How to test alternative text (alt-text)

A tester once posted a message in a group chat, stressed that they didn’t know how to achieve something huge! They had been given a catalogue of 3,000 products to check for the presence of alt-text. He didn’t know the most important thing about alt-text. An image only needs a description if it has information or context not available elsewhere. But more on that later. If you are reading and not hearing this, you might not be familiar with alt-text. Alternative text, also written as alt-text with or without the dash, is an image description that sits in the metadata for use by screen readers or text-to-speech technologies.

Back to our stressed friend. Three thousand products. Batteries of all shapes and sizes. AA, AAA, and a few of the chunky type, and rechargeable ones too. They weren't only wondering how to automatically check for alt-text, but also how to write three thousand descriptions. For photos of, essentially, similar objects in slightly different colours and packaging. When I asked, every scrap of information about lithium or alkaline, size, etc., was already sitting right there in the product description underneath.

The part I was able to help with was explaining the myth of alt-text. One that many designers, developers, testers, and others don’t know, because they heard of alt-text for those who can’t see. Thought… Great idea. And have added it ever since. WRONG! 

The myth that makes alt-text so misunderstood 

Here's the myth, and I want to end it once and for all. 

Every image needs alt-text. 

That statement. “Every image needs alt-text”. Is false. Not true. A misunderstanding. A great idea, taken at face value, and thrust forward with impunity. Here is the truth.  

NOT every image needs alt-text. 

Not every image needs alt-text. And I’ll explain why. But before we get to that. I’ll do you one better and give you another, better, more sensible statement. One that, if you take nothing else away from this article, please remember this.  

Every image needs A GOOD REASON TO HAVE alt-text. 

Let me explain why “Every image needs a good reason to have alt-text”. Some images are decorative. They add polish, mood, or visual rhythm to a page, and if you removed them tomorrow, nobody using a screen reader would lose any information. Forcing a description onto those images doesn't help anyone. It adds noise. Screen reader users have told accessibility teams this for years. That a page cluttered with "decorative border graphic" and "abstract blue swirl" read aloud between every paragraph is exhausting, not helpful. Take this blog post image. If you were only reading this and couldn't see the rather cute penguin painting, you would gain nothing by having alternative text. Which is why that image doesn't have any.

So the real question was never "does this image have alt-text?" Rather, it's "does this image carry information that isn't available anywhere else?" For a battery on a product page, with the details already written out in plain text beside it, the honest answer is usually no. Who cares if the battery is gold and black, or red and silver? Is it the right size for my needs?  

The principle is simple. Describe the important information, and if there is none, leave it be. 

What good alt-text actually looks like

RNIB, the Royal National Institute of Blind People, who know a thing or two about this given their whole purpose is supporting blind and partially sighted people, put it simply. Be descriptive. Keep it concise. Give context. And crucially, avoid redundancy. Don't repeat information that's already sitting in the text next to the image. 

Their own example has stuck with me since I first read it. Bad alt-text for a photo of a cat on a windowsill? Just "Cat." That's it. That's the whole description. Technically true, genuinely useless. Good alt-text for the same photo? "A tabby cat sitting on a windowsill, looking out the window with its tail curled around its body." Suddenly you've got a picture in your head, built entirely from words. 

That gap between "Cat" and the full description is where most testing conversations about alt-text go wrong. People assume the fix is always more text. Sometimes the fix is less. Sometimes it's none at all. 

Back to the batteries

So what should my tester friend actually have done with those 3,000 products? 

Ask what each image was doing. If it was purely showing what a fairly standard-looking battery looks like, and the information was already written out as text on the page, the image was 100% without a doubt decorative. A blank alt attribute (alt="") would have done the job, telling the screen reader to skip politely past rather than announce "image" and leave the user none the wiser. 

Side note: If you see alt-text that begins with a variation on “image of” or picture of” or even “image shows”, that is poor. The screen reader announces “image” before reading the text, so first saying it is a picture or image is redundant. That applies to duplicating information, and adding unnecessary details as well. Ok, carry on. 

If, on the other hand, a handful of those products had something genuinely visual going on. It could be unusual packaging. A diagram showing how to insert the battery correctly. A size comparison against a coin. Anything that's informative and not covered in text needs a proper description. Suddenly the task isn't three thousand descriptions. It's a much smaller, much more sensible number. 

That's the shift I want testers to make. Stop asking "has every image got alt-text" and start asking "does every image need it, and if so, is what's there actually useful." 

Why this keeps happening

It's worth knowing you're not alone in bumping into this. In the WebAIM Million survey, an annual scan of the top million homepages, missing alt-text is consistently the second most common accessibility issue found, sitting only behind low-contrast text. Not a rare edge case. Not a niche concern for one team. The second biggest digital accessibility issue on the web, year after year. 

Part of that, I suspect, is exactly the confusion my tester friend ran into. Teams either skip alt-text entirely because nobody owns the decision, or they overcorrect and demand it everywhere, which leads to lazy, repetitive descriptions that fail the RNIB test just as badly as having none at all. Neither extreme serves the person actually relying on that text to understand the page. Getting alt-text wrong isn’t usually because folks don’t care, but because they don’t know or misunderstand what it is there to do. 

Side note: I once had a discussion with a scientist about alt-text on graphs. The scientist was of the very strong opinion that the graphs' alt-text had to be very clear and specific, and cover all the data. After a little back-and-forth, I asked, 

“Is all the data shown in the text of the paper?” Of course they replied. “So what is the purpose of the graph?” To show trends was the response. “So if all the data is available, why doesn’t your alt-text say ‘the graph shows an upward trend, or downwards, or bell curve’ whatever the actual trend is?” It took the scientist a few moments to process this information. Then they smiled. “Thank you,” they said, smiling. “You have just saved me and my colleagues hours!” 

To sum up

Testing alt-text isn't a checklist exercise where every image gets a tick or a cross. It's a judgement call, made from image to image, asking whether information changes or is lost if that image disappeared. Decorative images can, and often should, stay silent. Informative ones deserve real descriptions, written the way RNIB suggest, specific enough that someone could close their eyes and still picture roughly what's there. 

And if you ever find yourself staring down three thousand nearly identical product photos wondering how you'll possibly describe them all, take a breath. You probably don't need to. The information might already be sitting right there, in the text, the whole time.