AI put a black box over the PDF text. Is the text actually gone?

Reopen the saved copy and check underlying content. Covering what you see can leave text available for extraction.

Do not treat a visible black rectangle as evidence that a PDF’s text was removed. In our local fictional test, the covered text was invisible in the rendered page but still appeared when we extracted text after reopening the file.

What we actually tested

We created a one-page PDF containing the invented value DEMO-4821, and a version with a black rectangle drawn over it. Both yielded DEMO-4821 to pypdf text extraction. A reference PDF generated without that value did not. The reference was not created by a redaction tool.

Use removal, not just visual marking

Adobe’s current Acrobat Pro guide describes selecting text or images for redaction, applying it, choosing whether to sanitize hidden information, and saving. We read that procedure but did not test Acrobat or an AI assistant. A marked area and an applied operation are different stages.

Check the saved file before sharing

Our recommendation is to keep the source private, reopen the intended release copy and try search and text extraction for the removed content. Also review hidden information using the editor’s relevant tools. A clean extraction result alone is not a full privacy audit.

Keep the finding narrow

The fixture contains simple text and a rectangle. It does not test scanned images, attachments, metadata, signed documents or every PDF editor. If AI says it removed confidential content, verify the actual saved artifact rather than accepting its description.

Published October14, 2026. Three wholly invented PDFs were created locally, reopened, extracted and rendered. No personal data, external AI upload, Acrobat session or general redaction-security audit was involved.

Sources: Adobe · Redact PDFs.