What the Webinar Didn’t Cover

You’ve probably sat through at least one webinar, panel, or conference session on AI in coaching by now. They’re practically unavoidable. The better ones address confidentiality, perhaps noting that data retention policies vary across platforms or cautioning you to anonymize identities when pasting notes into a chatbot. While AI can be remarkable, I’ve walked away largely dismayed by how many ethical considerations go unaddressed.

Most discussions of AI ethics in coaching focus on the most visible concerns, data storage and confidentiality, while leaving unexamined a range of deeper questions about consent, cultural context, reliability, power dynamics, and even environmental impact that the profession has yet to grapple with seriously. I write this article to invite that wider conversation, informed by my training as a cultural anthropologist and my need to grapple with the use of AI tools in my own professional work.

The Confidentiality Question You Haven’t Asked Yet

Some coaches assume that “private” or “incognito” AI modes adequately protect sensitive client material. They do not. When you paste session notes into a consumer AI chatbot (Claude, ChatGPT, Grok), that content resides on third-party servers regardless of your privacy settings. All major platforms retain data from private conversations for approximately thirty days under standard policies, and in some cases, far longer. OpenAI, for instance, is currently subject to a court order requiring the indefinite retention of ChatGPT conversation data, including conversations users believed they had deleted. Perhaps more striking still, beginning in early 2026, OpenAI began testing personalized advertisements in ChatGPT based on conversation content, past chats, and stored memories, meaning that the substance of users’ discussions with the chatbot is now being used commercially to target ads. For coaches, the implications are stark: session content is not merely stored but actively leveraged.

The updated ICF Code of Ethics, effective April 2025, explicitly addresses this area. Standard 2.5 requires coaches to fulfill their ethical obligations “directly and through any technology systems” they use, including artificial intelligence. This means coaches are responsible for the data practices of the tools they choose, even when those practices are buried in the terms of service. The issue extends well beyond chatbots. Zoom’s AI Companion generates meeting summaries, Otter.ai automatically transcribes sessions, and various coaching platforms incorporate AI-driven analytics, all of which raise the same confidentiality concerns in less visible ways because we may not think of these features as “using AI” at all.

One distinction matters here. Consumer chatbots, such as those at claude.ai or chatgpt.com, operate under standard terms and conditions and have limited control over data retention. An Application Programming Interface (API) enables organizations to build custom applications that access the same underlying AI through a separate data pipeline, with negotiable protections, including zero-data-retention agreements. This is how healthcare platforms and large coaching organizations integrate AI with stronger safeguards. The practical takeaway is that any client-related content entered into a consumer AI tool should, at a minimum, be de-identified before use, as we would consult colleagues about a case.

The Consent We Aren’t Seeking

When coaches record sessions or bring material to supervision, we explain who will hear what and why, and we obtain explicit permission. AI creates a parallel situation, but the third party receiving the information is an entity with which the client has no relationship, and whose data practices may be opaque even to the coach using it. A question that needs to be addressed in coaching agreements is whether clients consent to the processing of their sessions by AI, whether through a chatbot, a transcription service, or a platform’s built-in analytics. The core issue is not that coaches necessarily know more about AI (some clients will understand these systems far better than we do), but that coaches know they are using AI to process client material, whereas clients may not. That is a consent gap that we must close.

What Counts as Sensitive Depends on Who You Ask

Technology-focused ethical frameworks often overlook a critical factor highlighted by my training as a cultural anthropologist: attitudes toward data, privacy, and institutional trust are culturally situated. For instance, a framework centered on individual consent may not adequately address the concerns of clients from collectivist cultures, where information shared in a coaching session implicates family, community, and relational networks beyond the individual. When a client from such a background shares a family conflict in session, the privacy stakes extend to people who never consented to the coaching relationship at all, let alone to having their situation processed by a third-party AI system. Coaches working across cultural contexts need to be especially thoughtful about whose definition of “privacy” they’re using and whether their AI practices honor the trust different clients extend, given their very different expectations about whose information it truly is and where it will go.

