Best practices for AI use in CME

CME provider organizations remain responsible for working with faculty to manage any implications of AI use in the development of activities for CME credit.

The AAFP Credit System recognizes that artificial intelligence (AI) is increasingly used by CME provider organizations. To support responsible innovation while protecting educational integrity, the following best practices are provided for CME providers that choose to use AI in the development of CME activities. These best practices are offered to support consistency and shared understanding and are intended to complement, not replace, existing Credit System requirements and policies.

1. Acknowledgement of Use

CME provider organizations are encouraged to disclose to learners when AI is used to generate or meaningfully shape the development of educational content. At a minimum, disclosure is encouraged to indicate whether and how AI was used (e.g., generating educational content, creating case examples, idea or image generation, polishing of text, summarizing clinical guidelines, etc.). Disclosure is also encouraged to include a general statement indicating that AI-assisted content was reviewed by a qualified human reviewer to support accuracy and educational integrity.

2. Human Oversight and Validation

AI is intended to support, not replace, human judgment. As with all CME activities, CME providers remain accountable for the accuracy, balance, and scientific integrity of educational content. When AI is used, CME provider organizations are encouraged to be mindful of known limitations of AI, such as inaccurate or fabricated outputs. While some CME provider organizations may choose to use additional AI tools to assist with reviewing AI generated content, such tools are best viewed as supplemental rather than a substitute for human review and accountability.

3. Independence, Bias, and Data Awareness

CME providers are encouraged to be attentive to whether AI-assisted content may introduce bias, promotional language, or undue influence. CME providers are also encouraged to be mindful of data privacy and avoid entering sensitive or proprietary information into unsecured AI tools. When AI is used to support educational content development, adhering to a consistent internal approach that considers potential risks, bias, and limitations can help support thoughtful implementation. CME provider organizations are also encouraged to remain attentive to potential intellectual property considerations when using AI-generated text, images, audio, and/or videos.

4. Appropriate Use Based on Risk

AI is most appropriate for lower-risk, efficiency enhancing tasks (e.g., organizing content, idea generation, summarizing evaluations, refining non-clinical learning objectives). Higher risk uses, such as generating clinical content, should be approached thoughtfully and accompanied by appropriate human review.

5.Documentation and Review

CME provider organizations may find it helpful to be prepared to describe how AI was used and how AI-assisted content was reviewed, if questions arise during the CME credit application review and/or activity audit process. This guidance does not introduce new requirements but reflects existing expectations related to educational integrity and accountability.

6. Ongoing Education and Engagement

CME provider organizations are encouraged to stay informed about evolving uses and risks and to actively engage with the CPD community.