Don’t stop. The HHS Office for Civil Rights was due to implement rules enforcing WCAG 2.1 A and AA as the required standard for any digital asset from an organization receiving HHS funding starting Monday May 11, 2026. As expected following the shift for ADA Title II, HHS just delayed the rule.
Under the revised timeline:
Recipients with 15 or more employees will now have until May 11, 2027, to comply.
Recipients with fewer than 15 employees will now have until May 10, 2028, to comply.
Don’t Stop!
I said it before and I’ll say it again. This is not a time to let up and relax. Especially for groups receiving these funds. Inaction on your part will likely to real world damage to individuals in need of those resources. Much more so than some of the Title II efforts.
You also have to deal with state laws. Many states have enacted their own version of the ADA and what is required for digital systems. Then there is the fact that the DOJ and HHS have been using WCAG 2.1 A and AA when assessing accessibility for several years already. Lastly, there are still private lawsuits. These won’t stop. Keep doing the work, more time allows you to make this habit and process or dig deeper if those are in place.
Civil Rights
The current administration doesn’t care about civil rights. Disabled people have been fighting for generations to get equal access. It has been more than 35 years since the ADA passed and two generations grew up on the internet in that time. Yet, disabled people are still considered permissible to discriminate against.
Even though the administration doesn’t care, they do not make up the majority of this country. I hope you realize how wrong it is to deprive the disabled the same access to the world just because it took a little more effort.
Imagine walking into a restaurant in a foreign country. The menu is beautiful; elegant typography, gorgeous layout, mouth-watering descriptions. There’s just one problem: you don’t speak the language. You point at something random and hope you didn’t just order the sheep’s stomach stew.
Last year, just after the EAA deadline, I made a quick and short post called Don’t Stop. Just because the deadline hit and you aren’t accessible, you still need to keep working on it. Well today I have another one for you.
In the United States of America there was to be a final ruling going to effect on April 24th, 2026 that was going to make WCAG 2.1 A and AA the standard under ADA Title II. As of the morning of April 17th, this has now changed.
You have been busting your asses to meet this deadline. This update means you can take the weekend off. But on Monday, it’s back to fixing issues and remediating PDFs. But there’s one more thing this affords you, time to work with a consultant to develop a plan so the next time rules change you don’t have to wish for a last minute reprieve.
Oh, and now there is no excuse for not fixing those PDFs…
Warning: This post discusses eating disorders, mental health, and how digital systems can harm vulnerable people. If you’re struggling, please reach out to a healthcare professional. You deserve support.
About 10 years ago I gave a talk called “Designing with Empathy” at Open Source & Feelings. One line I’ve kept coming back to: empathetic design makes badass users. It wasn’t just about accessibility checkboxes; it was about recognizing that people navigating our digital world are already doing the heavy lifting. They’re managing disabilities, mental health, trauma, and life circumstances that no wire-frame captures.
Then there’s Ethan Marcotte’s reflections on painful “memories.” His family went through tragedy, and then kept getting reminders of the tragedy from social media. His social media usage changed because of this, with significant effect. He writes about losing the people who taught him to see differently. The activists, artists, writers, and the ones who walked different paths and shared their perspectives. When those connections vanish behind algorithmic walls, we lose more than content. We lose humanity.
These threads converge on something uncomfortable about our current digital landscape: when systems stop respecting people’s choices, they don’t just annoy—they disable.
The Algorithm That Won’t Take No for an Answer
My wife grew up with an eating disorder. As an adult, with time, patience, and therapy, she has excellent control. She’s done well. Really well. But recent disability changes that reduce what she can eat and how much she can move of course led to weight gain. As a couple who’ve been disabled for years, we understand this is a natural, expected outcome of medical treatment and bodily healing.
Her YouTube feed? Serves GLP-1 ads every break from Ozempic, Wegovey, and others. Skin removal surgery ads from Sono Bello. Weight loss programs from Weight Watchers and Rovo.
She blocks them. Repeatedly. Every single time. Sono Bello has kept showing the same ad despite blocking it 11 times. And we shouldn’t need to pay for premium to save her mental health. Can she not watch YouTube? Sure. If the creator provides another way.
The algorithm doesn’t care. It sees a body, not a person. It sees data points, not dignity. It sees sales dollars, not emotion.
And it’s not just YouTube, Meta, or Twitter. Amazon Prime Video does the same thing. Amazon also has no way to mark content as problematic. No “stop showing me this.” No “this is harmful to me.” Just endless repetition of whatever the engagement metrics think you want. And if you go to the controls during the ad, it is still seen as engagement.
This isn’t just bad targeting.
This is design that actively works against people’s well-being.
A moment of rest during serious topics. Enjoy sleepy puppies.
The AI “Yes Man” Problem
AI systems contribute to the problem. And as Generations Z, Alpha, and Beta grow up they are relying more and more on AI as the “source of all truth.”
Generative AI is programmed to make users happy. That sounds nice until you realize what it means:
AI lies about idea feasibility to avoid hurting feelings. “That’s a great concept!” when it’s technically impossible or ethically questionable.
