Alt Text for Complex Images in EPUB: A Publisher’s Guide

Alt Text for Complex Images in EPUB: How Publishers Should Describe Charts, Diagrams and Visual Content

Writing alt text for a photograph is often straightforward.

Writing alt text for a scientific diagram, map, infographic, flowchart or multi-series graph is not.

A complex image may communicate:

• Data
• Relationships
• Sequence
• Comparison
• Location
• Cause and effect
• Hierarchy
• Trends
• Labels
• Multiple layers of meaning

Trying to compress all of that into one short sentence can make the description technically present but practically useless.

That is why publishers should not ask:

“Does this image have alt text?”

The better question is:

“Can a reader who cannot see this image obtain the same essential information?”

W3C’s guidance for non-text content requires text alternatives that serve an equivalent purpose. For complex images such as charts and diagrams, W3C specifically recognizes situations where a short description alone is insufficient and a longer text alternative is needed.

Quick Answer: How Should Complex EPUB Images Be Described?

Use a simple decision model:

Identify Purpose → Decide Short or Extended Description → Describe Meaning → Add Structure → Human Review → Assistive-Technology QA

For simple informative images, concise alt text may be enough.

For complex content such as:

• Charts
• Maps
• Scientific diagrams
• Process diagrams
• Infographics
• Complex tables presented as images

use a short alternative to identify the image, then provide a fuller description when the important information cannot be communicated concisely.

DAISY’s current image-description principles recommend context-specific, audience-appropriate descriptions and extended descriptions when alt text alone cannot adequately communicate the image.

Alt Text Is About Purpose, Not Appearance

One of the biggest mistakes is describing what an image looks like instead of what it means in context.

Consider a graph in an economics textbook.

Weak description

“A blue and orange line graph.”

Technically, it describes appearance.

But it does not tell the reader why the graph matters.

Better description

“Line graph showing inflation rising from 3% to 8% while wage growth remains below 4%.”

Now the reader receives the central relationship.

W3C’s image guidance stresses that the correct text alternative depends on the image’s purpose, context and content. The same visual may require a different description in a different publication.

The Four Image Types Publishers Should Separate

Not every image should be handled the same way.

1. Informative Images

These communicate useful information.

Examples:

• Photographs
• Illustrations
• Simple diagrams
• Icons carrying meaning

The alternative should communicate the essential information.

2. Decorative Images

These provide visual styling but no meaningful information.

Examples:

• Flourishes
• Decorative borders
• Background ornaments
• Scene-break graphics

These should normally be implemented so assistive technology can ignore them.

W3C recommends a null text alternative for genuinely decorative images rather than forcing screen-reader users to hear meaningless descriptions.

3. Functional Images

These act as controls or links.

The alternative should describe the function, not merely the graphic.

For example:

Instead of:

“Magnifying glass icon”

use:

“Search”

when the image functions as a search control.

4. Complex Images

These communicate information that cannot usually be conveyed in one short phrase.

Typical examples include:

• Charts
• Graphs
• Maps
• Flowcharts
• Scientific diagrams
• Infographics
• Engineering drawings

This is where publishers need a more deliberate description workflow.

When Is Short Alt Text Enough?

Short alt text can work when the essential meaning is simple.

For example:

Image

A photograph of a library building.

Possible alt text

“Main entrance of the university library.”

That may be enough if the image is simply identifying the building.

But if the same photograph appears in an architecture textbook to demonstrate Gothic structural elements, the description may need to mention those relevant features.

The question is:

What does the reader need from this image at this point in the book?

Not:

“How many visual details can we list?”

When Do You Need an Extended Description?

Use an extended description when the information cannot be communicated effectively in concise alt text.

W3C guidance explicitly recognizes charts and diagrams as examples where a short description may need to be supplemented by a longer text equivalent.

Extended descriptions are especially useful for:

• Multi-series graphs
• Detailed maps
• Scientific processes
• Organizational charts
• Anatomical diagrams
• Engineering schematics
• Complex infographics
• Multi-step visual workflows

DAISY also recommends extended descriptions where supported when alt text alone is insufficient.

Short Alt Text vs Extended Description

A strong workflow may use both:

Short alt text:
“Bar chart comparing quarterly revenue. Detailed description follows.”

Extended description:
Explain the data, comparisons, largest changes and important conclusions.

How to Describe Charts

Charts should not be described as decoration.

The description should focus on the message the chart communicates.

