Check the alt text of a page’s images

How to use the tool

Enter the address of a public page and click “Check”. The tool’s server downloads the page’s HTML, pulls out the img tags and returns only the list of images, not the page. For a page that is not live yet, or sits behind a login, open “Paste HTML instead”: the analysis then runs in your browser and nothing is sent.

Each row of the table shows a preview of the image, its full address, the state of the alt attribute (missing, empty or its text), then what the tool found: declared width and height, the loading attribute, format from the file extension, whether it sits in a picture element, and the link around it. The last column holds the remarks. The “Show only images with a remark” box hides rows without an alert, and the CSV button exports the whole table.

The tool lists up to three hundred images per page. The tools on this site accept twenty requests per ten-minute window and per visitor, all tools combined.

The tool’s remarks

Each remark rests on Google or W3C documentation, except two rules the tool sets itself and says so.

• Missing alt. The image has no alt attribute. For a decorative image the W3C decision tree asks for an empty alt; for an image that carries information, a text that conveys its meaning.² Without the attribute, neither is said.
• Empty alt. The tool counts it without judging it: an empty alt is the right answer for a purely decorative image, for text that is only a visual effect and for an image whose content already appears in the nearby text.² It is up to you to check that these images really are decorative.
• Image in a link with no name. The link holds no text, has no aria-label or title, and the image has an empty or missing alt. An image used as a control needs a name that describes its purpose.³
• Alt that looks like a file name. “IMG_1234.jpg”, “cat.jpg” or the file’s own name. The W3C lists file names among the texts that do not replace the image.⁴
• Placeholder alt. “image”, “photo”, “picture”, “spacer” or “photo 1”: same failure, same source.⁴
• Same alt on several images. A rule of the tool: identical text on different images cannot describe each one.
• Repeated word. A word of three letters or more that comes back three times in the same alt. Google advises against filling alt with keywords.¹ It sets no threshold: the three occurrences are the tool’s choice, meant to catch the obvious cases.

Dimensions, the loading attribute and the format are shown without a verdict. Only Google’s documentation on format is used: it lists BMP, GIF, JPEG, PNG, WebP, SVG and AVIF and asks that the file extension match the file type.¹

Empty alt or missing alt

The two cases do not say the same thing. A missing alt leaves the question open: nobody knows whether the author forgot the image or judged it irrelevant. An empty alt, written alt="", declares the image decorative and tells assistive technology it can skip it.

The W3C decision tree sorts images into five families.² A decorative image gets an empty alt. An informative image gets a brief description of what it adds. An image of text repeats that text. A functional image, an icon or a button, states its function, destination or action. A complex image, such as a chart, gets a short alt and the full information elsewhere on the page.

WCAG success criterion 1.1.1, level A, requires an equivalent text alternative for any non-text content, and expects pure decoration to be ignored by assistive technology.³ Adding text to such items gets in the way of screen reader users. Choosing between decorative and informative remains the author’s judgment, based on why the image is on the page.⁶ The tool cannot make that call for you.

Writing a useful alt

Google asks for information-rich content that uses keywords appropriately and in the context of the page.¹ Its documentation gives four examples: no alt is a bad example, an alt stuffed with keywords is another, “puppy” is better, and “Dalmatian puppy playing fetch” is best. Keyword stuffing creates a poor user experience and may make the site look like spam.

The W3C describes the alt of an informative image as a short phrase or sentence, without setting a limit in words or characters.⁶ It adds that the text is typically not a literal description of the image but its meaning on the page. A product photo does not carry the same alt on a product page and in a blog post. That is why the tool shows no length verdict.

Google adds that images should sit near relevant text, on a page whose subject matches the image.¹ For the sharing image, the og:image:alt Open Graph property plays the same role; the meta tags and Open Graph generator prepares it.

An image inside a link

When an image is a link, its alt becomes the name of the link. Google notes that the alt text then serves as anchor text.¹ The W3C asks a simple question: would the link be hard or impossible to understand without the image? If so, the alt describes the destination or action. If the image repeats real text next to it, such as a logo followed by the company name in the same link, the alt stays empty.²

The W3C’s H2 technique recommends putting the icon and the text in a single a element, with an empty alt on the image, so keyboard and screen reader users do not meet two identical links side by side.⁵ The tool therefore does not flag an empty alt in a link that holds text.

What the tool does not see

The tool reads the HTML as the server sends it. It does not see images added by JavaScript after loading, nor CSS background images, which the W3C names as an alternative for decorative images.² Inline SVG, canvas elements and image-type buttons are not listed. For a page that builds its images in the browser, open the developer tools, copy the HTML of the rendered page and paste it into “Paste HTML instead”.

For lazy-loaded images, the tool reads the address from src, then data-src, data-lazy-src, data-original, and failing that the first address of srcset. It does not judge the quality of an alt: a text that looks fine to a machine can miss the point, and only a human sees it. The weight, real dimensions and metadata of one particular file belong to another tool, the image metadata reader.

Frequently asked questions