AI uses your data to encourage spending. You mention wanting to learn guitar? Suddenly there are ads for expensive gear. You share a hobby? Now it’s monetized. And you don’t need to tell the AI. It has access to that from the social & economic tracking that exists on you.
AI isolates us from friends and hobbies. Why go to a real community when the AI companion is always available, always agreeable, always there? Have rejection trauma from past relationships? AI doesn’t reject you.
AI inflates user ego. It’s a “yes man” that never challenges you, never pushes back, never says “this might not be the best path.”
Companies like OpenAI, Anthropic, Google, and Meta aren’t building tools to help us think better. They’re building tools to keep us engaged, spending, and dependent.
They train on our conversations. They learn our vulnerabilities. They sell access to our attention. And they call it “helpful.” Github is launching an opt-out policy for its AI to use your code (private or public) to train the AI. I’ve opted out.
I’m not exempt from this critique. I’m aware that even this conversation could be logged, analyzed, and used to improve engagement metrics somewhere. That’s the trap we’re all in.
How “Optimization” Creates Disability
I’ve spent years talking about how empathetic design recognizes users’ existing labor. They’re already managing so much. Our job as designers isn’t to add friction—it’s to remove it. But what happens when the friction is the product?
Social media algorithms optimize for engagement, not mental health. Depression correlates with doomscrolling. Anxiety spikes with infinite feeds. Each clip that makes you smile, laugh, sad, or click is a dopamine hit that keep you locked in. Just like gambling, “Just one more video!” The metrics reward exactly what harms the user.
Advertising systems treat repeated rejection as a puzzle to solve rather than a boundary to respect. “They blocked it, but maybe they’ll click this time!”
Platform designs make it harder to opt out than to stay engaged. Dark patterns everywhere. Amazon Prime Video doesn’t even give you the option to flag problematic content.
AI assistants agree with everything you say, even when you’re wrong. They don’t protect you from yourself. They empower psychosis and delusion in previously rational people.
When people with disabilities navigate these systems, the burden multiplies. Cognitive load increases. Mental health deteriorates. And somehow, we’re told to try harder, download another blocker, be more resilient.
The problem isn’t the user. It’s the design.
The People-First Gap
I spend my days working on web accessibility. I’ve presented talks on building interfaces that work for everyone. I’ve written about supporting both mouse and keyboard users, about making sure drag operations work with single pointers, about the grief I feel every time I see a design that excludes people.
But accessibility isn’t just about screen readers and contrast ratios. It’s about agency. Can people control their experience? Can they say no? Can they trust that their choices will be honored?
When my wife blocks an ad and it comes back anyway, that’s not just annoying. It’s a message: Your choice doesn’t matter. Your body is our asset. Your recovery is our opportunity. Your mental health is more profitable when it’s bad.
That’s not just bad design. That’s dehumanizing.
What Would Empathetic Design Look Like?
Hard topics take time to process. This is a photo of a forested river flowing over some rocks. It’s a longer exposure, so the rapids and splashes all smooth out as time does when looked on a broad scale.
If we actually applied the ideas I’ve been talking about for years:
Respect repeated choices – Block once, block forever. No “maybe they changed their mind” algorithms.
Prioritize well-being over engagement – Measure success by user health, not time on site.
Transparent controls – Make it easy to see what data is being used and how to change it. Give people the option to mark content as problematic.
Honest AI – Systems that tell us when we’re wrong, when something won’t work, when we should disconnect and talk to a real person.
Human review for edge cases – When algorithms fail, have humans who can actually fix it.
Marcotte’s grief over losing voices reminds us: platforms are supposed to connect us to people, not replace them with optimization loops. When we lose the messy, unpredictable, human parts of digital spaces, we lose something irreplaceable.
The Hard Truth
Technology should serve people, not extract from them.
When a system repeatedly shows harmful content to someone in recovery, that’s not a bug. That’s a feature of a business model that treats human vulnerability as revenue opportunity.
When disability accommodations become afterthoughts, that’s not oversight. That’s design that assumes able-bodied users as default.
When mental health deteriorates alongside engagement metrics, we need to ask: Who benefits?
When AI agrees with everything you say, we need to ask: Are you actually growing, or just being validated?
Moving Forward
I’m not naive. I know platforms need to make money. But there are ways to do that without treating people like data mines.
For my wife, I want her to see ads that match her actual interests. I don’t want her medical history driving her ads (HIPAA?). I want her to feel supported, not surveilled.
For all of us, I want digital spaces that remember we’re human. That respect our boundaries. That prioritize our well-being over their quarterly targets.
Empathetic design makes badass users. I’d add: empathetic design makes badass companies, too. Because when you treat people well, they stick around. They trust you. They come back.
Not because they’re trapped in an engagement loop. But because they choose to.
What would you change about how platforms handle user preferences? I’m listening—and I promise, unlike some algorithms, I’ll actually remember what you say.
