AI Content Rinse: Text & Metadata Cleaner

AI Content Rinse helps review content copied from AI tools and other sources.

• Scan post/page titles, content and excerpts for zero-width spaces, BOM characters and soft hyphens.
• Preview the proposed text changes before applying them.
• Review the current title, content and excerpt directly in the block editor, including unsaved edits. Apply the reviewed changes to the editor and use WordPress Undo or Save.
• Choose invisible-character removal and long-dash replacement independently. Both are enabled initially; preferences are saved for each administrator.
• Preserve HTML tags, block comments, attributes, shortcodes, code, Cyrillic and emoji joiners.
• Restore text backups when no subsequent editorial changes would be overwritten.
• Paste plain text or WordPress/HTML content, preview exact character changes and copy the cleaned result.
• Load a demonstration text with invisible characters and long dashes to try the preview. Replacing entered text requires confirmation.
• Search content, filter findings on the current page and review selected items before batch cleanup.
• Clean supported image metadata in existing files, including available resized versions and saved originals, without regenerating image sizes.
• Verify staged bytes before replacement and scan the saved files again. Report per-file results and incomplete operations.

This release removes JPEG comments/XMP, PNG text and caBX C2PA chunks, and WebP XMP. It preserves EXIF (including GPS), ICC colour profiles and source pixel data. Media IDs, filenames and URLs are retained. Each file must be local and at most 20 MB; up to 100 existing variants per attachment are checked. Missing or unsupported variants block the preview. Replacement uses a verified temporary staging file; no new attachment or permanent image copy is created. An attachment with multiple files is processed one file at a time; failures are reported and a fresh scan is required to retry.

No external service, account, tracking or AI API key is required. Findings do not establish AI authorship. Provenance detection currently recognizes PNG caBX blocks and textual OpenAI/GPT references inside them; it does not validate signatures. C2PA in other formats, visible logos and pixel-domain watermarks are outside this release. It does not rewrite prose, support page-builder data, or automatically clean uploads or posts. Image metadata removal has no undo. Existing browser/CDN caches may still serve old bytes until refreshed. Pasted text is processed locally on your WordPress server without saving a post.