Alt Text Coverage: Counting the Images That Still Say Nothing

2026-10-28

Alt text coverage is the first accessibility metric most teams report, and often the least honest. The number is easy to inflate: count any non empty alt attribute and almost any site can claim success while screen reader users still hear image final v2 dot jpeg read aloud.

A coverage metric is still worth having. It just has to count what matters, which means defining quality into the measurement instead of counting attributes like raffle tickets.

The three buckets

Meaningful alt text describes the image's job in the page: what a product looks like, what a chart concludes, what action an icon performs. Decorative images, borders, spacers, purely visual flourishes next to text that already says it, should carry an empty alt so they are skipped.

Junk is everything else: missing attributes, filenames, keyword stuffed descriptions, and the word image repeated. Junk is worse than a gap in some ways, because it performs coverage while delivering noise.

Counting honestly

Crawl your templates and a sample of real pages, including product pages with user generated images and marketing pages built by hand. Classify a sample manually to calibrate any automated counting, since tools can detect presence but not meaning.

Weight the metric by traffic. A missing description on your best selling product matters more than ten on an archived campaign page. Coverage reported for your top templates and top pages is the number leadership should see.

Product images are their own problem

Ecommerce alt text fails at scale: thousands of products, images arriving from suppliers, and descriptions that duplicate the product title word for word. Screen reader shoppers hear the title twice and learn nothing about colour, angle, or detail.

Set a pattern: alt text adds what the title does not say, like colourway, view, and key visible features. Supplier feeds should be checked against it on import, because fixing ten thousand images later is a project nobody funds.

Improving the rate over time

Work templates first, since one template fix covers every page built on it. Then work down the traffic list. Assign new content an alt text standard at creation, with decorative marking as an explicit choice in the CMS, not a default.

Retest quarterly and report the bucket mix, not just the headline. A rate that rises because the decorative bucket was audited is real progress. A rate that rises because filenames were pasted in is debt wearing a costume.

Frequently asked questions

What counts as good alt text for a product image?

A short description of what the image shows that the surrounding text does not already state: the colourway pictured, the view or angle, and any visible feature a sighted shopper would use to decide. Keep it factual and brief; it is information, not advertising.

Should every image have alt text written out?

Every image needs an alt decision. Decorative images should have an empty alt attribute so assistive technology skips them. Writing descriptions for decoration creates noise. The metric is correct decisions, not maximum text.

How do I measure alt text quality automatically?

Tools can flag missing, empty, filename like, and overlong alt text at scale. Judging whether a description is meaningful still needs human sampling. Combine a crawl for structure with a manual review of your highest traffic pages.

What coverage rate should we target?

Aim for complete correct decisions on your key templates and highest traffic pages first, then expand. An honestly counted rate in the high nineties on important pages beats a claimed hundred across the site that nobody has sampled.