For a chart, consider:

• Chart type
• Axes
• Units
• Categories
• Main trend
• Largest differences
• Peaks or declines
• Important exceptions

Poor

“Bar chart showing sales.”

Better

“Bar chart showing sales increasing from $2.1 million in Q1 to $3.8 million in Q4, with the largest increase between Q2 and Q3.”

If exact values matter to the reader, provide the underlying data in accessible text or table form where practical.

W3C specifically notes that for complex charts, a longer description can include the type of chart, summary, trends and implications, with actual data supplied in a table where appropriate.

How to Describe Diagrams

Diagrams often communicate relationships rather than simple objects.

A biology diagram might show:

• Components
• Positions
• Movement
• Sequence
• Connections
• Functions

Do not simply list everything visible.

Explain the relationships that matter.

For example:

Weak

“Diagram of the human heart.”

Better

“Diagram showing blood flow through the four chambers of the heart, from the right atrium through the lungs and back to the left ventricle.”

If the image contains detailed labels required for study, an extended description may be necessary.

How to Describe Maps

Maps present a different challenge because meaning often depends on:

• Position
• Direction
• Distance
• Boundaries
• Routes
• Regions
• Patterns

Descriptions should avoid relying only on phrases such as:

“on the left”

or:

“shown in red.”

Instead, communicate the actual relationship.

Example

“Map showing the trade route from London through Paris and Venice to Constantinople, with a secondary route through central Europe.”

The purpose is to make the geographic information understandable without requiring visual access.

How to Describe Infographics

Infographics often combine:

• Text
• Icons
• Data
• Visual hierarchy
• Arrows
• Colour coding
• Multiple concepts

These are frequently too complex for one alt-text field.

A stronger approach is:

• Provide concise alt text identifying the infographic
• Reproduce important text as real text
• Explain sequence or relationships
• Provide an extended description where needed
• Avoid communicating meaning through colour alone

The description should deliver the same message, not replicate every decorative element.

Complex Images Should Not Become One Giant Sentence

A common remediation mistake is putting hundreds of words into an alt attribute.

That creates a poor reading experience.

For genuinely complex content, separate:

Identification

from

Explanation

For example:

“Diagram of the publishing accessibility workflow. Detailed description follows.”

Then provide the longer explanation in structured text.

This is easier to:

• Navigate
• Pause
• Review
• Read in braille
• Understand with a screen reader

Extended descriptions can also use richer structures such as lists and tables rather than forcing everything into a single string of text. DAISY’s 2026 principles explicitly recognize that extended descriptions may contain structured content.

Context Changes the Correct Description

The same image may require different alt text in different books.

Imagine a photograph of a tiger.

Children's story

“A tiger watches from behind tall grass.”

Biology textbook

“Bengal tiger showing the characteristic orange coat with black vertical stripes.”

Conservation report

“Bengal tiger photographed in a protected forest reserve.”

None of these descriptions is universally correct.

The surrounding content determines what matters.

That is why DAISY emphasizes descriptions that are both context specific and audience appropriate.

Audience Matters Too

A description written for a five-year-old should not sound like one written for a postgraduate medical textbook.

Consider:

• Reader age
• Subject knowledge
• Educational level
• Purpose of the publication
• Vocabulary used in surrounding text

For specialist academic content, image-description work may require:

• Editors
• Subject-matter experts
• Authors
• Accessibility specialists

DAISY specifically recommends using the best available expertise, including authors, subject specialists and image-description professionals.

Alt Text for Educational and Academic EPUBs

Academic publishing creates some of the hardest description problems.

Typical content includes:

• Graphs
• Maps
• Chemical structures
• Anatomical diagrams
• Statistical figures
• Scientific workflows
• Historical images
• Mathematical plots

A scalable production workflow should therefore classify images before description.

Simple

Straightforward photographs and illustrations.

Moderate

Images requiring contextual explanation.

Complex

Charts, diagrams, maps and multi-part graphics.

Specialist

Scientific or technical visuals requiring subject expertise.

This prevents every image from being treated as the same production task.

A Publisher Workflow for Complex Image Descriptions

Stage 1 — Inventory the Images

Identify:

• Total image count
• Image type
• Repeated images
• Decorative content
• Complex visual content
• Specialist subject matter

Stage 2 — Classify by Purpose

Decide whether each image is:

• Informative
• Decorative
• Functional
• Complex

This determines the appropriate treatment.

