Alt Text That Works: Writing Image Descriptions for Scholarly and Educational Book

Most alt text fails in one of two ways. It says nothing useful, or it says too much. A screen reader user hears "image" or a filename, or sits through a 90-word paragraph about a photo that needed ten.

Describe the job, not the picture

Every image is in a book for a reason. A photograph of a lab bench may show equipment setup. A line chart may prove that a trend reversed. The alt text should serve that reason. Ask what the sighted reader is meant to take from the image, then write that.

Skip "image of" and "picture of". Screen readers already announce an image. Skip the caption too. If the caption says "Figure 4.2: Enrolment by region, 2015 to 2024", the alt text adds what the caption leaves out, such as which region grew fastest.

Sort images into three kinds

Decorative images carry no meaning: a chapter-opening flourish, a divider, a background texture. These get empty alt text (alt=""), so assistive technology skips them. Describing them wastes the reader's time.

Informative images need a short description, usually a sentence or two. A portrait in a biography, a map of a battle, a cell under a microscope.

Complex images hold data or structure: charts, diagrams, flow charts, annotated figures. A sentence cannot carry them. Give a short alt text that names the image and states the main finding, then put the full detail in a long description or an adjacent data table. EPUB Accessibility 1.1 allows this, and so does WCAG success criterion 1.1.1.

Where it goes wrong in scholarly books

Charts without data. "Bar chart showing sales" tells a student nothing. "Bar chart: online sales overtook print in 2022; full figures in the table below" gives the point and the route to the detail.

Equations as pictures. Maths set as an image needs MathML or a real description. Alt text reading "equation" fails every learner who relies on it.

Repeated descriptions. If a figure appears in the text and again in a summary box, the reader hears the same paragraph twice.

Guessing at identity. Describe visible facts. Do not name a person, assume a gender or infer an emotion unless the book itself says so.

Why automated drafts are not enough

Software can draft a description of what a picture contains. It cannot know why the figure sits in chapter four, what the author is arguing, or which detail an exam question will test. A draft from a tool is a starting point. Someone who understands the subject has to review it, and for technical content that person needs domain knowledge.

Make it part of production

Alt text written after the files are final is rushed and inconsistent. It works better when authors flag the purpose of each figure at manuscript stage, editors check descriptions alongside the copy, and the same style rules apply across a whole list. A short style guide settles most arguments: when to describe, how long, how to handle charts.

Where S4Carlisle fits

S4Carlisle writes and reviews image descriptions as part of EPUB and PDF production, with subject-matter checks for STEM and scholarly titles and tooling from the NINJA AI Ecosystem to find missing or weak descriptions across a backlist. Our workflows meet EPUB Accessibility 1.1 and WCAG 2.x AA requirements.

S4Carlisle holds Benetech GCA certification and brings more than 50 years of publishing experience to every title. To review a sample chapter with our team, write to sales@s4carlisle.com.