Generative artificial intelligence (A.I.) has exciting implications for digital accessibility. As the technology develops, we expect A.I. to address a variety of issues that impact users with disabilities (and as we’ve discussed on this blog, A.I. is already making a difference).
However, with any new technology, it’s important to understand the limitations — particularly when you’re trying to improve experiences for real-life users.
Recently, OpenAI began rolling out image recognition capabilities for its flagship product, ChatGPT. Users can upload an image and ask the chatbot to describe the image.
ChatGPT describes images with remarkable accuracy. Eventually, the tool could allow content creators to quickly generate alternative text (also called alt text) for thousands of images.
But there’s a catch: From a digital accessibility perspective, simply describing an image isn’t always sufficient.
Alternative text is an essential accessibility feature
To better understand the limits of image-recognition A.I., let’s briefly discuss how alt text impacts users.
People with vision disabilities may use screen readers, software that converts text to audio or braille. Screen readers announce the alt text of visual elements, which provides the user with an equivalent experience (provided that the alt text is descriptive).
Accurate alt text also has other benefits:
- The alt text can serve as “stand-in" content when an image fails to load.
- Alt text helps with search engine optimization (SEO) by providing search engines with crawlable content.
- Alt text may help with backend image organization. For example, if you’re creating content on WordPress, you can search through the alt text of the images in your library to find something that fits your article.
The Web Content Accessibility Guidelines (WCAG) requires alternative text as part of its first Success Criterion (SC), “Non Text Content.” To meet that criterion, text alternatives must serve an “equivalent purpose" to the image (or graph, or other non-text content).
That’s an important distinction: To be equivalent, alt text must be accurate and appropriately descriptive, given the context of the website.
Related: Alternative Text: What and Why
Generative A.I. can describe images, but may miss important context
A.I. chatbots can describe most images accurately and even read images of text. But once again, alt text isn’t just about accuracy; it’s about context.
For example, let’s say you’re writing alt text for a photo of a man holding a cat. Generative A.I. tools might describe the image like this:
Man in white shirt holding an orange tabby cat with a blue background.
There’s nothing wrong with that alt text — unless it’s on a veterinary website. If that’s the case, a sighted user might immediately identify the man as a veterinarian. That’s important context that should be part of the alt text.
Better alt text might read:
Veterinarian holding an orange cat.
This description contains fewer details, but it provides more useful information for real human users.
Most current A.I. alt text generators have trouble deciding what users need to know
A.I. tools may also describe every part of an image without considering which components are the most important. In other words, their descriptions might be…well, too descriptive.
For example, this alt text is accurate, but overwhelming for a screen reader user:
A hand is holding a red apple with one green leaf. A drop of dew is rolling down the side of the apple. The background is a mostly clear blue sky with one white cloud.
Depending on the context, “red apple" might be perfectly fine alt text — there’s no need to go into extreme detail.
Of course, you could train generative A.I. tools to recognize the context of each image, but that requires human work.
A.I. will simplify much of the hard work of accessibility, but humans will always play a role
Generative A.I. will become better at recognizing alt text, and current tools are still quite useful. In the near future, screen readers might use generative A.I. to fill in the blanks when content creators leave out alt text.
And since somewhat accurate alt text is arguably better than no alt text, there’s a defensible use case for generative A.I., particularly when websites have thousands of untagged images.
For now, though, we strongly recommend writing alt text yourself. The process only takes a few seconds, and if you get into the habit of considering accessibility when building your content, you’ll be able to avoid other common barriers.
To learn how A.I. will change accessibility, read: Will Generative AI Improve Digital Accessibility?
And if you’re ready to build a long-term strategy for digital compliance, we’re here to help. Send us a message to connect with an expert.