As early as 2020, I began exploring a growing challenge in editorial publishing: how to increase ROI on images while improving SEO and accessibility at scale. Medical content relies heavily on visuals, but across thousands of articles, one recurring bottleneck was consistent image captions and alt text — especially when speed, accuracy, and accessibility standards all needed to be met.

In early 2023, I helped initiate an internal AI experiment focused on image-to-text automation. The idea was simple but high-impact: upload original photography into an early large language model (initially using ChatGPT-style image reading), have the tool interpret what’s in the image, then generate a caption and alt text aligned with editorial tone and accessibility best practices. This approach could support SEO performance while reducing manual strain on editors and copy teams.

At the time, we quickly faced a major industry barrier: copyright and licensing restrictions. Many organizations didn’t have the rights needed to upload broad libraries of stock content into large language models, which limited what was possible. To stay compliant, we focused only on original photography and legally approved assets — proving that innovation could still move forward responsibly.

I presented the concept to senior stakeholders, including creative leadership and our internal AI committee, and the initiative gained momentum. While the program later expanded beyond my scope and was elevated through engineering for broader implementation, I’m proud to have been part of the initial strategy conversations and early testing — helping shape a scalable direction for AI-assisted metadata, accessibility, and editorial efficiency.