Free re-runs on still-flagged passages: finish the paint map, don’t wash the chapter again
The second AI-writing report is where panic gets expensive. A few continuous-prose spans are still painted; the LMS card still looks loud; the reflex is to paste the entire chapter into another “humanize” box and hope a private meter moves.
That reflex optimizes for throughput. It does not optimize for matching this attempt’s highlight map, keeping DOCX layout, or explaining to a supervisor which paragraphs actually changed.
This post is a recheck worklist—product-page wording about remaining flags, not a clearance certificate and not a detection-evasion guide.
Recheck literacy before you rewrite anything
Institutional AI-writing exports (Turnitin / iThenticate and peers) usually paint qualifying continuous prose. Lists, tables, captions, and reference blocks are often outside that map. Similarity still answers a different question than AI writing. Rewriting for one dial does not reliably move the other.
Before you spend a single credit on a second pass:
Download the full AI writing report for the same attempt you will resubmit.
Rank still-painted continuous-prose spans by confidence / intensity.
Edit those spans. Leave unflagged paragraphs alone unless they are factually wrong.
Freeze citations, verified numbers, formula objects, and claim strength unless the report itself painted them.
Prefer a DOCX / PPTX return path so styles and tables are not thrown away for a chat paste.
Track A: free re-runs until * under 20% (product page, not a promise)
HumanPen’s product page describes free re-runs on still-flagged passages when a recheck still shows ≥20%, toward * under 20%. That is a workflow and billing statement about remaining flagged continuous prose—not “guaranteed pass,” not “undetectable,” and not advice to bypass any school process.
In plain English:
The loop is scoped to spans the new report still paints.
Clean paragraphs stay out of the rewrite bill.
You bring your institutional export; we do not invent detector percentages beside the editor.
Track B: whole-document wash vs the free crumb
Many paste-box tools optimize for “rewrite everything you pasted.” After a recheck, that usually means paying again (or burning a subscription dump) for prose the new paint map never marked.
HumanPen’s mechanical difference:
Whole-document wash | Report-guided second pass |
|---|---|
Rewrites the full paste again | Changes still-flagged continuous prose from the new report |
Often bills by whole paste / monthly capacity | 100 free credits ≈ first ~1,000 words of rewrite for new accounts (≈ 1,000 words = 100 credits); then pay for words you actually rewrite |
Pure text out | DOCX / PPTX / TXT back when possible |
Private “human %” theater | Your Turnitin / iThenticate export is the source of truth for scope |
Put the free ~1,000 words on the highest-confidence remaining spans first. That is usually enough for a realistic section-level second pass—not a second thesis wash.
Five artefacts worth keeping for office hours
Attempt number + report timestamp for both exports.
A short list of spans that stayed painted vs spans that cleared.
Note of which citations / numbers / formulas you deliberately froze.
File identity: DOCX vs PDF, filename, word count.
The full PDF (or official export)—not a cropped badge screenshot.
Soft CTA
If you want a document-level loop—import the AI writing report, rewrite still-flagged passages under the product’s free re-run wording toward * under 20%, keep citations and layout where possible, and use the 100 free credits ≈ first ~1,000 words—start at https://humanpen.net (hub: lnk.bio/humanpen).
No honest tool should be sold as a guaranteed pass.
— Charles Li · HumanPen · https://humanpen.net
Join the Conversation
Share Your Thoughts
Related Articles
Freeze the bibliography before you touch the red spans
When an AI-writing report paints continuous prose, the instinct is to rewrite the whole submission. That usually breaks more than it fixes: citation styles drift, table captions get “helpfully” paraph

*% 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.