Ask the accessibility expert (en): Pilot projects for accessible publishing

The experts

Anke Kersten is an accessible image specialist at Dedicon. She had an active part in the APACE pilot project in the Netherlands and is the author of the end report on all APACE pilot projects. Julija Skerniškytė is an inclusive publishing training coordinator at the Lithuanian Audiosensory Library (LAB) where she coordinates and implements the library's training programs and events, raising awareness about accessible e-books and how to read them. Julija was involved in LAB’s pilot project with Vilnius University Press. 

Introduction

Elisa Molinari from Fondazione LIA does the introduction about APACE and since this is the final event of the project, she also says thank you to the European Commission and the Creative Youth Program for funding this project and allowing the project partners to work for two years on accessible publishing and building bridges between publishers and libraries. She is also grateful for all the experts, the host and the moderators who did a wonderful job in past editions of these Aks the Expert sessions, providing guidance and expertise and sharing their knowledge. Finally, she is thanking all the partners in the APACE project: the Lithuanian Audiosensory Library, Celia, Dedicon, the Italian Publishers Association, the Bulgarian Publishers Association, and the German Booksellers and Publishers Association, and Fondazione LIA.  

Questions and answers

What were these six pilot projects about?

Anke: The six pilot projects all followed a similar approach, focusing on the content of complex publications rather than on metadata or distribution. Each pilot began with contacting a publisher known for producing complex e-books, as these were more likely to contain accessibility issues.

If the publisher agreed to participate, a kick-off meeting was held to discuss expectations, roles, and timelines. Publishers were then asked to provide a complex e-book, after which the accessibility expert and the publisher jointly selected a small number of representative, highly complex samples, such as content involving formulas, music, extensive tables, complex images, multimedia, or foreign languages, because fully remediating an entire book would require too much time.

Each publication then underwent an initial assessment to determine its conformance to international standards such as WCAG and EPUB Accessibility 1.1, and the results were documented in a report that was reviewed together with the publisher. The expert supported the remediation process by helping the publisher address the most important issues and by offering training and guidance.

During the project, the expert also conducted an in-depth interview with representatives of the publishing house to understand their workflow, production processes, tools, and desired future state, covering aspects such as strategy, content creation, production workflows, quality assurance, and distribution.

After the remediation work and interviews were completed, a final evaluation meeting was held to reflect on the outcomes, the collaboration between the publisher and the expert, and possible next steps for improving accessibility in the publisher’s future work.

Did each individual project have a specific focus?

Anke: The focus of each pilot project was determined by the issues identified during the accessibility checks by the expert. After reviewing the assessment reports, the teams concentrated on the areas where publishers encountered the most difficulties. The most common problems included non-accessible non-text content, such as images or videos lacking alternative text, poor visual contrast, or graphics that relied solely on colour. Many publications also showed weak semantic structure and incorrect reading order, including mistagged headings, tables used for layout rather than data, and unmarked foreign-language passages. Another issue concerned complex mathematical and chemical formulas, which were often inserted as images or incorrectly tagged, making them inaccessible for screen readers and braille devices. These recurring challenges guided the priorities and focus areas of the various pilot projects.

How was the experience for Vilnius University Press and LAB? Was it insightful both organisations?

Julija: In this pilot, the LAB-team collaborated for the first time with Vilnius University Press, which also participated in the SPIRAL project as the publisher producing an accessible book. Unlike other pilots, this project involved remediating the entire book rather than selected samples.

The chosen publication, an already published but non-accessible EPUB on disinformation, contained several complex elements such as foreign-language passages, images, and graphs. By the end of the pilot, it had been fully converted into an accessible EPUB.

For the publisher, this was their first real attempt at creating an accessible EPUB; until then, they mainly produced PDFs or non-accessible EPUBs lacking features like alternative text, language markup, or pagination. Accessibility as a concept was still very new to them, they learned mostly through earlier library-led awareness activities. The process proved both challenging and eye-opening.

A major issue was the staff’s limited knowledge and skills, especially among layout designers, who worried about not having the programming-level expertise they assumed was required, with older staff showing resistance to changing established workflows. However, the pilot also helped them understand which tools are needed, how workflows shift when accessibility is incorporated, and what it means in practice to add accessible features.

From the perspective of the library and accessibility experts, the experience showed the value of letting publishers experiment hands-on rather than only learning in theory. Ultimately, with the right support, Vilnius University Press succeeded in producing its first fully accessible EPUB, an important achievement.

Was there any hesitation on the publisher’s side to participate?

Julija: The kickoff meeting with the university press began positively, with the administration showing enthusiasm about participating. However, attitudes shifted once the chosen book underwent an accessibility check and significant errors were revealed. When the team later met with the layout designers, their initial enthusiasm faded as they realized how much work accessibility remediation would require and that it was more difficult than expected.

