If removing the image changes nothing, question why it is there.
Visuals add file weight, production work, rights questions, accessibility decisions, and new ways for an export to fail. They earn that cost when they reduce cognitive effort or communicate something prose cannot express as efficiently. A decorative image can still be intentional, but it should not interrupt reading, carry hidden information, or receive a noisy description that creates no benefit.
Write the image brief before making the image
Complete this sentence: “After this visual, the reader can see/compare/do ______ more easily.” If the blank only says “the page looks better,” decide whether the visual is genuinely decorative and treat it accordingly.
Classify the image in its actual reading context.
| Image role | What the reader needs | Appropriate treatment | Common failure |
|---|---|---|---|
| Informative | The meaning or observation conveyed by the image. | A concise alternative that communicates the relevant information; use a caption for context useful to everyone. | Describing colors and objects while omitting the point. |
| Functional | The action or destination associated with the image. | Name the function, such as “Download the worksheet,” rather than the icon's shape. | Alt text such as “arrow icon” that never explains what happens. |
| Complex | The main takeaway plus access to the underlying relationships or data. | A brief alternative and a nearby extended explanation, data table, or equivalent account. | Trying to compress an entire chart into one exhausting text attribute. |
| Decorative | No information; it should not interrupt the reading flow. | Mark it so assistive technology can ignore it; in HTML-based contexts this commonly means an empty alt value. | Announcing “decorative flourish” on every chapter page. |
| Text in an image | The words and their purpose. | Prefer real text. When an image of text is essential, provide an equivalent and preserve legibility. | Embedding small paragraphs that cannot resize or reflow. |
The W3C Web Accessibility Initiative image tutorials organize image guidance around purpose and context. The same asset can need different alternative text in two chapters because the surrounding argument changes what matters.
Describe the contribution, not the inventory of pixels.
Start by reading the paragraph before and after the image. Ask what a reader would miss if the image did not load. State that information economically, without repeating a nearby caption word for word or adding “image of” when the format already announces an image.
Weak: “Screenshot of a menu”
This identifies the asset type but not why the screenshot appears.
Better: “Export menu with EPUB selected”
This preserves the interface state the instructions depend on.
Weak: “Line chart with blue and green lines”
This describes appearance while withholding the relationship.
Better: “Completion rises after reminders; the no-reminder group stays flat”
This communicates the takeaway; a nearby table can provide the exact values.
Captions and alternative text are not interchangeable. A caption may identify a source, frame the significance, or add information visible to every reader. Alternative text ensures the image's relevant contribution is available when the visual itself is not perceived. Avoid redundant wording when the same information is already adjacent and available to the reading system.
Take each visual through six explicit gates.
- Gate 01
Purpose
Record the reader task, comparison, proof, orientation, or pause the image supports. Remove visuals with no defensible role.
- Gate 02
Rights and provenance
Record the creator, source, license or permission, allowed uses, required credit, and the original file. “Found online” is not a rights record.
- Gate 03
Composition
Crop for the information readers need, preserve aspect ratio, remove accidental clutter, and ensure labels remain understandable at the intended display size.
- Gate 04
Text equivalent
Classify the image, write contextual alternative text, add a useful caption, and supply a longer explanation or data equivalent for complex visuals.
- Gate 05
Placement
Place the image close to the passage that introduces it. Check reading order and avoid directional language that fails when a reflowable layout changes.
- Gate 06
Artifact test
Inspect the exported file on more than one screen, at different text sizes where possible, and with images unavailable or nonvisual navigation in mind.
Different visual types fail in different ways.
Cover
The cover must work as both an identification surface and a small thumbnail. Treat retailer dimensions, file type, and content rules as destination-specific requirements and check them close to publication.
Screenshot
Crop to the interface state being taught, remove personal or secret data, and pair numbered callouts with prose that still makes sense when layout changes.
Diagram
Keep labels short, preserve a logical reading order, and explain the relationship in nearby prose. A diagram should reduce explanation, not force readers to decode a poster.
Chart
State the takeaway, label axes and units, avoid relying on color alone, cite the data, and provide exact values or an equivalent account when those values matter.
Photograph
Describe only details relevant to the surrounding claim. Verify releases or permissions where people, property, artwork, or private environments create rights questions.
Exercise or worksheet
Decide whether it must be completed on a fixed page. A companion PDF may serve this task better than forcing a page-dependent form into a reflowable chapter.
Review the exported edition, not just the source canvas.
- Every image has a recorded purpose and is placed next to the passage that uses it.
- Rights, licenses, model or property releases where applicable, and required credits are documented.
- Text inside images remains legible at realistic display sizes—or has been replaced by actual text.
- Information is not distinguished by color alone.
- Informative and functional images have contextual alternatives; decorative images are treated as decorative.
- Complex diagrams and charts have a nearby explanation or data equivalent.
- Captions identify sources or significance without needlessly duplicating the alternative text.
- Aspect ratios, crop, orientation, and reading order survive export.
- The edition remains understandable when an image is unavailable.
- File size and visual quality are balanced by inspecting the actual delivery file rather than trusting an export preset.
Rights and content rules belong in the image workflow.
Image quality is not only a design issue. Before publishing through Amazon KDP, review the current KDP Content Guidelines and confirm you have the necessary rights for the material you submit. If KDP will convert the manuscript, also check its current supported ebook manuscript formats and inspect what the conversion does to image placement and readability.
- W3C WAI: Images Tutorial — contextual guidance for informative, decorative, functional, complex, and text images.
- W3C: EPUB 3.3 — technical context for EPUB publications and their resources.
- Amazon KDP Content Guidelines — current content and publishing requirements for that platform.
Decide the format before designing image-heavy pages.
Reflowable and fixed-page editions place different constraints on captions, diagrams, tables, and exercises. Compare them in the ebook format guide, place the visual pass inside the complete ebook workflow, or explore the dedicated AI ebook generator with images page.