You revised the essay on your laptop—but the LMS still scored last night's upload
Across English-medium programmes in India and Southeast Asia, the same mismatch shows up every term. You finished a cleaner draft after dinner. You kept the file name almost identical. You opened Moodle, Canvas, or Blackboard again and felt sure the new version was "in." Hours later the AI writing card still described paragraphs you already deleted.
This post is about file-version literacy inside AI writing reports. It does not promise any score, and it is not a detection-evasion guide.
Why the percentage can describe a ghost file
An AI writing indicator scores the upload the tool received, not the document open on your laptop. If the portal kept the previous attempt, if you uploaded a PDF exported before the last edit, or if a co-author submitted from another drive, the report can be accurate about the wrong object.
Students in Delhi, Bengaluru, Singapore, Kuala Lumpur, Bangkok, and Jakarta often compare screenshots in WhatsApp before they compare timestamps. That is how a version mismatch turns into a rumour about "the tool."
Separate the three clocks
Clock | What it answers | Common failure |
|---|---|---|
File modified time on your laptop | When your copy last changed | Not the same as "submitted" |
LMS submission receipt / attempt number | What the portal accepted | Multiple attempts; wrong attempt selected for grading |
AI writing report date / report ID | Which binary the classifier scored | Report regenerated on an old file; delayed processing |
If those three clocks disagree, arguing about the percentage first is the wrong order of work.
Six checks before you rewrite a single paragraph
Download the exact file the LMS lists for that attempt. Open it. Search for a sentence you remember adding last.
Compare word count and page count with the draft on your laptop. A missing conclusion is often a version problem, not a "style" problem.
Note DOCX vs PDF. A PDF exported at 11:40 and a DOCX saved at 00:15 are different submissions even when the title on the cover page matches.
Read the report header for tool name, report time, and attempt. A classmate's card from another module is not a calibration stick.
Open the full AI Writing Report, not only the dashboard badge. Confirm which paragraphs are highlighted in this file.
Only then decide whether revision is about wording—or about resubmitting the correct version under your school's rules.
Similarity and AI writing still answer different questions
A low similarity score does not cancel an AI writing indicator. An AI writing indicator does not prove plagiarism. Version mismatch makes both instruments noisier, because both may be talking about text you no longer consider "yours" in the current draft.
Polished, template-trained English common in many India and SEA programmes can look "predictable" to an estimator. That is a pattern observation about qualifying continuous prose, not a courtroom finding about authorship—and it is meaningless if the scored file is yesterday's export.
If your school already asked for revision
Confirm the file identity in writing (attempt number, filename, timestamp) before you rewrite. Scope edits to flagged continuous prose in the file that will actually be re-checked. Freeze citations, tables, and verified facts unless they are wrong. A narrow, documented revision is easier to explain in office hours than a total rewrite that erases your draft history.
For a document-level workflow—import an AI writing report, change only flagged passages, keep citations and layout in DOCX / PPTX / TXT—see HumanPen. It still does not promise what the next LMS card will show.
Takeaways
The card scores an upload, not the file open beside it.
Align laptop time, submission receipt, and report ID before you debate a percentage.
No honest tool should be described as a "guaranteed pass."
— Charles Li · HumanPen · https://humanpen.net
Join the Conversation
Share Your Thoughts
Related Articles
*% Versus 21%: Do Not Over-Read the First Visible AI Number
The AI writing indicator does not always show a tidy percentage. A display band — not a gavel — explains *% and ~21%.
Authorship Flag vs AI Detection: Two Systems People Mix Up in Reports
Authorship and AI detection answer different questions with different methods. How to read each flag without mixing them up.
The AI Badge Did Not “Miss” Your Rewrite—It May Be Scoring an Older Upload
An AI writing percentage scores a submitted snapshot. Version mismatch — not a stubborn detector — explains most “I fixed it and the score stayed” nights.