Why MathML — Not Images — Is the Key to EAA-Compliant STEM Ebooks
If your STEM ebooks store their equations as pictures, they are not accessible — and since 28 June 2025, in the European Union, that is no longer just a quality problem. It is a legal one.
Here is the short version, for anyone who needs an answer before the meeting: an image of an equation cannot be read, navigated, or rendered in braille by assistive technology. MathML can. That single distinction is what separates a compliant STEM title from one that fails an accessibility audit. After three decades converting mathematically dense books and journals, we can tell you that mishandled equations are the most common — and most expensive — reason a STEM backlist fails to meet the new rules.
This article explains what “accessible math” actually means, why images fall short, how MathML fixes it, and what the European Accessibility Act (EAA) genuinely requires of the maths on your pages.
What does “accessible math” actually mean?
Accessible math means the mathematics in your book is encoded as structure, not as a snapshot. MathML — Mathematical Markup Language, a long-standing W3C standard — describes an equation as a tree of meaning rather than an arrangement of pixels. A fraction becomes a container with a clearly named numerator and denominator. A square root has an explicit argument. A summation carries its own upper and lower bounds as labelled parts.
To make that concrete, the simple fraction a/b is not a picture in MathML — it is labelled parts a machine understands:
<math xmlns="http://www.w3.org/1998/Math/MathML">
<mfrac>
<mi>a</mi>
<mi>b</mi>
</mfrac>
</math>A screen reader sees the mfrac and announces a fraction,
then lets the reader step into the numerator and denominator separately.
Scale that same principle up to matrices, integrals, and nested
subscripts, and you have equations a reader can actually explore.
Because that structure is preserved, a screen reader can speak the equation correctly and, crucially, let the reader move through it — drilling into a denominator, stepping across the terms of a sum — the same way a sighted reader’s eye scans a formula. An image gives the reader none of this. It gives them a locked door.
Why an image of an equation fails
Publishers often assume that adding alt text to an equation image solves the problem. It does not, and understanding why is the heart of this topic.
A block of alt text is a flat string. For anything beyond a trivial expression, it collapses a two-dimensional, hierarchical object into one long, unpunctuated sentence that a listener cannot pause inside, re-enter, or explore. Try describing a matrix, a continued fraction, or a limit with three nested subscripts in a single line of prose — then imagine hearing it read aloud once, at speed, with no way to go back into any part of it. That is the daily experience an image-plus-alt-text equation forces on a blind student.
Images fail in three further ways that matter for compliance. They do not reflow, so on a small screen the equation either shrinks to illegibility or forces horizontal scrolling. They cannot be rendered on a braille display, which reads from the underlying markup, not from pixels. And they cannot inherit a reader’s personal settings — font size, contrast, spacing — because there is no text there to restyle. An equation image is, functionally, a wall.
How MathML makes math readable — and navigable
Encode the same equation in MathML and the wall becomes a window. Modern assistive technology now reads MathML well: screen readers such as NVDA, JAWS, and VoiceOver can read equations aloud and let users navigate through fractions, subscripts, or matrices element by element. On Windows, the current sweet spot is NVDA combined with the MathCAT add-on, and NVDA (from version 2025.1) and JAWS (from its 2026 releases) now support accessible math in Firefox as well.
This is not a fragile, experimental setup any more. A turning point came in 2023, when MathML Core landed in Chromium and completed cross-browser support, so native mathematics rendering became viable in every major browser for the first time. The tooling that was patchy five years ago has matured into something publishers can rely on — provided the maths reaches the reader as MathML in the first place.
The reader’s payoff is control. Instead of a single flat readout, they hear the expression at the top level, then choose to enter it, walk term by term, expand a fraction, or step into an exponent. For a genuinely complex STEM equation, that navigability is the entire difference between “technically has alt text” and “actually usable.”
What the EAA actually requires for your maths
The European Accessibility Act came into force on 28 June 2025, and ebooks are explicitly within its scope. In practice, conformance flows through a chain of standards: the EAA points to the harmonised European standard EN 301 549, which in turn adopts the Web Content Accessibility Guidelines (WCAG) 2.2 at Level AA, and for EPUB specifically the W3C’s EPUB Accessibility 1.1 specification maps those success criteria onto ebook structure.
You do not need to memorise that chain. You need to know what it means for equations: maths must be programmatically determinable and perceivable to assistive technology, which for all practical purposes means real MathML, not an image with a caption. An accessibility conformance report that lists your equations as images will not survive scrutiny — and under the EAA, non-conformant titles can be pulled from the EU market. The pressure is not theoretical; the U.S. is tightening comparable requirements over 2025 and 2026, so the same fix protects you in more than one market at once.
Build accessibility in at the source, not at the end
The costly way to do this is to bolt accessibility onto finished files, title by title, forever. The durable way is to hold your maths as MathML in the source XML — for journals, inside JATS; for books, in your production DTD — so that every output inherits it automatically.
When equations live as MathML in the primary structure, the EPUB gets navigable maths, the HTML gets navigable maths, and a PDF produced to PDF/UA gets navigable maths — from one source, with no per-format rework. Accessibility stops being a remediation cost and becomes a property of your pipeline. This “born-accessible” approach is why we convert to MathML at the XML stage rather than patching EPUBs at the end: it is cheaper across a catalogue and it is the only version that scales to a backlog of thousands of titles.
Five equation mistakes we see in STEM backlists
- Equations saved as PNG or JPEG. The default failure. Every such image is a title-level compliance risk.
- Raw LaTeX left in the text. LaTeX is an excellent source, but shipped as-is it is not readable by assistive technology; it must be converted to MathML for delivery.
- Alt text used as a substitute for structure. Fine for a lone inline symbol; inadequate for anything a reader needs to navigate.
- Invalid or “presentation-only” MathML. Malformed markup, or MathML that encodes only appearance with no semantic intent, reads poorly and can break in some readers.
- No validation step. Files are never checked with a tool such as the DAISY Consortium’s Ace checker, so errors surface months later at the retailer’s ingestion gate.
Frequently asked questions
Is alt text enough to make an equation accessible? No. For anything beyond a single symbol, alt text flattens a structured expression into an unnavigable sentence. MathML is required for the reader to explore the equation.
Do I have to redo my entire backlist at once? No. Triage it. Convert your highest-value and EU-selling titles first, fix the equations as MathML during that pass, and sequence the rest. A structured backlist remediation plan spreads the cost sensibly.
Should we author in LaTeX or MathML? Author in whatever your team knows — LaTeX is fine as source. What matters is that equations are delivered as valid MathML, which a good conversion pipeline produces automatically from LaTeX.
How do we check whether our ebooks pass? Validate every title with the DAISY Ace checker and confirm each equation is present as MathML, not an image, then record the results in an accessibility conformance report.
Get your hardest title checked — free
If you are staring at a STEM backlog and unsure how much of it will fail, the fastest way to find out is to see one done properly. Send us your most equation-heavy title and we will remediate a sample chapter to EPUB Accessibility 1.1 / WCAG 2.2 AA, with full MathML, at no cost, so you can judge the quality before committing to anything. Thirty years of complex STEM conversion sits behind every sample we return.