What should a legal team do when stamping strips bookmarks, form fields, or metadata from PDFs?

What stamping can silently remove

A PDF is more than its visible pages. It can carry bookmarks that map a long document's structure, interactive form fields with typed values, annotations such as comments and highlights, embedded file attachments, a text layer from OCR, hyperlinks, and document-level metadata like title, author, and creation date. Some stamping tools rebuild each page as a flat image or rewrite the file in a way that drops these objects. The output looks the same on screen, and the loss goes unnoticed until someone needs the bookmark or the form value.

The most consequential losses are form field values and the text layer. A form whose fields are dropped rather than flattened produces a blank form: the typed answers disappear. A text layer that is discarded turns a searchable document into an image, breaking keyword search and any redaction that relied on text. Bookmarks and links are less critical to the record but painful to lose on long documents such as agreements with many schedules, where the bookmark tree is how anyone navigates. Related: What exactly is Bates numbering and why do law firms rely on it for documents?

Keep reading: What exactly is Bates numbering and why do law firms rely on it for documents?, How do you keep Bates numbering consistent across multiple separate document productions?, When should you apply confidential or privileged watermarks to a batch of documents?. See how PdfStampr helps you bulk pdf stamping and bates numbering.

Deciding what should survive and what should not

Not everything should survive, but be precise about which metadata you are talking about, because two different things share the name. The first is the original document's metadata: author, dates created and modified, custodian, original file path, and the other fields captured at collection. That is evidence. ESI protocols generally require it to be preserved and produced, through the load file fields or with the native file, and stripping it from what you produce is a recognized deficiency and a spoliation risk, not something a protocol asks for. The second is the document properties of the production PDF you generate: the Title, Author, Producer, and creation fields that the stamping tool or the producing firm writes into the image file. Those are yours to set, and they should not leak reviewer user names, workstation paths, or internal system names. Beyond metadata, comments and annotations added by your own reviewers are work product and generally must be removed before production, and hidden layers and embedded attachments can contain material that was never reviewed. So the right question is not how to preserve everything, but which objects carry evidentiary or practical value and which carry risk. Related: What are the best practices for adding exhibit labels to documents before a hearing?

A reasonable default for a litigation production is to preserve the text layer, flatten form fields so their values become permanent page content, keep bookmarks where the protocol permits, remove annotations and comments, carry the original metadata through the load file or the native rather than relying on the image to hold it, set the production PDF's own document properties to neutral agreed values, and detach or separately produce embedded files. For a transactional or internal use, such as stamping a draft agreement, you usually want to keep bookmarks and links intact and leave form fields interactive. The stamping settings should follow from that decision rather than from tool defaults. Related: How do you keep Bates numbering consistent across multiple separate document productions?

Configuring the stamping pass

Look for three capabilities in whatever tool you use. First, content-preserving stamping, which adds the stamp as a new page object without rewriting existing objects; this keeps bookmarks, links, and the text layer. Second, explicit flatten options that let you choose to flatten form fields and annotations separately. Third, a metadata setting that lets you clear or set document properties. If a tool only offers a single flatten-everything mode, it is fine for simple scans but not for structured documents.

Order the operations deliberately. Flatten form fields first so the values become fixed content, remove annotations and comments, set the production PDF's document properties, then stamp. If OCR is needed, run it before stamping so the stamp text does not get mixed into the recognized text of the page, or run it after with the stamp region excluded, depending on what the tool supports. Do a dry run on a handful of representative files and compare the input and output object lists before processing the whole set. Related: How can a small legal team batch-stamp thousands of pages without losing a whole day?

Checking the output

Verification here is object-level, not just visual. Open a sample of stamped files and confirm bookmarks are present and navigate correctly, that form values appear on the page and are no longer editable, that a text search finds a known phrase, and that the document properties show the intended neutral values rather than a reviewer's user name or an internal path. Many PDF tools can list page objects, fonts, and attachments, which makes this check fast for a batch.

When something has been lost, do not try to reconstruct it in the stamped file. Go back to the pre-stamp set, adjust the settings, and re-stamp. Because the Bates range is assigned by the stamping pass, re-running it on the same ordered set produces the same numbers, so a corrected batch replaces the flawed one cleanly. This is one more reason to keep the unstamped, finalized set intact until the production has been delivered and accepted.