For as long as I can remember, styling classes has always been pushed as "best practice" for writing CSS. Avoid IDs because they're too specific. Avoid styling tags directly because it's too broad.
You can't get away from it. It's even baked into most of the linting tools we use every day. It's sensible advice, but I do think it has quietly trained several generations of developers to build things that look correct, yet communicate no semantics.
This is a common WCAG failure that can easily fly under the radar. It's easy to miss, easy to ship, and a lot more common than you'd expect. I've seen it on government websites, e-commerce platforms, and well funded projects that clearly had accessibility in mind. The lang attribute is set correctly. The visible text is properly translated. But buried in the markup, the alt attributes are all hard-coded in English.
Around 6 months ago, I wrote a blog post called Mission Impossible - Accessibility Job Roles. I publicly pulled apart a real job advert, and the vacancy holder reached out. What happened next is not what I expected.
5 reasons why WCAG AA compliance does not mean your website is accessible
Excerpt:
There's a common misconception, that if you meet the Web Content Accessibility Guidelines (WCAG) 2.2, to level AA, then your website is accessible.
WCAG 2.2 AA is often considered the holy grail and the target to aim for. When, in reality, it should just be a baseline or starting point. Because, what many people don't realise, is that you can have a "fully-compliant" website that is inaccessible to most people.
I recently read a super interesting, yet slightly terrifying study, by researchers at the University of Texas at Austin, Texas A&M, and Purdue, which shows that Large Language Models (LLMs) can suffer permanent cognitive damage when they're trained on low-quality, high-engagement social media posts.
Imagine listening to your favourite podcast. You rewind it to go over something you missed, but each time you replay it, it’s somehow different.
This sounds frustrating, right? But, it’s likely this is what would happen if we just stuffed large language models into screen readers, in a lazy attempt to avoid having to publish accessible content.
In my opinion, "best practice" has had its day. It’s time we stop hiding behind it, and start trying to create some urgency around things we know to be accessibility issues. We need terminology that prompts action, not apathy! Because if somebody can’t use your service, it’s not "best practice", it’s a problem!
In this post, I'll break down a real job advert and point out the things that concern me, in the hope it will help you to avoid taking on an accessibility role which is an impossible task, and likely detrimental to your health!
I've developed a weird fascination with challenging language models, thinking up edge-case questions or scenarios to really work out if it can do what is being claimed, or whether it's just reasonably convincing on the surface.
Most language models are useful, but usually, they aren't that great once you drift outside of the realms of common knowledge and into nuanced territory.
So, since the launch of axe Assistant, I've spent a few days testing it out! This post documents my findings!
The problem with AI, especially large language models (LLM), is that most people don't really understand how they work. This is not an accident! It's in many companies' interests to keep the mystery alive! They use smoke and mirrors to market their products as smarter, safer, or more "human" than they actually are.
As humans, we love to personify things, because our only reference point is often ourselves. When something talks like us, we assume it thinks like us too.