The experience highlighted an important lesson: attitudes toward accessibility can differ greatly within a publishing house. Management may be eager to embrace new opportunities, while the staff responsible for the hands-on work may feel far less enthusiastic.

What are some best practices regarding image description that came out of the pilot projects? 

Anke: Across multiple pilot projects, accessibility assessments repeatedly revealed that publishers struggled with images. While the lack of alternative text was a common issue, an often-overlooked problem was the accessibility of the images themselves. In data visualizations, such as graphs and diagrams, errors frequently occurred, including reliance on colour alone to convey information, insufficient contrast between adjacent colours, and poor text–background contrast. One pilot therefore focused specifically on improving accessible data visualizations in academic content, checking all visual elements for correct colour use, contrast, labeling, and distinguishable structures. A checklist for accessible data visualizations was created during the pilot project and this has been made available through the European Accessibility Directory.

Beyond visual accessibility, informative images must also include meaningful text alternatives, but writing high-quality descriptions is time-consuming and difficult for many publishers. Accessibility experts can provide valuable training and guidance, as demonstrated in the pilot with Vilnius University Press. Another pilot explored the use of AI to generate alternative text: images were extracted together with their surrounding context, enabling generative AI to produce descriptions that could support and accelerate accessibility workflows, particularly in educational publishing. However, the quality of AI-generated alt text depends heavily on the contextual information provided, and a human-in-the-loop remains essential to ensure accuracy, quality, and educational appropriateness. Overall, the pilot showed that with proper tools, context, and expert or AI-assisted support, publishers can significantly improve the accessibility of images in their publications.

Raising the bar for accessibility also raises the quality of the publication. Was that something the publishers were particularly attentive of? And were there any other best practices that came to light that generally raised the quality of the digital publication for those publishers?

Anke: A major insight from the pilot projects was that improvements made for accessibility, often introduced with blind or low-vision readers in mind, ended up benefiting all readers. Publishers saw clearly that features such as more accessible data visualizations improved the overall quality of the publication for everyone.

Another general best practice that emerged is the value of having an accessibility expert assess a publication, even when publishers believe their content is already accessible. In every pilot, expert assessments revealed issues that publishers had not noticed themselves. Such evaluations, which are also available outside the pilot context, help publishers understand the real accessibility level of their content and provide guidance on how to resolve remaining issues, making publications even more accessible in the end.

What key lessons did LAB learn from their pilot project? 

Julija: A key lesson was that publishers must first understand why accessibility matters and how it improves the reading experience before they can meaningfully engage in the technical aspects. Internal motivation and attitudes proved just as important, if not more important. than technical skills. Many staff members were hearing about accessibility for the first time and needed support to understand its purpose and impact. Without this understanding, resistance to changing established workflows was stronger.

The project also showed the importance of practical, hands-on support: the University Press emphasized that without the library’s training, guidance, and detailed methodological materials, creating an accessible EPUB would have been far more difficult.

The most lasting change was the publisher’s adoption of alt text for non-text content, demonstrating the effectiveness of the provided training.

Another major insight was that accessibility requires more time, planning, and resources than initially expected. Even deciding where to begin took longer than planned, and progress only accelerated once layout designers and administrators were brought together and received step-by-step instructions.

Overall, the main lessons were the need for stronger skills, sufficient planning time, and above all, building internal motivation and understanding.

At the end of the pilot project, did the publisher felt equipped and able to start making this a reality in their standard workflows?

Julija: Despite the pilot project it is unlikely that Vilnius University Press now produces all books accessibly, which highlights a broader issue: temporary projects alone are not enough, and lasting change requires a cultural shift within publishing houses. To support this, LAB continuously invites university and other publishers to recurring events and trainings on inclusive publishing.

Technical skills and attitudes are closely linked, and motivation is often low because producing accessible e-books is time-consuming and costly. There is no single tool that covers all accessibility needs, meaning publishers must use multiple tools and coordinate between them. Without tools, skills, financial resources, or experienced staff, it is difficult for publishers to stay motivated. One effective approach is to show publishers the real-world impact of accessibility by letting them observe readers using their books, demonstrating the difference between accessible and inaccessible content. This practical exposure is a key way to influence and motivate publishers toward accessibility.

One of the other types of complex content that was part of these pilot projects was maths. How was that topic of making equations and math formulas accessible approached?

Anke: Two pilot projects focused specifically on improving the structural accessibility of publications: one addressing overall semantic structure and the other concentrating on mathematical formulas.

The general semantics pilot emphasised the importance of correct heading levels (H1–H6) with a consistent hierarchy and highlighted that tables must never be used for layout purposes, as this prevents screen readers from interpreting content properly; tables should only be used for actual data.

