Cleaning a coaching transcript sounds like a compliance project. In practice it is closer to a short pass with Find and Replace, because a coaching session is mostly your client thinking out loud, not a report full of other people’s details.
Here is what I take out before a transcript goes into any tool, the routine I use to do it, and why I keep the job in my own hands rather than passing it to an AI tool.
Why it is lighter than you might expect
Most of the identifying detail in a coaching transcript sits in a few predictable places. The client’s name tends to appear in the file name and in the opening greeting, and perhaps once or twice when a family member, a colleague or a company comes up. The body of the session is usually about decisions, patterns, emotions and self-reflection. That is the part you want to keep, and it rarely names anyone.
So this is not full anonymisation, and I would not call it that. It is identifier removal: take out the details that point to a specific person, organisation or location, and leave the coaching substance in place.
What counts as an identifying detail
Two kinds are worth watching for. Direct identifiers are the straightforward ones: names of people, company names, locations, email addresses, phone numbers, URLs, and a file name that carries the client’s name. Indirect identifiers are the ones that point to someone through context rather than a name.
| Type | Example |
|---|---|
| Direct | a name, a company, a city, an email address |
| Role-based | ”my co-founder, who used to be CFO at…” |
| Event | ”after the acquisition announcement last Tuesday” |
| Relationship | ”my husband, who also sits on the board” |
| Rare context | ”our 12-person medtech company expanding to Malaysia” |
A single rare detail can identify a client more surely than their first name. That is why a human glance still matters at the end.
The two-minute routine
Rename the file
The file name is the most common leak.
Instead of Sarah_Jones_CEO_AcornHealth_session_2026-06-14.txt, save it as coaching-session-2026-06-14.txt.
Fix the greeting and the close
Names cluster at the start and end. Replace “Hi Sarah, good to see you again” with “Hi [CLIENT], good to see you again”, or remove the greeting lines from the transcript altogether.
Find and Replace the rest
Search for each of these and swap in a label: the client’s first name and surname, the company name, a unique product or service name, an identifying city or country, and the names of colleagues, co-founders, a spouse, investors or board members, along with email addresses, phone numbers and URLs.
| Identifier | Replace with |
|---|---|
| Client name | [CLIENT] |
| Coach name | [COACH] |
| Company name | [COMPANY] |
| Colleague name | [COLLEAGUE] |
| Co-founder name | [CO-FOUNDER] |
| Investor name | [INVESTOR] |
| Spouse or family name | [FAMILY MEMBER] |
| City | [CITY] |
| Country | [COUNTRY] |
[EMAIL] | |
| Phone | [PHONE] |
| Website | [URL] |
Check the first page and the story-heavy parts
Identifiers gather in a few spots: the file name and title, the greeting, the “what has happened since last time” update, descriptions of workplace situations, any concrete example involving other people, and the closing admin. A quick read of those is usually enough.
Upload only the cleaned text
Upload the cleaned transcript, not the audio, not the original, and not a file with the client’s name in the title.
Keep the coaching substance
It is possible to take out too much. A useful read of the session needs the real language: the client’s thinking and dilemma, the emotional words, the metaphor, the behaviour pattern, the decision context, and your own questions. Generalise the labels, keep the meaning.
Original:
“When we opened the Singapore office, I realised I was still trying to run everything myself.”
Too flat:
“When we opened the office, I realised I was trying to run everything myself.”
Light redaction:
“When we opened the [REGIONAL OFFICE], I realised I was still trying to run everything myself.”
The light version protects identity and keeps the leadership context that makes the moment worth analysing.
Why I do not hand this to an AI tool
It is tempting to paste the transcript into a general AI tool and ask it to anonymise the text. I step around that, for a few reasons.
The first becomes plain in hindsight: to anonymise the original, you have to send the original, identifiable transcript to that tool. The exposure you wanted to avoid has already happened. As an example, OpenAI says API data is not used to train its models unless you opt in, yet abuse-monitoring logs may hold content for up to 30 days. “Not used for training” is a different question from “not processed” or “not stored”.
The second reason is that AI redaction misses things, the indirect identifiers above in particular. The third is that a general anonymiser can over-flatten the text and strip the coaching evidence along with the names. A manual Find and Replace is often faster, and it keeps the judgement with you.
A word on “anonymise”
I am careful with the word. Replacing “Sarah” with [CLIENT] reduces exposure, and it is good practice, though it does not make the transcript legally anonymous if you can still link it back to your client. The UK ICO and the European Data Protection Board both treat that kind of pseudonymised text as personal data. So the honest description is identifier removal and light redaction, kept proportionate, rather than a promise of full anonymity.
Where this fits with CoachFlow
This is the step that sits between your session and any tool you let read it, CoachFlow included. CoachFlow starts after the session, on the text you have chosen to use, and it is built so that names and identifying detail are not displayed back to you. Your part is the quick clean-up. The tool’s part is the read.
If you want to see what a cleaned transcript turns into, start a 7-day free trial at coachflow.space. You are not charged when you sign up, and you can cancel during the trial at no cost.
A simple rule keeps this light: when you are unsure whether a detail would identify someone, take it out. The session is about your client’s thinking, and that is the part worth keeping.