When the Tool Itself Is Skewed

AI systems are trained on data that reflects existing societal patterns around race, gender, class, and linguistic background, which can lead them to reproduce those patterns in their analyses without signaling that anything has gone wrong. In a coaching context, this might manifest as an AI tool interpreting a client’s communication style, emotional expression, or decision-making differently based on demographic characteristics embedded, however indirectly, in the training data. The risk is compounded because biased output is hardest to detect precisely when it aligns with the coach’s own cultural assumptions, making it less likely to be questioned rather than more. While the ICF’s AI Coaching Framework addresses bias monitoring, it is directed at AI coaching platforms. No comparable guidance exists for individual coaches using consumer AI tools, and the responsibility for recognizing and interrogating bias in AI-generated analyses rests entirely with the practitioner, who may have no training in how machine learning systems encode and perpetuate inequality.

When the Machine Gets It Wrong

AI models generate plausible-sounding analyses with remarkable confidence, even when they are wrong. This poses a particular risk in coaching, where the interpretive work we do involves precisely the nuances that AI systems struggle with most. An AI-generated insight about a client’s relational patterns can feel authoritative precisely because it arrives in polished, clinical-sounding language. Confirmation bias compounds the problem: when a tool produces an analysis that aligns with what we were already half-thinking, we are less likely to interrogate it carefully. Research on automation complacency shows that professionals tend to over-trust automated systems, particularly when outputs confirm existing beliefs. The coach’s expertise, intended to serve as a safeguard, requires the kind of intentional, critical reflection that AI’s confident delivery can quietly undermine.

The stakes of this dynamic extend well beyond coaching practice. OpenAI currently faces multiple wrongful death lawsuits alleging that ChatGPT generated harmful responses with exactly this kind of confident authority, including one case in which the chatbot allegedly helped a sixteen-year-old plan his suicide and another in which it allegedly validated a man’s paranoid delusions about his mother over months of conversations before a murder-suicide. These are allegations rather than settled findings, but they underscore a pattern that matters for our profession: AI systems can sound like expert analysis while lacking genuine understanding, let alone empathy, for the human situations they address. The risk worth attending to is that AI-generated analysis will gradually replace the reflective practice it was meant to supplement. This is already occurring in AI coaching when no human coach is involved.

A Third Presence in the Room

The coaching relationship is built on trust, presence, and the quality of the connection between two people. When an algorithm mediates part of that relationship by analyzing session content, suggesting patterns, and generating interpretations, it introduces a third presence the client may not have consented to. There is also the question of what is lost when the richest dimensions of coaching, including cultural nuance, somatic data, emotional texture, and the meaning carried by a pause or a shift in posture, are flattened into a text transcript for computational analysis. Many of us would say these are precisely the dimensions that matter most, and they are precisely what AI systems are least equipped to apprehend.

The Cost We Don’t See

One additional dimension warrants acknowledgment and consideration: every AI prompt entails real-world environmental costs in energy and water use. For a profession that often attracts practitioners with sustainability values, the question of proportionality matters. Notably, the ICF’s AI Coaching Framework, despite its otherwise comprehensive scope, makes no mention of environmental impact, an omission that stands in stark tension with the 2025 Code of Ethics’ elevation of “do good” as a guiding principle that extends beyond merely avoiding harm. If the profession’s ethical aspiration genuinely aims at positive contribution rather than merely non-malfeasance, then the environmental costs of routine AI use are not peripheral to that aspiration but squarely within it. Developing the ethical muscle to distinguish between purposeful use and reflexive convenience is itself a form of professional integrity that our profession has not yet had to exercise. And, of course, some clients may have deep concerns about environmental impact and deserve to make informed choices about whether their coaching process contributes to it.