The second pilot dealt with the challenges of complex mathematical and chemical formulas, which often include many symbols and special characters. Publishers sometimes insert formulas as images to avoid these complexities, but this makes them inaccessible, particularly for braille users. The recommended solution is to use MathML, which ensures that formulas are both visually correct and machine-readable for screen readers. Newer versions of InDesign now allow MathML to be incorporated directly into InDesign files, meaning publishers do not need to overhaul their workflows to XML in order to produce accessible formulas.

Were disabled users involved to try and access these extracts of these different publications? Was that feasible as part of this project?

Anke: Although the project team initially hoped to conduct user testing for all pilot publications, the wide variation and high complexity of the content, especially the data visualizations and mathematical formulas, made comprehensive testing unfeasible. Instead, they selected one publication for end-user evaluation to gain an initial understanding of potential issues.

The tester was instructed to assess the readability rather than the subject matter, since the content itself was challenging. The results were encouraging: with MathML in place, the end user could correctly read all mathematical formulas, and screen readers and text-to-speech tools rendered them accurately. While this was a positive outcome, the team noted that expanding such end-user testing to all pilot projects and including more content types would be valuable in the future.

Are there AI or other tools that publishers are now using in their workflows to make content accessible?

Anke: Several useful tools emerged from the pilot projects that can support publishers both in producing accessible content and in performing quality assurance.

For the production process, tools such as Adobe Premiere can generate AI-based subtitle files, while Speech Lab can translate English audio into subtitles in other languages, both important for meeting WCAG requirements.

For data visualizations, the Color Contrast Analyzer is essential for checking whether colour contrast meets accessibility standards.

Publishers who still work in Word are advised to use coded Word files and templates, which limit design freedom but greatly reduce errors in semantic structure and reading order. 

For quality assurance, the Ace by DAISY tool is widely used and helpful as an initial automated check, but it detects only around 30% of potential issues and does not fully identify problems such as incorrect language markup. Therefore, manual checks remain crucial, and accessibility experts can assist with more thorough assessments.

Additional tools include Bookshelf, which allows publishers to test EPUB files with screen readers to ensure correct semantic reading, and Sigil, which is useful for manually inspecting and correcting EPUB files, especially for validating complex structures like MathML or ChemML in formula-heavy publications.

What would you do differently if there was a next round of pilot projects?

Anke: The pilot projects demonstrated a strong and effective model for collaboration between accessibility experts and publishers, and the team expressed interest in conducting another round in the future.

While occasional initiatives are helpful, establishing a more structured and ongoing approach would better support publishers in advancing their accessibility practices. One major lesson was that the three-month timeline for each pilot was too short for the amount of work involved. Although the short duration was intentional to avoid overburdening publishers, the successful cooperation showed that working together more deeply, and over a longer period, could further strengthen publishers’ workflows beyond individual parts of a publication.

For future rounds, we would also aim to incorporate more extensive end-user testing across all pilot projects, not just selected ones, to ensure that accessibility improvements truly meet readers’ needs.

Julija: We would also allocate more time and begin by introducing publishers directly to end users, allowing them to see firsthand what it means to read their books with assistive technologies. They would also prepare training materials and methodological guides in advance to avoid delays.

Overall, the project successfully produced one more accessible e-book and provided valuable learning experiences for the library, especially as it was among their first collaborations of this kind. The pilot reinforced that outcomes depend heavily on each publisher’s motivation and perspective: some publishers are eager to make their books accessible, while others need more guidance and encouragement. Selecting the right publisher and helping them understand that accessibility is not just a legal obligation under the European Accessibility Act, but a cultural and ethical value within the publishing house, is therefore essential.

Was this the first to go under the bonnet in the production process, or were the libraries already familiar with that side of things? 

Julija: The pilot project offered LAB an opportunity to better understand accessibility work from the publisher’s perspective, particularly because this was their first time collaborating with an academic publisher. Academic books tend to be more complex than typical fiction titles, so the experience provided valuable insight into the challenges specific to that sector. Although LAB was already familiar with the overall accessibility process and had worked with other publishers before, this collaboration gave LAB a deeper understanding of how accessibility efforts function within a university press context.

Anke: Dedicon did projects like this in the past before, but, to be able to really get insight into the publisher's workflow, to see how they were working, and especially with academic content, that was new for Dedicon as well. These pilot projects really helped to build a bridge between the publisher and the accessibility expert, so we know how to find each other now, and that is a great takeaway from these pilot projects.

Where can we find more information about the pilot projects? 

An overall report is made and available on the European Accessibility Directory website. In this report you can read in more detail about all the different pilot projects that were conducted, and what the outcomes of all the pilot projects were. Both the general and the specific best practices are also listed there.