Yes.
We get asked this a lot, and the answer is yes, AI can replace humans in roles such as auditors, user testers, consultants, project managers, and more within digital accessibility.
In our recent post, the future of AI in digital accessibility, we talked about hybrid automation. With hybrid automation, the services market evolves into a layered approach of layer 1: automated rule sets / AI and then layer 2: human supplementation (contributing materially to what AI cannot do) and review.
This hybrid layered approach will be commercially available in 2026 for at least a few services and activities in digital accessibility. And we expect hybrid automation to be commercially viable for accessibility audits by 2027.
The next wave of super scans (think traditional automated rule sets combined with artificial intelligence) will be better than current scans, but still won’t be advanced enough to flip the script — where humans review automation and not the other way around.
None of the above commercially eliminates humans from the picture, but once your start to project from here, you can see why there will be less and less accessibility jobs.
| Year & Development Stage | What It Means for You |
|---|---|
| 2025: Humans with AI Tools | AI is leveraged by accessibility professionals and teams to make work more efficient. For example, Accessibility Tracker AI is used by teams to fix issues faster. |
| 2026: Agentic AI Beta | AI agents can think independently and create accessible templates based on WCAG-conformant examples. Early agentic AI is flawed and error-prone causing doubts about its efficacy, but it improves. |
| 2027: Hybrid Automation | The role of accessibility professionals transitions from fully immersed to primarily AI review and supplementation. AI does most of the work while professionals make sure the final deliverable is ready for showtime. |
| 2028-2030: Full Automation | AI can automate most accessibility roles that humans once occupied. AI is so good that it now fulfills most accessibility services. |
Table of Contents
Note: This article concerns real AI in accessibility, not “AI-powered” widgets. Also, hybrid automation does not refer to the current incorrect notion that you combine “automated testing” and manual testing to audit a digital asset.
Conceptually
Before we continue, just stop and think about AI’s capabilities right now.
- AI can understand prompts and respond.
- AI can see and analyze images.
- AI can watch video and interpret.
- AI can listen to audio and process.
- AI can parse code and correct.
How far of a leap do you think it is for AI to do anything on a computer or phone that humans already do?
We’re doing the same things over and over again.
We’re going to a Shopify store, looking for a product, adding the product to the cart, entering our information, and checking out.
We’re opening the YouTube music app on our phone, searching for a song, and pressing play.
If AI can go through these motions, it can also evaluate accessibility while it does so.
Can it learn WCAG?
Of course.
Can it learn practical accessibility and usability issues?
Definitely.
Can it learn how to correct those issues?
Surely.
Really all that’s left for AI is the ability to fuse all of its current capabilities and execute processes more accurately for longer spans (e.g., maintain 99% accuracy over 20, 50, and 100 steps, not just one). This will take time, but it’s clearly within reach.
AI Outlook
Here’s the Accessible.org current AI outlook:
2025: Humans with AI tools (like Tracker AI) speed things up. Time and money saved. Much more efficient than the past. Humans are still completely involved in all aspects of general accessibility (audits and remediation), but now work is being completed much faster. Some AI specialty tools nearly automate specific tasks (e.g., closed captions, alt text descriptions).
2026: Agentic AI rolls out. Beta versions are flawed and error-prone, but show promise. AI can think for itself and carry out more tasks like creating similar code templates based on a template that has been deemed WCAG conformant. Read more on AI agents in digital accessibility. The differentiation with agents is they will execute on tasks on their own (humans won’t be needed to prompt).
2027: Hybrid automation takes hold. Humans still essential, but they’re doing even less work. Their role has been reduced to a review and supplementation capacity.
Here’s a very loose analogy of hybrid automation that is off, but still makes the point:
Do you remember when Walmart and Target used to have a human cashier for every checkout lane? Have you noticed that now there are far less human cashiers and now many more self-checkouts with just one person to oversee them all?
2028 – 2030: AI can automate or nearly automate most of the accessibility services that were once provided by humans.
Why Will AI Replace Humans So Fast?
Because AI can simulate human interaction and replicate the processes humans use to evaluate and fix accessibility issues.
For example, AI can keyboard tab through a website and visually scan the site as it’s doing so. AI can also test with a screen reader and determine where an accessibility issue exists. AI will be able to check for closed captions and audio descriptions (and AI agents will be able to correct them).
