date:** [September 15, 2026, 5:52pm UTC](https://humanizeaiforum.com/t/flagged-for-ai-on-something-i-wrote-myself-here-is-what-helped/203/3 "2026-09-15T17:52:07Z")
false positives on human-written work happen more than the tools’ marketing suggests. formal academic writing, ESL writing, and any piece written in a highly structured conventional style are particularly vulnerable. the tools are measuring writing patterns, not authorship. ---
**Author:** ![Arclion003](https://yyz1.discourse-cdn.com/flex033/user_avatar/humanizeaiforum.com/arclion003/32/22_2.png) [@Arclion003](https://humanizeaiforum.com/u/Arclion003)\ **Post date:** [September 16, 2026, 1:35pm UTC](https://humanizeaiforum.com/t/flagged-for-ai-on-something-i-wrote-myself-here-is-what-helped/203/4 "2026-09-16T13:35:07Z")
the “prove you wrote it” request puts the burden of proof in the wrong place but it’s what you’re going to be asked to do. the evidence that works: draft history from your word processor, writing notes, search history from your research, anything that shows the piece being developed over time. ---
**Author:** ![ErnestPrice](https://yyz1.discourse-cdn.com/flex033/user_avatar/humanizeaiforum.com/ernestprice/32/24_2.png) [@ErnestPrice](https://humanizeaiforum.com/u/ErnestPrice)\ **Post date:** [September 17, 2026, 2:56pm UTC](https://humanizeaiforum.com/t/flagged-for-ai-on-something-i-wrote-myself-here-is-what-helped/203/5 "2026-09-17T14:56:07Z")
the tools themselves can help your case. running the same piece through multiple detectors and getting inconsistent results is evidence that the detection is unreliable. a piece that scores 68% on one tool and 30% on another is a piece the detection community disagrees about. ---
**Author:** ![tallest.tower.2016](https://avatars.discourse-cdn.com/v4/letter/t/c89c15/32.png) [@tallest.tower.2016](https://humanizeaiforum.com/u/tallest.tower.2016)\ **Post date:** [September 18, 2026, 5:40am UTC](https://humanizeaiforum.com/t/flagged-for-ai-on-something-i-wrote-myself-here-is-what-helped/203/6 "2026-09-18T05:40:06Z")
went through this. here is the process that worked. first: ran the piece through [Proofademic](https://proofademic.ai/). the sentence-level output showed exactly which sentences were flagging. free first 1,000 words, no card. I could show the professor that the flagged sentences were transitional phrases and topic sentences, the most formulaic parts of any academic paper, not the analysis sections where my actual thinking was. second: pulled my draft history from Google Docs. the version history showed the paper being written in non-linear order with deletions and rewrites, which is how humans write, not how AI generates. third: ran it through three other detectors and got three different scores. used that disagreement as evidence that the score was not reliable. the professor accepted the explanation. the key was having specific evidence for specific sentences, not just asserting I wrote it. ---
**Author:** ![maya.glitch](https://yyz1.discourse-cdn.com/flex033/user_avatar/humanizeaiforum.com/maya.glitch/32/39_2.png) [@maya.glitch](https://humanizeaiforum.com/u/maya.glitch)\ **Post date:** [September 19, 2026, 12:13pm UTC](https://humanizeaiforum.com/t/flagged-for-ai-on-something-i-wrote-myself-here-is-what-helped/203/7 "2026-09-19T12:13:07Z")
the draft history step is the one most people don’t think of. if you write in Google Docs or any cloud processor, the version history exists. it shows the work being built, which is the most direct evidence you can produce.