MarkCleaner

Field notes on publishing AI-assisted text without the artifacts detectors and platforms flag.

MarkCleaner is a working notebook for content teams that generate with ChatGPT, Claude, or Gemini and then publish the result. Every page here targets a single gap the big "watermark remover" guides leave open — and leads with the answer.

Latest

Remove Invisible Characters Safely: Why Normalization Is Not Removal

A reproducible NFC/NFKC check that distinguishes normalization from deletion and flags meaningful joiners for review.

ChatGPT Hidden Characters: Find Where They Entered Your Text

A local, stage-by-stage method to trace code-point changes across a draft, editor, CMS, and published page.

How to Clean ChatGPT Text Before You Publish

A free, no-membership workflow: strip metadata, strip invisible characters, then break the statistical signature. Includes a 30-second pre-publish checklist.

Compare ChatGPT Text Before Removing Characters

A local before-and-after code-point check to identify copy-and-paste changes without calling them proof of a watermark.

What we cover

The three layers real detectors read — provenance metadata, invisible Unicode, and the token-distribution signature — and the free, repeatable routine to handle each before anything goes live. No paid plan required for normal publishing.