Plenty of coaches now use some AI somewhere in their practice. A transcription tool on calls. A chatbot to draft session notes. An app that sends clients reflection prompts between sessions. Often the client has no clear idea any of this is happening.
The International Coaching Federation (ICF) has put out an updated white paper on this, dated October 2026: Coaching + AI: A Partnership to Unleash Potential. Near the end it lists "Principles for Human-Centered Developmental AI," which ICF says apply "wherever AI touches a coaching experience."
Three of those principles turn into practical questions you can settle with a client before the first session. This post walks through each one and ends with a short checklist you can add to your coaching agreement.
What the white paper says, briefly
A bit of context first, all from the ICF document:
- ICF describes three modalities: coach-led (AI doesn't take part in the coaching relationship), hybrid (the coach stays accountable while AI supports parts of the practice), and AI-led (a user talks to an AI system with no coach accountable).
- In the hybrid model, ICF says the coach "is responsible for defining the operating procedures including disclosure, client consent, and oversight of how AI is used."
- The 2025 Code of Ethics already covers technology. Standard I:2.5 asks coaches to meet their ethical and legal obligations "directly and through any technology systems I may utilize," and it names artificial intelligence in that list.
- In July 2026, ICF convened experts in coaching ethics and ethical AI to form an Ethical AI Standards & Governance Initiative working group.
If you use any AI tool in your work, you're probably in the hybrid model. That puts disclosure, consent, and oversight on you.
Principle 1: Transparency
ICF's wording: people "must clearly understand when they are interacting with AI, what the system is designed to do, its limitations, how data is used, and where human oversight exists."
It adds that AI systems "should never imply human consciousness, empathy, wisdom, or professional accountability where none exists."
What this looks like at session start
Tell the client, in plain words, which tools you use and for what. Not a vague "I may use technology." Something closer to:
"I record our sessions with a transcription tool so I can write accurate notes. I sometimes use an AI assistant to help summarise those notes. I read and edit everything myself. Here's where the recordings are kept and when they're deleted."
That example is a template, not a quote from anyone. Change it to match what you actually do.
Cover five things, because those are the five ICF names:
- When the client is dealing with AI, not you.
- What each tool is for.
- What it can't do.
- How their data is used and stored.
- Where you step in and check the output.
If you send clients AI-generated prompts or messages between sessions, be clear those come from a tool. Don't let a chatbot sign off as if it were you.
Principle 2: Psychological Safety
ICF says developmental AI systems should be designed "to avoid manipulation, emotional dependency, exploitative persuasion, unsafe attachment dynamics, and harmful reinforcement loops."
It also names specific risks: "anthropomorphic attachment, sycophancy, authority inflation, and emotional over-reliance."
Sycophancy is the one many coaches will recognise. ICF puts it bluntly elsewhere in the paper: general LLMs "are designed to deliver a pleasing output to the user, which is contrary to the challenge necessary for personal growth."
What this looks like at session start
- If you give clients an AI tool to use between sessions, explain its role. It's for reflection or reminders, not a stand-in for you.
- Agree on how much to use it. A client messaging a bot at midnight every night is a sign to check in, not a success metric.
- Ask clients to bring what came up in the tool to your next session, so you can talk it through together.
- Tell them plainly that the tool may agree with them too easily. Real coaching includes challenge, and that part comes from you.
Principle 3: Human Accountability by Design
This is the strongest line in the document. ICF writes that human governance of developmental AI "is not a corrective fallback for when something goes wrong," but "a structural requirement from the outset."
It goes on: coaches and organisations using AI "bear professional responsibility for the tools they use: their outputs, their limitations, and their potential to cause harm." And: "Responsibility for developmental outcomes cannot be delegated to an algorithm."
ICF also says systems in emotionally sensitive contexts "must include built-in pathways to human support, ethical review, mental health escalation, or professional intervention."
What this looks like at session start
- Make it clear that you own your notes, summaries, and any advice, whether or not a tool helped draft them.
- Tell the client how to reach you, a real person, if something a tool says worries them.
- Have a referral path ready for when a client needs more than coaching, such as a therapist or mental health service. Write it into the agreement so it is there when you need it.
- Check any AI output before it reaches the client. If you didn't read it, don't send it.
The checklist
Add these lines to your coaching agreement or your first-session script. Adjust the wording to your own practice.
Transparency
- I've listed every AI tool I use in this engagement and what it does
- I've explained what each tool can't do
- I've said where recordings, notes, and transcripts are stored, and for how long
- I've explained when and how I review AI output
- The client knows when they're talking to a tool and when they're talking to me
Psychological safety
- If the client uses an AI tool between sessions, we've agreed on its role and limits
- I've told them the tool may agree too easily, and that challenge comes from our sessions
- We've agreed they'll bring anything important from the tool back to our sessions
Human accountability
- The client knows I'm responsible for anything I send them, AI-assisted or not
- They have a way to reach me directly with concerns
- I have a referral path for needs beyond coaching
- The client has said yes to the AI use described, and can opt out
Consent
- The client can say no to any tool and still get coaching from me
A note on opting out
The last item matters. Nowhere does ICF say clients must accept AI tools. Its whole framing is that the coach-client relationship comes first. If a client doesn't want sessions recorded or notes run through an AI tool, the simplest response is to work without it for that client.
Where to read more
The full white paper is short and worth reading yourself: ICF, Coaching + AI: A Partnership to Unleash Potential (October 2026). ICF also has a resource page for it.
None of this needs new software or a lawyer. It needs a few honest sentences at the start of the engagement and the habit of checking what your tools produce before a client sees it.