Batch find and replace text across hundreds of PDFs

Under the hood

Why content-stream find and replace is different from search-and-replace in a viewer

A PDF reader's Find dialog walks the document, highlights matches, and asks the user to retype each one. That works for one file. It does not work for two hundred. PDF Batch Editor instead opens each PDF's content stream — the sequence of text-showing operators (Tj, TJ, ', ") that put glyphs on the page — locates every match, and rewrites the operators with new strings.

The result inherits the original font, size, and color. The replacement is written into the same content stream as the original, so the file is still a valid, signable, searchable PDF when you are done. Text-fitting modes adjust the replacement string when its rendered width differs from the original — Adaptive combines font scaling and horizontal compression for the best visual fit, Preserve Width keeps the original bounding box and adjusts character spacing, Fit to Page rescales the font, and None writes at the original size and lets the layout shift.

When the file already carries a digital signature, PDF Batch Editor uses an incremental update by default: changes are appended to the file after the existing %%EOF, leaving the original revision intact. Signatures attached to pages you did not modify stay valid. Signatures attached to a page you did modify are invalidated — which is exactly what the PDF specification requires.

De-identification

Forensic-grade removal — the original text is gone, not just hidden

For routine editing — updating dates, fixing typos, rebranding boilerplate — incremental save is exactly right: it keeps signatures valid on pages you didn't touch, and the original text living in a prior revision of the file is harmless. For de-identification, that prior revision is the problem. A scanned name, a patient identifier, an account number: if any of it survives in the file bytes, a forensic tool, hex editor, or PDF revision-history walker can pull it back out. This is the root cause of nearly every public PDF redaction failure on record — from the 2016 court filing to ongoing FOIA disclosures.

PDF Batch Editor's find & replace edits the text-showing operators inside the content stream directly — the replaced characters are genuinely gone from the page content, not hidden behind a black rectangle or a white overlay. When the output mode includes the post-replace optimizer pass (default for de-identification workflows), the file is then collapsed to a single clean revision and every superseded original is dropped. After this step the original text cannot be recovered from the file bytes by any forensic tool, content-stream extractor, hex editor, or PDF revision-history utility — the bytes simply do not exist in the file.

Verifying this is easy: a single, clean revision contains exactly one %%EOF marker. Two or more means prior revisions — possibly with the original text — are still present. Open one of the cleaned outputs in any hex editor and search for the original value; it won't be there.

This is the foundation of the app's HIPAA-conscious workflow for medical document processing — clinics use it to swap patient names with internal codes (or strip them entirely) before sending charts to third-party processors, with the assurance that what they're sending genuinely no longer contains the PHI.

Scanned & image-only PDFs

Built-in OCR — find and replace inside scans, no separate step

Earlier versions could only touch the content stream — the actual text operators drawn on the page. A flatbed scan has none: it is a picture of a document, and there was nothing to find. You had to run a separate OCR tool first. That is no longer the case. PDF Batch Editor now classifies every file as you load it — born-digital, scanned, or mixed — and routes each one to the right engine automatically.

When a page is a scan, a built-in OCR engine rasterizes it, recognizes the text in the image, and locates every occurrence of your search term — the same terms and CSV pairs you would use on a digital file. Because the text only exists as pixels, the replacement is not a content-stream edit: PDF Batch Editor burns the matched region out of the page image and draws your replacement text on top. The original characters are gone from the pixels — not whited out, not boxed over — which makes this the right tool for de-identifying scanned charts, statements, and forms, not just relabeling them.

A mixed document — a born-digital contract with a scanned signature page, say — is handled in one run: digital pages go through content-stream replacement, scanned pages through OCR-and-burn. You define the pairs once and PDF Batch Editor picks the correct mechanism per page.

