Information conveyed by color alone

We look for places where colour alone carries a distinction and nothing else does: typically a link outside the navigation that differs from the surrounding text only by colour and has no underline, or a status conveyed only as green and red. The report gives the count of places found on that page.

For links, since 19 September 2026 we compare the colour against the surrounding text (the parent element). A link that has the same colour as the text around it is not distinguished by colour at all, so it gets no finding; until then it did, which was the claim upside down. We also leave out an element with a background or border of its own (a button built on a link, where the area and the border are non-colour signals) and an element that is off-screen or the size of a screen-reader label (typically a “Skip to content” link, an accessibility feature rather than a barrier).

What the check does not do (and here the list matters especially, because this is a heuristic rather than a measurement):

• It does not understand meaning. It cannot tell whether the colour conveys anything at all. Coloured text used purely decoratively can trigger the finding even though it distinguishes nothing.
• It cannot see context. A green “In stock” badge carries the information in words too, but if the word is in an image or the distinction rests on shape, the check will not count it.
• It does not count icons and shapes as an alternative signal.
• It does not check states: our check cannot judge a link that underlines only on hover. But let us not make the standard stricter than it is: under WCAG this need not be a failure at all. Sufficient technique G183 permits distinguishing an inline link by colour alone, provided its contrast against the surrounding text is at least 3:1. A permanent underline is our recommendation, because it is more reliable. It is not a requirement of the standard.

Treat this finding as a prompt to look, not as a list of defects. The finding is page-level; severity is warning.

For visibility in AI answers there is no effect. But it has a standard behind it, one of the oldest and most fundamental in accessibility: WCAG 1.4.1 (Use of Color), level A. That is the lowest, most binding level.

The standard puts it simply: colour must not be the only visual means of conveying information, indicating an action or distinguishing an element. It does not forbid using colour. It forbids relying on it alone.

Who it affects: people with a colour vision deficiency, people reading on a monochrome screen or in sunlight, and screen reader users, to whom colour conveys nothing at all, because nobody reads it out.

The commonest case is a link without an underline inside a paragraph. It looks clean and is common in designs. And for somebody who cannot tell the colour apart, it is text they do not know is clickable.

Where our check is weak: it is a heuristic and cannot read intent. Look through the findings before changing any styles. Some will be false alarms and some will reveal something you did not know.

A rule that resolves most cases: add a second signal to the colour. It need not be prominent, it needs to be present.

For links in body text this often becomes a trade-off between appearance and clarity. A compromise that works: a subtler underline rather than none. A thinner line or a lighter underline colour keeps the look and keeps the signal.

p a {
  text-decoration: underline;
  text-decoration-thickness: 1px;
  text-underline-offset: 3px;
}

You can check this without any tool: take a screenshot of the page and convert it to greyscale. What cannot be told apart in grey cannot be told apart by somebody who does not see colour.