Alt text is the written description attached to an image, and it is what a screen reader reads out in place of the picture. The test is not whether you described it. It is whether the reader gets what the image was there to give them, which is why the same photograph needs different alt text on different pages. A chart's alt text is the finding. A product shot's is the product. An image sitting alone inside a link is a special case, because there the words have to name where the link goes rather than what the picture shows. And decoration gets an empty alt attribute, which tells assistive technology to skip it.
In practice
Most images fall into two buckets, and it is worth knowing where the split is not clean. An image carrying meaning gets text saying what it tells the reader. An image carrying nothing gets an empty alt attribute so it is skipped. The exceptions go wrong most often. An image that is the only content inside a link or a button needs a name describing what the control does, and an empty attribute there leaves the control with no name unless something else on it supplies one. Nearly half the images found with no alt text in the wild are linked images, which turns one failure into two.
The common failure is not a missing attribute. It is a useless one. Filenames, the word image, and a description of what a chart looks like rather than what it says are all technically present and practically worthless. Naming the object instead of the job is the shape of it. PDF icon describes the picture. What the reader needed was that the link hands them a PDF.
Two details catch careful people. Marking an image as decoration means an empty alt attribute and no title attribute, because a title puts the image back into the accessibility tree and undoes the whole thing. And the alternative has to be in a language the page has declared, which normally means the language of the page around it. That is where automatically generated alt text falls over on a multilingual site, because it tends to come back in English whatever the page is written in, with nothing marking the change.
Why it matters
Missing alt text was found on 53.1% of the top one million home pages in 2026, second only to low contrast text. That is the machine-detectable half of the problem. The other half is the half no machine reaches, because a scanner can see that alt text exists and cannot tell you whether it is true. Context decides the answer and context is exactly what software does not have. A maple leaf might mean Canada or it might be illustrating a leaf, and only the page around it settles which.
The test
Turn the images off and read the page. If the words that replaced a picture tell you what the picture told you, they are right. If you get a filename, the word image, or a description of what a chart looks like, they are not. And if a link goes silent, that image was doing a job nobody wrote down.
Where this shows up on the site
Related terms
- Screen readerA screen reader is software that reads the screen out loud, or sends it to a braille display, for people who cannot see it.
- Semantic HTMLSemantic HTML means using HTML elements for what they mean rather than for how they look.
- Automated testingAutomated testing is software that checks a page's code for the accessibility failures a machine can prove.
- CaptionsCaptions are a synchronized text version of everything you would hear in a video, covering speech and the non-speech sound that matters as well.
- Data visualisationA data visualisation is any graphic that turns numbers into a picture.
- Decorative imageA decorative image is one that serves only an aesthetic purpose, carries no information, and does no work.
Knowing the word is the easy part.
Find out where your own site stands. The free scan checks 10 pages in a real browser against all 90 supported automated rules, separates 27 best-practice checks from its WCAG findings, and names the rule behind every result.