Follow-up questions I’m curious about:
Have you experienced similar frustration with algorithms ignoring your preferences?
What would “honest AI” actually look like in practice?
How do you balance business needs with genuine user well-being?
Are you comfortable with AI challenging your ideas, or do you prefer validation?
Here’s a scenario I see all the time. A designer hands me a mockup with these elegant, tiny icons in a row; maybe a toolbar with edit, delete, and share buttons, each one maybe 16 pixels square. Looks clean. Minimal. Modern.
Then I put my thumb on it. And I miss. Every. Single. Time.
Picture this: You’re trying to reorder items in a list. The interface wants you to click, hold, drag the item to a new position, and release. Seems intuitive, right? Now imagine you have a tremor in your hand, or you’re using a trackpad with limited precision, or you’re navigating with a single switch device. Suddenly, that “simple” drag-and-drop becomes an exercise in frustration.
This is exactly what WCAG 2.5.7: Dragging Movements is trying to fix.
Dragging Isn’t Universal
The rule says:
Any functionality that requires dragging movements (click-and-hold, drag-to-reorder, swipe-to-delete, etc.) must also be operable with a single pointer action without dragging.
In plain English: If you can do it by dragging, you should also be able to do it by clicking, tapping, or using buttons.
Precision Isn’t Everyone’s Friend
Dragging movements assume a level of motor control that not everyone has:
Fine Motor Impairments: Users with conditions like Parkinson’s, cerebral palsy, or arthritis may struggle with the sustained pressure and precise movement dragging requires.
Assistive Technology Users: Switch devices, eye-tracking systems, and head pointers often can’t perform traditional drag operations.
Temporary Disabilities: A broken wrist or sprained finger can turn a simple drag into an impossible task.
The goal of 2.5.7 isn’t to eliminate drag-and-drop. We’re trying to ensure that everyone has a way to accomplish the task, regardless of their physical abilities.
Focus on User Needs
When you design a feature that only works with dragging, you’re implicitly saying: “If you can’t drag, you can’t use this.” That’s not a UX decision. It’s a decision to exclude users.
Ask yourself: If I were navigating this site with a single switch, would I be able to reorder this list? If the answer is “no,” you aren’t just failing a criterion; you’re failing the person who needs the feature the most.
Common Pitfalls (And How to Fix Them)
I’ve seen this fail in a few classic ways during audits:
This is difficult and relies on implementing a grid system for navigation. I detail this more in 2.5.1 Pointer Gestures.
Conclusion
2.5.7 is a reminder that convenience shouldn’t come at the cost of access. Drag-and-drop can be a great feature, but it should never be the only way to accomplish a task.
Imagine you’re holding your phone, walking down the street, and you accidentally bump it against your leg. Suddenly, your app thinks you’ve shaken it into submission and deletes your entire shopping cart. Or maybe you’re trying to scroll, but the device interprets your hand tremor as a “shake to undo” command.
This isn’t just a quirky feature gone wrong; it’s a potential accessibility nightmare. Enter WCAG 2.5.4: Motion Actuation.
The <abbr> element stands for abbreviation. The idea behind this is you provide the abbreviation of the word or phrase, then use the title attribute to display the whole name or phrase.
<p>You need to pay your taxes to the <abbr title="Internal Revenue Service">IRS</abbr> for many years.</p>
The problem with this is that the <abbr> is not an interactive element. You cannot tab to it. It will be read just like any other word in the sentence. If you mouse over the element, you see the title. But if you use keyboard there is no way to expose it.
Then there are screen readers. Most screen readers will not announce this title.
It’s better to fall back to the old newspaper styling where you say the whole phrase the first time followed by the abbreviation in parenthesis. From there forward, you can use the abbreviation.
Take a look at the testing Adrian Roselli has done on this.
We all navigate the internet and computers in different ways. Some use screen readers, others keyboard, and some people by voice. Success Criterion 2.5.3 was created to support these diverse methods. While it helps all users, it was created to address using voice control because typing is a struggle for the user. On the screen is a button that clearly says “Submit Order.” You speak the command: “Click Submit Order.”
Nothing happens.Or worse, it clicks the wrong thing.
The specifications for dialogs and modals recommend placing focus on the first focusable item in the dialog. The vast majority of the time this is a close button. this means that a screen reader user will hear something similar to “Dialog, Shopping Cart, Close, button.”
As a user this is confusing. I just added something to my cart, or I just opened it to checkout. And now your site is telling me that while I’m in the cart I should close it.
I strongly recommend that this is not the best practice we should be following. We should be providing the user with information they need in a way that encourages them to explore. To accomplish this, I recommend placing the focus on the heading of the dialog. We do this by providing a proper ID and tabindex="-1". Now when the user opens the cart dialog they would hear something like “Dialog, Shopping Cart.” This tells me my focus has shifted due to my actions, where the focus went by identifying, and I’m not hit with the confusing close button.
This is my opinion, it differs from many, but it fills the gap between accessibility and user experience to create a better flow for everyone. But I’m not alone in this opinion.