OCR runs entirely on your machine — no page image is ever uploaded to a cloud service, which matters when the scans are medical records or signed agreements. Recognition quality depends on scan resolution; for faint or low-DPI source material you can raise the rasterization DPI. Pages that are not a single full-page image are flagged for review rather than altered unsafely, so a stray logo or stamp never gets silently painted over.

Pattern matching

Regex, case sensitivity, and whole-word matching

Each pair has independent regex, case-sensitivity, and whole-word toggles. The regex engine is .NET System.Text.RegularExpressions — full support for character classes, quantifiers, anchors, lookaheads, lookbehinds, and capture groups. Replacement strings reference captured groups with $1, $2, etc.

A few patterns that come up in real document sets:

Reformat US dates from MM/DD/YYYY to ISO YYYY-MM-DD
  Find:    \b(\d{2})/(\d{2})/(\d{4})\b
  Replace: $3-$1-$2

Mask all but the last four digits of an account number
  Find:    \b\d{12}(\d{4})\b
  Replace: ************$1

Update a versioned product name across an entire library
  Find:    \bAcme Suite v\d+(\.\d+)*\b
  Replace: Acme Suite 2026

Strip a draft watermark line from every page header
  Find:    ^DRAFT — do not distribute$
  Replace: (empty)

Whole-word matching uses an IsLetterOrDigit boundary check on the characters before and after the match. Searching Smith with whole-word on matches Smith, but skips Smithfield and Smithy. Case-sensitive matching is on by default; flip it off to make Acme, ACME, and acme equivalent.

Multi-replace

CSV-driven substitutions for hundreds of pairs

Annual updates are rarely one substitution. They are dates, addresses, contact names, internal codes, version strings — sometimes hundreds of pairs across a contract library. Build the change set in Excel or Google Sheets, save as CSV, and import:

find,replace
2025,2026
Q4 2025,Q1 2026
123 Old St.,500 New Ave.
"Acme Holdings","Acme Holdings, Inc."
v3.4,v3.5

Each row becomes one pair, applied in order. Pairs chain through temp files internally, so a later pair sees the result of an earlier one — useful when a substitution depends on an earlier one. Mix imported pairs with hand-edited pairs, expand any row after import to override its options individually, and export the current pair list back to CSV for version control or sharing with colleagues.

Per-pair controls

Page ranges, signatures, formatting, and conditional skipping

Each find/replace pair has its own independent settings, accessible by expanding the pair row.

Edge cases

Scanned PDFs, signed pages, encoding, and form fields

Scanned and image-only PDFs. These are handled automatically — no separate OCR step. PDF Batch Editor classifies each file on load; when a page is a scan, the built-in OCR engine rasterizes it, recognizes the text, locates your search term in the image, and burns the matched pixels out, drawing the replacement on top. See the Scanned & image-only PDFs section above for how this works. Recognition depends on scan quality, so raise the rasterization DPI for faint or low-resolution source material; pages that are not a single full-page image are flagged for review rather than altered.

Signed PDFs. Incremental save (default on) preserves the prior revision, so signatures over pages you do not modify continue to verify. Signatures attached to a modified page are invalidated by design — that is the entire point of a signature. If you need to keep all signatures intact, restrict the pair's page range so it does not touch a signed page.

Custom encodings and Identity-H fonts. Most PDFs use a ToUnicode CMap that maps glyph IDs back to readable characters. PDF Batch Editor extracts text through this CMap, so well-formed PDFs work transparently. PDFs generated by older or hand-rolled tools sometimes ship without a CMap or with a broken one — in those cases the live preview's match count may under-report even though the underlying content stream still contains your search text. If a file's preview shows zero matches but you can clearly read the text on the page, the PDF likely lacks a usable text-extraction layer.

AcroForm fields. Find/Replace targets the visible page content stream. AcroForm field values live in a separate dictionary; once a form widget's appearance has been generated and embedded into the page (the usual case after a form is filled), the visible text is matched and replaced like any other content. Empty, unfilled forms are best handled by the dedicated batch form-fill module.