These are all behaviors that can be programmed for and mimicked by AI.
And AI is iteratively becoming more intelligent and more capable. The recursive loop for advancement – where AI is now used to train better AI, which in turn creates even more capable AI – is accelerating. Each generation of AI helps build the next generation faster than before. Basically, AI is becoming better, smarter, faster, and it’s doing so at an exponential rate because it’s increasingly involved in its own improvement.
Won’t People Still Want Real Humans to User Test?
No. User testing is expensive and if people can reduce their cost from $550 a user testing session to $5 or even free for an AI session, they’re not going to pay $550.
Realistically, once we reach the point where AI can largely mimic the experience of most users with disabilities, the only use of humans in user testing will be for the sake of involving humans in user testing.
But AI Won’t Have the Lived Experience of a Person with a Disability
This is true, but AI can learn from people with disabilities and other experts on what creates an accessible experience. It’s not that there isn’t value in the experience, it’s that AI will be able to extract and harness that value.
And what forever will be true is AI won’t have the same experience and perspective — but it doesn’t need to to improve usability and accessibility.
AI Doesn’t Have to Be as Good as Humans
Whether it comes to audit services, user testing services, or even remediation, you might argue that humans will be better in some way, form, or fashion. Or maybe that humans will always be needed as a review.
Maybe. Maybe a human will catch those one or two issues that AI overlooks.
But it won’t be enough of a difference.
If AI can automate to nearly flawless accessibility, consumers will be fine with that. Digital experiences will be more accessible than ever, let’s say 99.9% accessible, and the law of diminishing returns will apply.
What About Digital Accessibility Companies?
Will digital accessibility companies even be needed? Maybe, but probably not.
Right now and in the near future, there will be AI sold by companies that’s specialized for helping clients and customers make their digital assets accessible to people with disabilities.
The question of whether accessibility companies go extinct comes down to whether or not AI for accessibility is even capturable — will a company be able to throw a fishing net over AI, bag it, and sell it as “Accessibility AI”?
Our inclination is to say no. AI is like a wild horse. You can’t really tame it to where only you get to reap the rewards commercially.
Yes, maybe for now, but eventually that horse bucks free and anybody can walk up to their favorite AI store and get whatever they want.
Our bet: Over/under 4.5 years — we’ll take the under that there won’t be a need to go to Accessible.org.
What To Do?
The above is horrible news and great news depending who you are.
For accessibility auditors and user testers, they’ll need to find new work.
For people with disabilities, the digital world is about to open up like it never has been for (which, if you think about it, is cause for celebration: the whole point of the industry will be realized in a short amount of time).
For Accessible.org clients and other consumers, the cost of accessibility and compliance will sharply decrease in the coming years (it already has, Accessibility Tracker’s AI tools have reduced the amount of technical support hours clients need from us).
Of course, the digital accessibility marketplace is not alone in this reality: AI is increasing efficiency now and will start replacing more people in the near future.
As for what you can do next as an accessibility professional, look into spaces where humans value and connect with humans vs. where you’re providing digital services that are viewed as mostly a swap of money for digital services.
Here’s an example outside of accessibility: let’s say you write content for SEO purposes. There’s no connection here. It’s just a digital trade. As soon as the buyer can get the same or better article for less from AI, they’re going to drop you.
In the future, AI will eat up virtually all digital services.
On the bright side, currently accessibility services are in higher demand than they’ve ever been due to the European Accessibility Act (EAA), ADA Title II web rule, and ongoing ADA website lawsuits resulting in the ADA website compliance sector.
This gives accessibility professionals a few years to adjust to the new AI realities.
Counterpoints
Let’s look at cars. Specifically electric cars and self-driving cars.
Electric cars were buried for decades after it was thought they would take over. And now that they’ve finally been realized, we still see very practical problems and flaws in execution.
Self-driving cars still aren’t here and Elon Musk has been claiming we’re just 2-3 years away for a decade now.
Another tech example: Bitcoin.
Crypto enthusiasts have been claiming bitcoin would be a part of daily commerce for nearly a decade and BTC still isn’t close to actually being used. All that happens is people speculate and hoard it.
The point is technology can be feasible and on the cusp of realization and yet still not come through in real life. This might also be the case with AI.
This post only reflects what we see as we research and work on integrating rapidly advancing AI technology into our products, services, and operations.
