Q: What is the TRAPP Framework? A: TRAPP is a simple decision-making framework from the PHCC Academy and Health Resources in Action that helps public health teams use AI with intention. This AI in public health communication guide lays out five plain-language checks—Task, Risk, Accuracy, Privacy, and People—to support thoughtful choices before, during, and after using a model. Q: What are the five checks in the TRAPP Framework and what do they assess? A: The five checks are Task, Risk, Accuracy, Privacy, and People. They help teams choose appropriate tasks for AI, assess potential harms and trust impacts, verify sources and facts, protect sensitive data and follow laws and organizational rules, and ensure human review and responsibility. Q: When is it appropriate to use AI according to this guide? A: AI is most useful for low-risk, repetitive, or early-stage work such as brainstorming message angles, drafting plain-language summaries of approved content, creating audience-specific versions, outlining social copy, emails, and FAQs, and adjusting reading level or tone. The guide advises caution for medical or clinical advice, sensitive topics that affect stigma or rights, and any content that requires new, unverified facts. Q: How should teams verify the accuracy of AI outputs? A: Feed the model approved source text when possible, ask the model for citations, and manually check those references. Cross-check key facts with official data or your organization’s guidance before publishing. Q: What privacy precautions does the TRAPP Framework recommend? A: Do not paste personal health information or restricted data into public tools and use privacy-safe, organization-approved platforms. Mask, aggregate, or exclude private details, follow applicable laws and organizational rules, and check language for bias, stigma, and readability across audience groups. Q: Who should take responsibility for AI-generated content? A: Assign a reviewer and an approver for every AI output and keep humans accountable for final judgment. Document prompts, edits, and sources, credit human authorship, and record when content was last reviewed. Q: What common pitfalls should teams watch for and how can they be fixed? A: Common pitfalls include letting AI invent facts, sharing sensitive data, using one-size-fits-all language, over-automation, and unclear ownership. Fixes include providing source text and verifying claims, masking or excluding private details and using secure tools, testing for readability and involving community reviewers, keeping human review for tone and harm risk, and defining drafter, reviewer, and approver roles. Q: Which metrics should teams track to measure the impact of using AI under TRAPP? A: Track quality with factual accuracy rate, correction count, and time-to-publish; equity with readability scores and community feedback; trust with complaints, misinformation corrections, and engagement sentiment; and efficiency with staff hours saved on drafting and versioning. Use periodic audits of AI outputs and tools and share lessons learned across programs to strengthen consistency and oversight.