Redact PDF Free Online Before Using AI: Automatically Remove Names and Personal Data
Before you upload a PDF to ChatGPT or another AI, HILO automatically finds the names and personal data and replaces them. There is no need to hide anything by hand.
You get an editable Word file: lighter and clearer for ChatGPT and other AI tools (it uses fewer tokens).
Your document is not passed to any AI: it does not leave your computer and is not uploaded or stored. The name detector runs inside your browser. You can check it: after the first document, disconnect from the internet and it keeps working.
Options
Document language:
Citations of published works (“(Smith, 2019)” and the reference list) stay as they are. Turn it off for a blind review.
An italic paragraph, in the document’s language, that explains to the AI what the codes in brackets are and asks it not to try to guess the data.
Do you already have a HILO lookup table (from a dataset or another document)? Load it and the same people and data keep their codes.
First time using the anonymizer? To process your document without it leaving your phone, HILO has to download, only once, the dictionaries and files it needs (about 140 MB). We recommend using Wi-Fi. On a phone, the process may take longer than on a computer.
Do not close this tab.
People
ID numbers, phone numbers, and emails
Uncheck what you do not want to replace. The names you see here do not leave your computer.
To review
If you do not review them, they are covered anyway. Review them so you do not cover words that are not data.
Before downloading, look at this
These words were left in the anonymized document and could be people or data. Replace the ones that are.
From PDF to a Word file without data, in three steps
1. Load it
Drag in the PDF. If it is scanned (a photo of paper), HILO tells you right away.
2. Review
Review the names and data it found on every page, including the headers.
3. Use it with AI
Download a lightweight Word file with clean text and give it to the AI.
An editable Word file, and what cannot be read
Redacting a PDF by hand with a black box (or a black highlight) often leaves the text underneath: it can be copied or “unredacted”. HILO does not cover anything: it finds the names and data on its own and builds a new Word file with the text already replaced, with the paragraphs and tables. The PDF’s comments, annotations, and properties are not carried over.
Scanned PDFs (photos of paper) cannot be read yet: they have no text, only images. HILO tells you if that is the case. More in the guide How to redact a PDF properly.
What happens to each part of the PDF
The new Word file carries the text, not the drawing of the page:
- paragraphs and tables, with the data already replaced;
- repeated headers and footers, only once;
- images are not carried over, and you are told how many there were;
- nor are the PDF’s comments, annotations, or properties.
That is why nothing is left under any black box: whatever is not in the Word file is simply not there.
Why a Word file works better than a PDF for AI
A PDF brings line breaks, columns, and repeated headers that the AI has to untangle. With the Word file:
- the file is lighter and fits in whole;
- the AI spends fewer tokens on formatting;
- tables arrive as tables;
- you can fix something before sending it.
Documents worth anonymizing before using AI
Plans
Start free. When you want more, subscribe or buy hilos on their own.
- Keyword search, coding, and annotating, with no limit
- Unlimited downloads of the edited document
- Error diagnostics and tools for datasets
- Topic and code summaries
- Kappa calculator
- AI features (concept maps, assisted coding, checklist, logic model, variable operationalization), with your hilos
- Anonymize documents: 1 without an account, 5 per week with your account (ES · EN · PT)
- 325 hilos per week for AI features (renewed every Monday at 00:00 UTC)
- Change tracking (redline)
- Change history and traceability
- Anonymize documents: 50 per week (ES · EN · PT)
Payment is processed by Paddle. HILO doesn't see or store your card details.
The week renews on Monday. Each anonymized document you download counts once per file.
Because it is safer and more useful: a retouched PDF can hide the original text, and a Word file can be edited and pasted into the AI. If you need a PDF, save it as PDF from Word.
Not yet: a scanned PDF is a photo of each page, with no text. HILO tells you and does not process it.
The text, paragraphs, and tables are kept; the exact page layout is not. The idea is to have the content ready to work with.
They are not carried into the Word file, and HILO tells you how many there were, in case any had data.
No. It is read in your browser and the Word file is built right there.