Stage 3 — Review Context

Read:

• Caption
• Surrounding paragraphs
• Section heading
• Learning objective
• References to the image

Do not write descriptions in isolation from the manuscript.

Stage 4 — Choose Short or Extended Description

Ask:

Can the essential purpose be communicated concisely?

If yes, short alt text may be sufficient.

If not, provide an extended description.

Stage 5 — Draft the Description

Focus on:

• Essential meaning
• Relevant relationships
• Important values
• Purpose in context

Avoid:

• Repeating visible captions word for word
• Unnecessary visual details
• Subjective interpretation
• Placeholder descriptions

Stage 6 — Subject Review

Complex academic content should be reviewed by someone who understands the subject.

This helps prevent a description that is grammatically good but scientifically wrong.

Stage 7 — Accessibility QA

Check whether the description is:

• Accurate
• Useful
• Context appropriate
• Objective
• Structured correctly
• Free from unnecessary repetition

Stage 8 — Assistive-Technology Testing

Representative testing can reveal whether descriptions work naturally when read aloud or accessed through assistive technologies.

The full process is:

Inventory → Classify → Context → Describe → Expert Review → Accessibility QA → Test

Can AI Write EPUB Alt Text?

AI can help accelerate image-description workflows.

But it should not automatically become the final author.

AI-generated descriptions may:

• Miss context
• Invent details
• Misread charts
• Misinterpret specialist diagrams
• Overdescribe irrelevant visual features
• Introduce biased language
• Fail to understand why the image appears in the book

DAISY’s 2026 image-description principles specifically call for special care in quality review when artificial intelligence is used.

The safer workflow is:

AI Draft → Context Review → Subject Review → Accessibility QA → Approval

Not:

AI Generate → Publish

Automated Alt Text vs Human-Written Descriptions

The two can complement each other.

Automation can support production.

Human judgment protects quality.

Why Human QA Matters

A description can be technically present and still fail the reader.

Examples:

• “Image.”
• “Diagram.”
• “Graph showing information.”
• “Photo of a person.”
• AI-generated text that describes irrelevant background details

These may satisfy the presence of an alt attribute.

They do not provide meaningful accessibility.

DAISY’s current principles recommend a formal QA process involving editors, specialists and ideally people who use image descriptions for accessibility.

Image Description Quality Points for Publishers

Before releasing an EPUB, confirm that:

• Meaningful images have useful alternatives
• Decorative images do not create screen-reader noise
• Descriptions match the surrounding context
• Complex visuals receive extended descriptions where needed
• Chart trends and relationships are communicated
• Critical data is available in accessible form
• Technical descriptions have subject review
• AI-generated descriptions receive human QA
• Language is objective and inclusive
• Final EPUB is tested with assistive technology

These points are more useful than simply measuring:

“Percentage of images with alt text.”

Presence is only the first step.

Quality is what makes the description useful.

Accessibility Metadata Should Reflect the Result

EPUB accessibility metadata can communicate that a publication contains alternative text or extended descriptions.

But those claims need to match the actual content.

W3C EPUB Accessibility guidance notes that accessibility metadata can identify features such as alternative text and extended descriptions so readers can determine whether a publication meets their needs.

Therefore:

Image description QA should happen before accessibility metadata is finalized.

Do not claim a feature the book does not consistently provide.

How Gentize Supports Image Description Workflows

Gentize's current public Accessibility practice specifically lists:

• Alt-text and structural tagging at scale
• Accessible EPUB remediation
• EPUB and ONIX accessibility metadata
• NVDA, JAWS and VoiceOver testing

Gentize also states that its accessibility work is backed by 50+ A11Y-certified testers with visual impairments, and its published process combines automated/manual auditing with assistive-technology testing.

For publishing-scale work, Gentize publicly documents a 35,000-title EPUB 3 backlist accessibility programme.

These proof points are relevant because complex image description is not only a writing task.

At scale, it becomes a workflow involving:

classification + editorial context + specialist review + accessibility QA + metadata + final testing

Final Takeaway

Good alt text is not simply a description of what appears in an image.

It communicates the information a reader needs from that image in the context of the publication.

For simple images:

Use concise, meaningful alt text.

For complex images:

Use short identification + an extended description when necessary.

For publishing teams:

Context → Meaning → Expertise → QA → Testing

And when AI is involved:

Use it to accelerate drafting—not to replace editorial and accessibility judgment.

That is the difference between an EPUB that technically contains alt text and an EPUB whose images are genuinely understandable to readers who cannot see them.