When HTPBE? flags a PDF as modified, can it tell a harmless re-save or compression apart from real content changes?

Not by itself. HTPBE? reports that a file was changed after it was created and names each piece of evidence it found, but it does not label a change as harmless or harmful. A re-save, a compression pass and an edit to a figure are all changes to the file, and each can produce a Modified result. What you get is the kind of change that was seen, so you can judge it against what you know about the file’s history.

What the result does tell you

A Modified result lists its findings by name, and the names are grouped by the kind of evidence they report (see How does HTPBE? determine whether a PDF was modified?). The group helps you tell an ordinary step in a document’s life from something that needs a closer look:

• The file’s own history and dates. Findings that the file was saved again, rebuilt or processed by another tool, or that the dates recorded in it disagree. This is where an ordinary re-save or export shows up. A single finding here that matches a step you know happened, such as the sender saving the file again, reads very differently from several unrelated findings.
• Changes to what is on the page. Findings that values were covered over or replaced, that characters were altered, or that an image was edited and saved again. These point at the page content itself, and plain saving does not usually produce them.
• Signatures. Content changed after a digital signature was applied, or a signature removed so the document looks unsigned.

On a Modified result the confidence level adds to this. Certain means conclusive evidence, such as dates that contradict each other or changes after signing. High means strong evidence that a person should still look at.

What the result does not tell you

• Whether a change was harmless. HTPBE? states a fact about the file, not a motive. A re-saved file is not the original, and the result says so even when nothing you can see has changed. Treating re-saves as harmless would also let real edits through, because an edit usually arrives in a re-saved file too.
• Whether the content is the same as before. HTPBE? does not read the text, amounts, names or dates written in the document, so it cannot compare what the file said before and after a change.
• Why the file was processed. The result describes the file you uploaded. The reason it was opened, saved or compressed is something only the people who handled it can tell you.

Compression and similar processing

Compressing, converting or printing a PDF to a new PDF creates a new file built from the old one, and the verdict describes that new file, not the original. Depending on the tool, the result can be Modified, or Inconclusive when the file went through an online editor or converter, because a file made that way could have been created from scratch and the structure alone cannot settle it.

How to decide whether it matters

1. Read the finding names, not only the verdict. Note which kind of evidence they belong to.
2. Retrace the file’s journey. List who saved, compressed, merged or signed it. A step you can name often accounts for findings in the history group.
3. Check a copy from the source. Ask the issuer for the file straight from their system, or download it again from the portal it came from. If that copy is Intact and yours is Modified, the changes were made after the file left the source.
4. Report a finding that is wrong. If a finding names something that did not happen to the file, open the check in your dashboard and use Report an Incorrect Result. We re-analyse the PDF and review it manually.

When the result matters

If a Modified file is part of a decision, such as a loan, a tenancy, a claim or a hire, do not accept it on the strength of its appearance and do not reject the person automatically either. Treat the findings as a reason to ask for the original and to let the sender explain how the file was handled.

The ordinary causes of a Modified result are listed in Why does my PDF show as modified even though I didn’t edit it? and how to read each verdict is covered in How should I interpret PDF analysis results? To run the same check on your own files or inside your own workflow, see the HTPBE? API and pricing.