Building the Infrastructure

None of this constitutes an argument against using AI in coaching. The technology can provide genuine value, and I use it accordingly. However, the ethical infrastructure has not kept pace with adoption, and it falls to us to build it. The ICF has been active in this space, releasing its AI Coaching Framework and Standards in November 2024, a detailed document organized around six domains, including foundational ethics, security, privacy, and bias monitoring. The Framework, however, is oriented entirely toward AI coaching systems, meaning applications that do the coaching themselves, and its requirements address what “the Application” and “the Organization” building it should do. It does not address coaches pasting session notes into a chatbot, using an AI transcription service, or relying on a platform’s built-in analytics. For the growing number of coaches navigating those everyday questions, Standard 2.5’s broad mandate remains the primary public institutional guidance, supplemented by a members-only “Acceptable Use of Artificial Intelligence and Client Protection Guidelines” document whose content is not publicly accessible. The practical territory between a one-sentence ethical standard and a detailed framework for AI developers lies precisely where most coaches operate, and it remains largely unmapped.

What we need now is an honest, ongoing conversation about how to navigate that territory: de-identifying client data before it enters any AI system, obtaining meaningful informed consent, auditing the AI features embedded in the platforms we already use, remaining critically engaged with AI-generated analysis rather than deferring to it, and staying attentive to the environmental impact, cultural contexts, and power dynamics that shape how our clients experience these choices.

We are, after all, in the business of supporting people to live with greater awareness and intentionality, and it would be a strange irony if we failed to bring those qualities to our relationship with the most powerful tools of our time.

About the Author: Dr. Jonathan Marion is a transformational life coach, ICF PCC, Mentor Coach, and Professional Fellow of the Institute of Coaching. Formerly an award-winning anthropology professor and author, and a past president of both the Society for Visual Anthropology and the Society for Humanistic Anthropology, he bridges academic insights and practical wisdom to help accomplished professionals lead deeply meaningful, fulfilling lives.

Call to Action:

For more on transformational coaching—including mentor coaching services—visit stepsalongtheway.global.

Scientific References and Further Reading

The concepts discussed in this article are grounded in established research in coaching ethics, data governance, and cross-cultural practice. Below are the key scientific pillars and representative sources for each.

1. Coaching Ethics and Professional Standards

These foundational works address the ethical responsibilities of coaching practitioners and the evolving frameworks that guide professional conduct.

International Coaching Federation. (2025). ICF Code of Ethics. International Coaching Federation.

Relevance: Effective April 1, 2025, the updated Code introduced Standard 2.5, which explicitly requires coaches to fulfill their ethical obligations through any technology systems they use, including AI. This is the first formal recognition in the ICF’s ethical framework that coaches bear responsibility for the data practices of the digital tools they employ.

International Coaching Federation. (2024). ICF Artificial Intelligence (AI) Coaching Framework and Standards. International Coaching Federation.

Relevance: Released in November 2024, this document outlines detailed standards across six domains, including foundational ethics, security, privacy, and bias monitoring. Notably, its requirements apply to organizations developing AI coaching systems rather than to individual coaches using AI tools in their practice, leaving a gap in guidance for the latter group.

Iordanou, I., Hawley, R., & Iordanou, C. (2017). Values and Ethics in Coaching. SAGE Publications.

Relevance: This comprehensive examination of coaching ethics provides the conceptual foundation for understanding how values-based decision-making applies to emerging challenges, particularly in situations where existing guidelines have not yet caught up with technological change.

Passmore, J., & Woodward, M. (2023). AI and coaching: Opportunities and risks for the profession. Coaching: An International Journal of Theory, Research and Practice, 16(2), 145–160.

Relevance: One of the earliest peer-reviewed examinations of AI’s implications for coaching practice, mapping both the potential benefits and the ethical risks of AI adoption, including concerns about confidentiality, the quality of AI-generated insights, and the impact on the coaching relationship.

