NISO Opens ARM Accessibility Standard for Comment

Two universities fix the same inaccessible textbook in the same semester. Neither knows the other did it. Both pay for the work, both pay for the expertise, and the student at the third university is still waiting.

That duplication is not a failure of effort. It is a failure of description. There has never been an agreed way to record what was fixed in a remediated document, which means there has never been a reliable way to find out whether someone already fixed it.

NISO is proposing one. Its draft Accessibility Remediation Metadata standard, known as ARM, opened for public comment on 24 September 2026 and closes on 9 November.

What is the NISO ARM standard?

ARM is a standardized vocabulary for describing what was done to make a document accessible, and what the remediated version actually supports. It covers the technical detail of the remediation alongside the bibliographic information, in a form that both people and software can read.

It is a draft. Nothing is required yet, and nothing will be until the public comment period closes and the working group takes the result through the full ANSI and NISO standardization process.

Why does remediation get duplicated?

Duplication happens because there is no framework for sharing it. NISO’s own framing of the problem is blunt: most published material is not fully accessible, educational institutions and government bodies are legally required to provide accessible versions, so remediation happens, and it happens repeatedly at individual institutions because nobody can easily tell what anyone else has already done. Over and over again.

A library that remediates a title has produced something genuinely valuable. Without a common way to describe it, that value stops at the institutional boundary.

What problem is ARM actually solving?

Not accessibility itself. Findability.

The distinction matters, and it comes through most clearly in an example from Bill Kasdorf, a consultant and co-chair of the ARM Working Group. A blind reader may need mathematical content encoded in MathML and long descriptions for complex images. A reader with dyslexia may need a reflowable EPUB that supports font customization. Those are different requirements, met by different work, producing different files.

A label saying “this document is accessible” is useless to both of them. It does not say whether the maths is encoded, whether the images are described, or whether the text reflows.

ARM exists so that those features can be described in detail, which makes it possible for a person or an institution to find a version that meets their specific requirement rather than a general one. In a university library or a classroom, that is a daily problem.

Where did ARM come from?

It did not start from scratch. ARM builds on the FRAME project and the platform it produced: Educational Materials Made Accessible, or EMMA for short. Both were funded by the Andrew W. Mellon Foundation.

FRAME set out to solve exactly the problem described above. Reduce redundant remediation work, encourage institutions to share what they had already fixed. To do that, it had to build a metadata model for describing the remediation process in detail, because sharing is impossible without description.

EMMA’s model worked for EMMA. The ARM Working Group was formed to extend it, refine it, and standardize it so it works for the broader accessibility community rather than one platform.

That lineage matters for two reasons. The model has already been tested against real material rather than designed in the abstract. And ARM is an extension of something that exists, not a proposal for something new.

Who is behind it?

The working group is co-chaired by three people, and the mix tells you something about the intended scope.

• Bill Kasdorf, a consultant with a long background in publishing standards.
• Jacob Jett, Research Librarian at the Association of State Floodplain Managers.
• J. Stephen Downie, Professor and Executive Associate Dean at the School of Information Sciences at the University of Illinois Urbana-Champaign.

Publishing, libraries, information science. Remediation metadata has to travel between the people who produce documents and the people who have to make them usable, and that is precisely the gap these three sit across.

What would change if ARM is adopted?

Three things, in rough order of how soon they would show up.

1. Remediation becomes describable in a consistent way. Jett’s framing is that ARM captures bibliographic and technical information in a standardized vocabulary that is both human-readable and machine-processable. The second half carries the weight. A description a system can parse is a description that can be indexed, searched, and exchanged without anyone doing it by hand.
2. Searching by requirement becomes possible. If the metadata records that a file has MathML and long image descriptions, a reader who needs those can look for exactly that.
3. Sharing results becomes practical. Jett describes ARM as laying a foundation for sharing the results of remediation efforts. That is the duplication problem from the top of this piece, approached from the description end rather than the platform end.

None of this happens automatically. A standard only works if enough people implement it, and adoption is the open question the working group itself keeps returning to.

What happens next?

The public comment period runs through 9 November 2026. The draft sits on the ARM project page, and NISO has said it will work with the group afterwards to evaluate and integrate community feedback before completing the full ANSI and NISO standardization process.

Stephen Downie’s comment on the draft is the practical one: additional feedback will help ensure ARM meets a broad range of needs, which in turn will support wider adoption once the standard is finalized.

That is the argument for commenting rather than waiting. A metadata standard gets shaped by the people who tell the working group what their material actually looks like and what their users actually need. Institutions remediating at volume, and the people relying on the results, hold specific knowledge the draft can still absorb.

After 9 November, the shaping is done. The only question left is whether to implement.