分享前清除 PDF 元数据：一份 PDF 里藏着什么 / How to Remove PDF Metadata Before Sharing

一句话结论 / Short answer

清除 PDF 元数据是发出文件前的最后一步，不是设置里勾一次的开关。你用来生成 PDF 的每个程序都会往里盖点东西：作者名、公司名、创建时间、修改时间、原始文件名，手机扫描件有时候还带坐标。这些信息不跟着页面内容显示，只跟着文件走，收件人打开属性面板就能看到，而多数人从不打开。

Removing PDF metadata before you send a file takes under a minute, and most people never do it. The cost shows up later: a proposal that still carries a former colleague's name as author, a contract draft whose timestamps put the drafting four months before it went out, a phone scan whose coordinates survive the export. Every application that writes a PDF stamps something into it, and almost nobody opens the properties panel afterwards.

判断顺序很短：先看这份 PDF 交给谁，再决定清到哪一档。交给外部的外发文件清到只剩必要字段；内部传的草稿可以留着协作信息；要走合规流程的，清除动作本身也要记一行，因为清掉的东西同样属于留痕范围。

The order of decisions is short. Look at who receives the file, then decide how far to strip it. A file going outside gets reduced to the fields it needs. A draft circulating inside can keep its collaboration details. Anything moving through a compliance process needs the stripping recorded too, since what you removed belongs to the same record. That last point is the one written up in /blog/pdf-file-audit-trail-requirements.

下面把元数据拆成八个字段，讲清每类清理方式停在什么地方，最后给一套当天能跑完的七步流程。

一份 PDF 里记了什么 / What a PDF records

八个字段按外泄后的麻烦程度排列，前面四个最常见，后面四个最容易被忽略。

Eight fields, ordered by how much trouble they cause once they leave. The first four are the common ones, and the last four are the ones people overlook.

最后两行值得单独说。XMP 扩展字段单看一条没什么，合起来能拼出一份完整的协作名单和修改节奏；地理位置字段是唯一一条无法用常识推断出来的信息，它直接把拍摄地点交出去。

Those last two rows deserve a note. An XMP field means little on its own, and a set of them reconstructs the collaboration list and the revision rhythm. The location field is the one piece nobody could guess, and it hands over where the picture was taken.

三类清理方式停在哪里 / Where each method stops

清理 PDF 元数据的路子不多，差别集中在两件事上：清到哪一层停手，以及文件在这个过程中去了哪里。

There are few ways to clean PDF metadata, and they differ on two things: how deep they go before stopping, and where the file travels while it happens.

第三行的代价写在第四列。为了让别人帮你清一份文件里的名字，你先把整份文件交了出去，而那些名字里带着的信息本来就是你想守住的东西。这类取舍在 /blog/privacy-first-pdf-workflow 里按场景拆过一遍。

The cost of row three sits in column four. To have someone else remove the names from a file, you hand over the whole file first, and the names are exactly what you were protecting. That trade is broken down by scenario in /blog/privacy-first-pdf-workflow.

第四行也不是万能的。浏览器本地处理守住了文件去向，代价是批注层、附件层和隐藏文字层要单独动手，多数工具不会替你一并清掉。怎么自己验证文件没有外传，见 /blog/devtools-network-tab-privacy-guide。

作者名与常漏的四个字段 / Author name and four others

作者名清除是最常被做对一半的那件事。清完信息字典就以为结束了，剩下的四处还留着。

1. 作者名：把系统用户名改成对外姓名，比事后逐个清更省事，新文件从生成时就带对的值
2. 原始文件名：另存为之前先定名，内部代号不要进档
3. 批注与修订：批注作者是独立字段，删掉批注内容不等于删掉作者名
4. 嵌入附件：嵌在 PDF 里的 docx 自带一套元数据，主文件清干净它也还在
5. 隐藏文字层：OCR 层与白字层不显示在页面上，全选复制就出来

1. Author: changing the system username to the name you use outside beats clearing files one by one, and every new file then carries the right value from creation
2. Original file name: settle the name before save-as, and keep internal codenames out of it
3. Comments and revisions: a comment author is its own field, so deleting the comment text does not delete the name attached to it
4. Embedded attachments: a docx sitting inside the PDF carries its own metadata, and it survives a clean main file
5. Hidden text layers: the OCR layer and white-on-white text do not show on the page, and select-all plus copy brings them straight out

第一条改的是源头，其余四条只能逐个处理。改用户名的收益在时间上后置：下次重装系统或换机器时要记得再改一次。

发出去之前的七步 / Seven steps before sending

按顺序走完，第七步那行记录可以直接进外发清单。前三步解决清到什么程度，后四步解决清没清干净。

Run them in this order, and the line from step seven goes straight into the outbound checklist. The first three settle how far to strip; the last four settle whether the stripping worked.

1. 先定接收方：外部、内部还是合规流程，这一档决定后面清到什么程度
2. 另存一份副本，在副本上清理，原文件保留
3. 清信息字典：作者、公司、标题、关键词、创建与修改时间
4. 清 XMP 扩展字段与自定义属性
5. 单独处理批注层、附件层和隐藏文字层
6. 打开属性面板复查一遍，确认剩下的字段是有意留的
7. 需要留痕的，把清理前后的字段清单连同时间记一行

1. Fix the recipient first: outside, internal, or a compliance process, since that decides how far to strip
2. Save a copy and clean the copy, keeping the original
3. Clear the info dictionary: author, company, title, keywords, creation and modification dates
4. Clear the XMP packet and any custom properties
5. Handle the comment layer, attachments and hidden text separately
6. Open the properties panel once more and confirm what remains is there on purpose
7. Where a record is required, write one line with the before and after fields and the time

合并这一步在 pdfmergenext.shop 全程本地完成，挑页码、调页序、清元数据都不需要把文件交出去。第七步那行记录由你自己写，谁也替不了。挑页语法见 /blog/pdf-page-selection-1-3-5-syntax，加密文件的处理顺序见 /blog/encrypted-pdf-can-you-merge。

Merging on pdfmergenext.shop runs on your device, so selecting pages, reordering and clearing metadata never means handing the file over. The line from step seven is yours to write. Page selection syntax is in /blog/pdf-page-selection-1-3-5-syntax, and the order for encrypted files in /blog/encrypted-pdf-can-you-merge.

更多 PDF 隐私与合规的判断细节见 /blog，合并时元数据的去留规律见 PDF 元数据保留，外发前商业机密的那份清单见 商业机密保护，GDPR 选型前的检查项见 GDPR 合规 PDF 工具。