AI Usage Guidelines

AI in written content guidelines

How and when we use AI in content development is an evolving practice as we adapt to new technologies and applications. As we evolve, our guidelines will continue to stem from our “true north” principles for content development: 
  1. Patient trust: Patients need to maintain full trust in our medical authority and expertise. 
  2. Credibility and reputation: We need to uphold the reputation of the institution and mitigate risk through credible content that’s authentic and true to our brand. 
  3. Accuracy: We have a responsibility to provide the most accurate information available to patients and the community. 

Content parameters

Content covered in these guidelines includes but is not limited to: 

  • Patient-facing articles, web content, social media, emails and other communications 
  • Patient-facing marketing materials, advertising, collateral and signage, etc. 
  • External facing publications, electronic communications, press releases 
  • Internal communications – both electronic and print – to faculty, staff and students 

Permitted uses for AI in content

Key Takeaway: Use AI for augmentation of content, but not authorship.  

AI benefits the use of content by assisting with creativity, creating efficiency and improving scale.  

  1. Brainstorming assistance: to generate ideas, concepts, themes, headline ideas, etc. 
  2. Content review: to suggest opportunities for streamlining, summarizing research or other publications, simplification, reading level appropriateness, etc. 
  3. Content framework development: to assist in developing or refining internal plans or documentation (such as communication plans, research abstract summarization, awards submissions, production plans, generating transcripts, briefs/summaries, etc.) 

How much AI augmentation is permitted? 

Sometimes, a sentence suggested by AI is terrific, and there may not be a better rewrite option. But as a general rule of thumb, no more than 20% of AI changes/edits/suggestions should be accepted in any content piece. 

Limited use cases for AI content development 

AI may support content creation in some low-risk environments. Accountability for the accuracy, appropriateness and quality of content remains with the designated/qualified human reviewer prior to implementation. Lower-risk use cases include, but are not limited to:

  • Generate provider bio copy (with human review) 
  • Develop starter copy for Google paid campaigns and paid social marketing campaigns (with human review) 
  • Awards submissions (with human review) 
  • Starting internal memos/announcements/talking points/FAQ documents (with human review) 
  • Transforming background documentation into HUB posts (with human review) 
  • Social media: caption and alt text generation (with human review) 

Review and risk for AI in content 

Because AI carries inherent flaws, human reviews are required for all content, but with varying levels based on risk stratification:  

  • High-risk (patient/clinical/public-facing): multiple human reviewers 
  • Lower-risk (drafts, internal): flexible level, but human review still required 

Non-permitted uses for AI in content

Never use AI in the following ways:

  • Using tools that are not university-approved and have not passed our medical center risk assessment.
  • Creating content that will be attributed to a human being. 
  • Generation of clinical content (e.g., What are the signs of a heart attack? or How prostate cancer is treated) as this information can contain inaccuracies and/or falsifications. 
  • Entering PHI, trade secrets, sensitive internal documents, strategic plans, embargoed releases, or confidential research into any non-approved AI tools. (Note: M365 applications, including Copilot, are approved for the storage of S1-S4 data, including PHI.)

AI-generated image and layout guidelines

AI-generated images, illustrations or layouts (e.g., brochures, flyers, posters, etc.) are not currently approved for patient-facing use due to accuracy, attribution and quality assurance requirements. This includes patient education, marketing and branded materials. However, AI can be utilized for image/video enhancements or for creating/modifying assets and layouts for internal audiences (e.g., internal education materials, presentations, etc.) with care. 

The same standards apply to AI-generated and AI-enhanced images as any original content created on behalf of the organization. AI can be a helpful tool to support creativity and efficiency, but people are responsible for the final result. 

  • Never use AI to change the context or meaning of content (e.g., adding or removing people in an existing photo). 
  • A human must review it before use to ensure it is factually accurate, appropriate in tone, clear and suitable for the audience.
  • The image must align with our brand standards in terms of visual style and avoid misleading information or bias.
  • Any use of images that are the intellectual property of the University, including Brutus, the athletic logo, the Buckeye helmet, etc., cannot be used without trademark and licensing approval. Learn more about Ohio State Intellectual Properties and approval.
  • Consider adding a disclaimer of “Photo Illustration generated by AI,” “Created with Copilot” for transparency.
  • When in doubt or uncertain about semblance to copyrighted materials, choose a more traditional image source from our Digital Asset Management System or email brand@osumc.edu for further assistance.

Marketers and communicators can find additional guidance on the responsible use of generative AI in the University’s AI for Marketing Guidelines.

Last updated: September 3, 2026