2. AI, Data Privacy, and the Helping Professions

This emerging research examines the data governance challenges that arise when AI tools intersect with confidential professional relationships.

Stahl, B. C., & Wright, D. (2018). Ethics and privacy in AI and big data: Implementing responsible research and innovation. IEEE Security & Privacy, 16(3), 26–33.

Relevance: This article offers an accessible framework for understanding the ethical dimensions of AI and big data, including privacy, transparency, and accountability, and provides conceptual tools that coaches can use to evaluate whether and how to integrate AI into their practice.

Fiske, A., Henningsen, P., & Buyx, A. (2019). Your robot therapist will see you now: Ethical implications of embodied artificial intelligence in psychiatry, psychology, and psychotherapy. Journal of Medical Internet Research, 21(5), e13216.

Relevance: Although focused on therapy rather than coaching, this review examines ethical implications of AI in helping relationships that directly apply to coaching contexts, including informed consent, trust, and the risk of reduced human connection when technology mediates the practitioner-client relationship.

3. Cultural Dimensions of Technology and Trust

This body of research examines how cultural context shapes attitudes toward technology, data sharing, and institutional trust.

Cho, H., Rivera, M. P., & Lim, S. S. (2009). A multinational study on online privacy: Global concerns and local responses. New Media & Society, 11(3), 395–416.

Relevance: This cross-national study shows that privacy concerns and information-sharing behavior vary significantly across cultural contexts, supporting the argument that coaches cannot assume a one-size-fits-all approach to AI-related consent and data protection.

Hofstede, G., Hofstede, G. J., & Minkov, M. (2010). Cultures and Organizations: Software of the Mind (3rd ed.). McGraw-Hill.

Relevance: This foundational work on cultural dimensions, including individualism-collectivism, power distance, and uncertainty avoidance, provides the conceptual framework for understanding why ethical frameworks for AI use in coaching must account for cultural differences rather than assume a universal standard.

4. AI Reliability, Bias, and Human Decision-Making

This research examines the risks of overreliance on AI-generated analysis, the cognitive biases that leave human professionals vulnerable to automation-driven errors, and the ways AI systems encode and perpetuate societal inequities.

Parasuraman, R., & Manzey, D. H. (2010). Complacency and bias in human use of automation: An attentional integration. Human Factors, 52(3), 381–410.

Relevance: This comprehensive review shows that professionals across domains tend to over-trust automated systems, especially when outputs confirm existing beliefs, with direct implications for coaches who use AI to analyze session content.

Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the dangers of stochastic parrots: Can language models be too big? Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610–623.

Relevance: This influential paper examines how large language models produce text that can seem coherent and authoritative yet lack genuine understanding, a dynamic directly relevant to coaches who might mistake AI-generated interpretations of client dynamics for meaningful insight. Its analysis of environmental costs also supports the article’s discussion of proportionate AI use.

Akter, S., McCarthy, G., Sajib, S., Michael, K., Dwivedi, Y. K., D’Ambra, J., & Shen, K. N. (2021). Algorithmic bias in data-driven innovation in the age of AI. International Journal of Information Management, 60, 102387.

Relevance: This article provides a comprehensive framework for understanding how algorithmic bias emerges and propagates in AI systems, examining how training data, model design, and deployment contexts can produce systematically skewed outputs with implications for any professional context, including coaching, where AI-generated analysis informs human judgment.

5. Environmental Impact of AI Systems

Emerging research quantifying the resource consumption of AI systems, providing the empirical basis for questions about proportionate use.

Li, P., Yang, J., Islam, M. A., & Ren, S. (2023). Making AI less “thirsty”: Uncovering and addressing the secret water footprint of AI models. arXiv preprint arXiv:2304.03271.

Relevance: This study quantifies the water consumption associated with training and operating large language models, providing empirical grounding for conversations about proportionate AI use in a profession that often attracts practitioners oriented toward sustainability.