WCAG 1.4.5 Images of Text says that when your technology can produce a given visual appearance with real text, you must use real text rather than a picture of text. Headlines, buttons, pricing tables, and quotes belong in styled HTML — not baked into JPGs or PNGs — so users can resize, recolor, and have them read aloud.

What 1.4.5 actually requires

The normative wording from the W3C is short: “If the technologies being used can achieve the visual presentation, text is used to convey information rather than images of text,” with two exceptions.

  • Customizable. An image of text is allowed if the user can visually adjust it — change the font, size, color, and spacing to their needs.
  • Essential. An image of text is allowed when a particular visual presentation is required to convey the meaning. The W3C names logotypes — text that is part of a logo or brand name — as always essential.

The key phrase is “if the technologies being used can achieve the visual presentation.” On a modern website using HTML and CSS, you can reproduce almost any heading, label, or block of body copy with real text and web fonts. So on the web, 1.4.5 effectively means: if it’s words a user needs to read, render it as text, not as an image.

Who 1.4.5 protects

The whole point of this criterion is letting people adapt how text looks. Real text bends to the user; a flattened image does not.

  • Low-vision users zoom pages to 200% or 400%. Real text reflows and stays crisp; an image of text turns into a blurry, pixelated mess and may force horizontal scrolling.
  • People who set custom styles — high-contrast themes, dark mode, bigger line spacing, a dyslexia-friendly font — get nothing from baked-in text. Their stylesheet can’t touch pixels in a JPG.
  • Screen reader users (JAWS, NVDA, VoiceOver) hear nothing from an image of text unless someone wrote alt text, and even then they can’t navigate, search, or select within it.
  • People with visual tracking or cognitive disabilities who rely on adjusted spacing and alignment lose those controls entirely when the text is an image.

Concrete failures — and the fix

These are the patterns we see most when auditing real sites.

1. A headline saved as a PNG. A designer exports a hero headline as an image to get an exact font. At 400% zoom it pixelates, and a screen reader skips it. Fix it with text and CSS web fonts:

<!-- Fail: text baked into an image -->
<img src="hero-headline.png" alt="Summer Sale — 40% Off">

<!-- Pass: real text styled with CSS -->
<h1 class="hero-headline">Summer Sale — 40% Off</h1>
.hero-headline {
  font-family: "Your Brand Font", system-ui, sans-serif;
  font-size: clamp(2rem, 6vw, 4rem);
  color: #1a1a1a;
}

2. A pricing or comparison table exported from a design tool. Flattening a table into one image destroys the data for assistive tech and breaks zoom. Rebuild it as a real <table> with proper headers so screen readers can announce rows and columns.

3. Buttons and navigation as image sprites. “Add to Cart” or menu labels saved as graphics fail 1.4.5 and usually fail color contrast too, since a checker can’t read the pixels. Use real <button> and link text styled with CSS.

4. Promotional banners with the offer baked in. “Free shipping over $50” inside the banner image is unreadable to a screen reader and un-resizable. Put the text in HTML and let the image be a background.

5. Quotes and testimonials as graphics. Pull-quote images look nice but lock the words away. Mark them up as real text in a <blockquote>.

When you genuinely must keep an image — a true essential case, or a logo — give it accurate alt text to satisfy 1.1.1 Non-text Content. That handles the screen reader, though it still can’t restore resizing, so reserve it for real exceptions.

How to test for images of text

You don’t need special software to spot most violations.

  1. Try to select the text. Click and drag across a headline or label. If you can’t highlight the words, it’s probably an image of text.
  2. Zoom to 200% and 400%. Use your browser zoom (Ctrl/Cmd and +). Real text stays sharp and reflows; an image of text blurs and pixelates.
  3. Switch on a custom or high-contrast theme. Words that don’t recolor are baked into an image.
  4. Run a screen reader over the section. In VoiceOver or NVDA, navigate the page. Content that’s announced only as “image” (or skipped) is a candidate failure.
  5. Use an automated scanner for a first pass. Tools flag <img> elements and missing alt text, but they can’t reliably tell decorative photos from images of text — a human still has to judge each one. That’s why thorough work pairs scanners with manual review.

Why this matters legally

ADA web accessibility lawsuits keep climbing. Plaintiffs filed over 4,000 digital accessibility cases in 2024, roughly 60% in federal court and the rest in state court, concentrated heavily in New York and California, according to UsableNet’s year-end report. E-commerce sites took the brunt — about 77% of those filings — and previously sued companies were targeted again, so a quick fix that misses things invites a repeat claim.

Images of text are easy for a plaintiff’s tester to document: they zoom in, the headline blurs, and the screenshot writes itself. Courts and the DOJ treat WCAG 2.1 AA as the practical yardstick for ADA Title III web claims, and 1.4.5 sits squarely at that AA level. For the W3C’s own framing of who this helps, see the Web Accessibility Initiative.

This is general information, not legal advice. For your specific exposure, consult a qualified attorney.

Fixing it the right way

There’s no shortcut here, and an accessibility overlay can’t help. An overlay is JavaScript that runs after your page loads; it cannot reach inside a JPG and turn flattened pixels back into readable, resizable text. The only real fix is to rebuild the content as HTML and CSS — which is exactly the hands-on, anti-overlay remediation Curbcut does.

Our approach is plain: [EXPERT_NAME] and the [AGENCY_NAME] team inventory every image of text, decide which are truly essential, convert the rest to styled text, and add correct alt text to what stays. When you need proof of conformance, we document it in a VPAT. If you’d rather not hunt down every baked-in headline yourself, start with a free scan and we’ll show you precisely where text is trapped in images — then remediate it for you.