Describe the image don't just label it

There is a little box in Joomla wherever you add an image that says “Image Description (Alt Text)”. You put some words in it, save your article and congratulate yourself because you have now made your website accessible.

Except you haven't.

Well, perhaps you have. But there is a good chance that what you have actually done is label the image rather than describe it. And those two things are not the same.

We have reduced the whole question of alternative text to one very simple test: is there something in the alt attribute? If the answer is yes, then apparently we can all go home. The accessibility box has been ticked and the job is done.

The person using a screen reader doesn't care whether the box has been ticked. They care about what the image is telling them.

The difference between a label and a description

Imagine that you are reading an article about web accessibility. There is a photograph of somebody sitting at a laptop. The alt text says “Person using laptop”.

There is nothing technically wrong with those words. They aren't a lie. There is, indeed, a person using a laptop. But they are almost completely useless.

Why is the photograph there?

Perhaps the person is using a screen reader. Perhaps they are testing a website. Perhaps the article is about remote working. Perhaps it is about computer security. Perhaps it is about the ergonomics of working from home.

The photograph doesn't tell us why it was selected. The person writing the article does.

If the purpose of the photograph is to show somebody using a screen reader, then “Person using laptop” isn't really an alternative to the image. It is simply a label attached to it.

“Developer testing a website using a screen reader” actually communicates something.

The difference isn't the number of words. It is the information those words convey.

We've become obsessed with describing what an image contains when what we should really be doing is describing what the image contributes.

It may seem like a subtle distinction, but it changes the way you write alt text completely.

The problem with “Image of…”

One of the habits I find particularly irritating is starting alt text with “Image of” or “Photo of”.

The screen reader already knows it is an image. That's its job. Telling someone who is listening to a screen reader that they have encountered an image is like standing next to someone looking at a photograph and announcing, “This is a photograph.”

No shit, Sherlock.

What they don't know is what the photograph communicates.

If I write “Photo of a woman sitting at a computer”, I've told you what you could reasonably infer from the picture.

If I write “Woman using a screen reader to navigate the Joomla administrator interface”, I've given you information that the image was intended to communicate.

That is the difference between describing the contents of an image and providing an alternative to the image.

A screenshot isn't useful just because you name it

Anyone who writes software documentation will have seen this.

You include a screenshot to show somebody where they need to click. You add alt text saying “Joomla Media Manager screenshot” and move on.

You've just told someone that the screenshot is a screenshot of the Joomla Media Manager. You haven't told them what the screenshot is showing.

The person who can see the screenshot can look at it and find the button. They can see the field. They can see the option that you are talking about. The person who cannot see it gets told that it is a screenshot of the Media Manager.

Thanks, but no thanks.

The screenshot is there to demonstrate that the Image Description field is found in a particular part of the interface. That's the useful information. That's what needs communicating.

Once you start thinking about alt text in this way, you realise that there isn't a universal description for an image. The appropriate text depends on why the image exists in that particular piece of content.

AI can see the image. It can't necessarily see the reason.

Modern AI systems are remarkably good at looking at an image and describing what they see. Give one a photograph of somebody sitting at a desk and it might tell you about the person, the laptop, the desk, the coffee cup and everything else in the frame.

That's impressive. But it doesn't tell you why the photograph was chosen or what it is there to illustrate.

The same photograph could be used in an article about remote working, accessibility, cybersecurity, office furniture or the alarming amount of coffee consumed by developers. The visual content is identical. The editorial purpose isn't.

AI can be useful. There is nothing wrong with using it to get a starting point, particularly when the alternative is an empty alt attribute or, even worse, somebody typing “image” into the field because they have been told it can't be left blank.

But the generated description should still be treated as something to review, not as an accessibility certificate.

The accessibility checker can't read your mind

And this exposes another problem with the way we test accessibility.

Automated accessibility tools are useful because they can check things that are relatively straightforward to measure. They can inspect the HTML and tell us whether an image has an alt attribute. They can identify an image where the attribute is missing, and some tools can flag text that looks suspiciously generic.

But they cannot reliably tell us whether the words we have entered are actually the right words.

If I put alt="image" on every photograph on my website, an automated test can see that the alt attribute exists. The HTML is valid. As far as that particular test is concerned, I have done what was required.

If I put alt="Person using laptop" on an image that is actually being used to demonstrate somebody using a screen reader, the test can see that there is text. It cannot understand the editorial purpose of the photograph well enough to decide that I have completely missed the point.

We can automate the checking of whether we have provided an alt description. We cannot reliably automate the judgement that the description is actually meaningful in the context of the page.

A green accessibility report does not necessarily mean that you have written good alt text. It can simply mean that you have successfully satisfied the things the automated test knows how to measure.

There is a world of difference between those two things.

This isn't an argument against automated accessibility testing. Far from it. Automated testing is incredibly useful. It catches errors that humans would otherwise miss and gives developers a way to identify problems across an entire website. It can tell you that an image is missing an alt attribute, which is considerably better than not checking at all.

What it can't tell you is whether you have thoughtfully described what the image contributes to the content or whether you have simply typed something into the box to make the warning disappear.

The software can check whether the box has something in it.

It can't check whether you understood why the image is there.

That still requires a human.

Sometimes the best description is nothing

There is another reason why I dislike the idea that every image must have some meaningful-sounding text attached to it.

Sometimes the image doesn't communicate anything. It is decoration.

Maybe it is a stock photograph at the top of an article. Maybe it is a background image. Maybe it is there because the page looked a bit boring without it.

If the image doesn't add information, there is no reason to invent some.

An empty alt="" is perfectly legitimate. It tells assistive technology that the image can be ignored.

That is considerably more useful than forcing somebody to listen to “Decorative image” every time they encounter one.

The purpose of accessibility isn't to make sure every field in the CMS contains something. The purpose is to make sure people aren't excluded from the information being communicated.

Sometimes that means adding information. Sometimes it means not adding noise.

Think about what the image is doing

Don't start by looking at the image and asking yourself, “What can I see?” Start with the article and ask yourself, “Why is this image here?”

If you removed the image, what would somebody who cannot see it lose? That is what your alt text needs to replace.

It might be a description of what somebody is doing. It might be the information contained in a chart. It might be the point demonstrated by a screenshot. It might be the name of an organisation represented by a logo.

Or it might be nothing at all because the image was never carrying any information in the first place.

None of this requires complicated accessibility technology. Joomla already provides the field. The browser already understands the alt attribute. Screen readers already know how to use it.

The hard part is us.

And I'm not going to pretend that I've suddenly solved this problem either. I still catch myself writing descriptions that are really just labels. Sometimes I look at an image and my first instinct is still to describe what I can see rather than think about why the image is there.

The difference is that I'm now much more aware of the problem. I try to stop, think about what the image is contributing to the article and then write the description accordingly. I don't always get it right, but I try.

Ultimately, that's all anyone can really be asked to do: be aware of the problem, think about the people who will experience our content differently from us, and make a genuine effort to describe rather than simply label.

We have to stop thinking of alt text as something we put on an image and start thinking of it as part of the content itself.

Because that is what it is.

It isn't a label. It isn't a caption. It isn't a box that needs filling in so that an accessibility checker turns green.

It is the bit of your article that needs to survive when the image doesn't.

So describe the image. Tell your readers what the image is telling us.

Don't just label it.