JSTOR introduces alt text tools for digital image collections

JSTOR Digital Stewardship Services has launched new workflows to support alt text across its digital collections. The release of the alt text tools allows participating institutions to author, edit, export, and publish descriptive image text directly on the platform. Consequently, cultural heritage organisations can now make visual research materials vastly more accessible to users.

The update applies to Tier 2 and Tier 3 participants within the JSTOR Stewardship infrastructure. In addition, Tier 3 charter members can use JSTOR Seeklight to generate draft alt text automatically for single items or large digital backfiles.

Meeting global web content accessibility standards

Providing descriptive alt text is essential for individuals who cannot perceive images visually. It forms a foundational requirement of the Web Content Accessibility Guidelines (WCAG) for non-text digital assets. However, creating meaningful descriptions item by item remains a major operational challenge for archives managing millions of items.

The new suite of tools addresses this issue by embedding accessibility directly into existing institutional workflows. As a result, libraries can scale up their compliance efforts without completely altering their established metadata pipelines.

Keeping expert review at the centre of AI workflows

Although JSTOR Seeklight uses AI to generate initial image descriptions, the tool is designed as a starting point for expert review. Library staff and archivists maintain full control over the final output. The system includes built-in uncertainty indicators to highlight complex items that require human intervention. Furthermore, flags automatically identify images containing people, ensuring that institutional context and professional judgement guide sensitive metadata decisions.

Roger Schonfeld, Managing Director of JSTOR Digital Stewardship Services, emphasised that alt text is vital for discovering visual archives. He explained that while Seeklight began with automated metadata generation, the broader vision encompasses comprehensive accessibility. By combining transcripts for text with alt text for images, the platform helps institutions advance discoverability while keeping expert human oversight at the centre.

Co-designing tools with the information community

The new capabilities were developed through close collaboration with librarians, archivists, and digital collections professionals. These stakeholders provided direct feedback through extensive testing.

Consequently, this initiative continues JSTOR Seeklight’s community-informed approach to AI-assisted processing. By balancing automated speed with expert review, JSTOR provides academic institutions with a practical framework for sustainable open research.