A client list, a customer export or a booking sheet can go to ChatGPT, Claude, Gemini or Copilot once the people in it are replaced with placeholders. That takes more than clearing one column: names also sit in notes, sheet names, comments and hidden sheets.
This page is about spreadsheets, whatever your profession. For a general list of what to mask in any text, see PII masking for AI prompts. For bookkeeping files in particular, see AI for accountants.
Which columns carry personal data
The obvious column is the name. These are the ones people leave in:
| Column | Why it identifies someone |
|---|---|
| Name, first name, last name | Direct. Check for a separate "contact person" column too. |
| Company | Names the client, and often the person behind a small business. |
| Email, phone | Direct. An email address usually contains the name. |
| Address, postcode, city | A full address points to a household. A postcode plus a rare job can too. |
| Account, IBAN, tax and ID numbers | Unique to one person or company. |
| Date of birth | An identifier on its own. |
| Notes, comments, "reference" and "description" | Free text. First names, flat numbers and "the lady from the bakery" end up here. |
| Customer or case number | Meaningless to the AI, but it links the row to your own system. Replace it if the file leaves your team. |
Columns that only describe the work, such as category, status or region, can usually stay. The test is the same as for a document: could someone who sees only the sheet work out who these people are? Combinations count. A row with a rare occupation, a small town and a date can identify a person without a name.
What hides in a workbook
- Hidden sheets. A sheet can be hidden from the tab bar and still hold the full client list. Excel also has a "very hidden" state that the Unhide dialog does not offer: Microsoft's reference says the user cannot make such a sheet visible again from the interface.
- Hidden rows and columns. They are still in the file. Microsoft's Document Inspector lists "hidden rows" and "hidden columns that contain data" as content it finds.
- Comments and notes. They carry reviewers' names and often quote the cell.
- Sheet names and the file name. "Smith family 2025" is a name too.
- Headers and footers, document properties, custom XML data. All of these are on the Document Inspector list as well.
- Formulas. A formula can contain a literal name, for example in a lookup.
- Pictures. A pasted screenshot or a scanned form is not text, and Find does not read it. See personal data in scans and screenshots.
Document Inspector removes this kind of content, but it does not replace names in cells. Microsoft recommends running it on a copy, because removed data may not come back, and Undo may not restore it.
Doing it by hand in Excel
- Work on a copy with a neutral file name.
- Unhide everything first. Show all sheets, rows and columns, so you see what you are cleaning.
- Build the key. On a separate file, write each real value next to its placeholder: Person_1, Company_1 and so on. Keep it outside the chat.
- Replace with Ctrl+H. In Options, set Within to Workbook, so all sheets are covered. Search the longest form of a name first, then the shorter ones. Match case off, so "SMITH" is caught as well.
- Check the free-text columns by eye. Sort or filter the notes column and read it. Replace finds only what you typed.
- Handle comments separately. In the Replace tab Look in offers only Formulas, so comments and notes need deleting or editing one by one. Microsoft's Find and Replace page describes these options.
- Search again for every original name. The count should be zero. Then search for "@".
- Remove the hidden content with Document Inspector, on the copy.
Doing it per client is manageable. A list of two hundred people with different spellings is where it stops working, because every Replace is one name at a time and the key table grows by hand. See placeholders across several documents if the same client appears in several files.
CSV and ODS
CSV is plain text. It cannot hold hidden sheets or comments, so there is less to miss. A free-text field can still contain commas and line breaks inside quotes, and the names in it are as real as in any other column. Open the file in a text editor once and read the first rows before you paste anything.
ODS, the LibreOffice spreadsheet format, has sheets and can hide them, just like XLSX. Check the same places: notes, sheet names and hidden sheets.
Getting real names back in the answer
The AI will answer in placeholders: "Person_14 has not replied since March, and Company_3 is the only one with two open cases." To turn that into something you can act on, you need the key. By hand, that means looking up each placeholder in your table, or running Replace in reverse on the answer.
Two rules keep this working. The same person must have the same placeholder on every row and every sheet, or the answer will treat one person as two. And the key stays with you: do not upload it with the file. The restore workflow covers the round trip.
If you only need a method, not an answer about your own rows, send a sample. Ten rows with placeholders are enough to ask for a formula or a pivot layout, and you apply it to the real file yourself.
With ClientSafe
ClientSafe reads Excel (.xlsx), LibreOffice (.ods) and CSV files, and hidden sheets in XLSX files are read too. You add the file to a client's case. Names, companies, addresses, phone numbers, emails, account and ID numbers and dates of birth become placeholders, and the same value gets the same placeholder across the whole case. You review the matches, untick false hits and mark anything it missed. The safe copy is copied for pasting into the chat, or saved as a Markdown .md file. The documents are processed on your PC.
When the answer comes back, paste it into ClientSafe and the real names are restored.
Frequently asked questions
Can I just delete the name column?
Only if no other column identifies the person. Email addresses, notes and the file name often still do.
Is a CSV safer than an XLSX?
It has fewer hidden places, but the cells themselves are the same. The names inside it still need replacing.
Does Document Inspector anonymize the data?
No. It removes hidden content and properties. It does not replace names in cells.
Does the AI see hidden sheets in an uploaded file?
Assume it can. Replace or remove hidden content before the file leaves your machine.
Should I upload the file or paste the data?
Whichever you choose, send the safe copy. A file you upload brings its file name, sheet names and properties with it.
ClientSafe reads spreadsheets together with Word documents, PDFs, presentations and emails, and keeps one set of placeholders for the whole client.