Not long ago, Anthropic engineer Thariq mentioned that, under the assumption that “humans no longer manually edit code,” HTML is more suitable than Markdown for interacting with AI.
That made me wonder: if AI is reading a high-quality accessible website, would it be easier for the AI to “understand” the website, and then produce better results when interacting with humans?
First, let’s quickly look at the benefits of good web accessibility
Here are a few points from Taiwan’s “Accessible Web Development Service”:
Five major benefits of adopting accessibility Legal compliance: Meeting domestic legal requirements and protecting the rights of people with disabilities is our responsibility and obligation. Expanding service reach: Reaching around 1.2 million people with disabilities, as well as older adults, helps increase service coverage. Corporate social responsibility (CSR): Showing a commitment to social inclusion helps improve an organization’s public image and trust. Better user experience: Good accessible design makes websites easier to use and improves satisfaction for all users. SEO optimization: Accessible websites with clear structure are more search-engine friendly and can help improve rankings.
The first point is a basic protection. Points 2, 3, and 4 are mostly about “serving people.” But for point 5, I think we should no longer see it only as “SEO optimization.” It should now become AXO optimization.
How does web accessibility help AI interact with websites better?
When an AI Agent operates a website, what it fears most is not an ugly interface. What it fears is not knowing: “What is this?” “Can I click this?” “What happens after I click it?”
In the past, when we added a proper label to an input field, it was to help screen readers understand what that field was for. But now, AI Agents may also rely on that label to judge: “Is this a search field?” “Should this field contain a name, an email, or an address?”
There are many things humans can understand at a glance, but AI may not.
If accessibility is done well, AI Agents can use the accessibility tree to get cleaner, more concise information that is closer to what they actually need for interaction. LogRocket BlogLink also mentioned that this helps AI Agents avoid having to guess their way through a messy DOM full of noise.
Take booking tickets as an example. While website A is still forcing the AI to parse a bunch of unclear elements, website B might already allow the AI to complete the booking successfully. For users, website B is faster. For AI service providers, website B also saves more resources.
What changes will this bring?
No Hacks mentioned in the article “HOW AI AGENTS SEE YOUR WEBSITE (AND HOW TO BUILD FOR THEM)”:
Last year, before the European Accessibility Act took effect, I half-joked that it would be ironic if the thing that finally got people to care about accessibility was AI agents, not the people accessibility was designed for. That's no longer a joke.
I find this line really interesting, because it points out a new reality: accessibility technology is no longer only serving humans. It is also helping all kinds of “non-visual users” understand websites.
Because of this, the biggest change may be that more companies will start paying attention to accessibility.
But the motivation may shift from “legal compliance” to “business interest.”
If you do not do it, you may fail to serve users who rely on AI Agents to complete tasks. And that means you may lose business opportunities that could have turned into actual conversions.
Compared with SEO, AI Agents may have a more direct impact on revenue.
SEO is more like bringing people to the destination. After that, the website still needs to rely on content, flow, and persuasion to convert users. But AI Agents may directly help users get things done.
This also leads to a second impact: website analytics may change dramatically. As AI Agent usage grows, familiar metrics like “click-through rate,” “time on page,” and “conversion rate” may all need to be redefined.
This shift in motivation could theoretically create healthy competition, and it may also improve the overall accessibility environment on the web.
After all, the first step is still to increase the number of websites that actually meet accessibility standards.
So, are the users of web accessibility really changing?
I don’t think they are changing. I think the scope is expanding, just like what happened with SEO in the past.
Although AI is now bringing a new wave of attention to accessibility, the focus will eventually come back to humans. A website is not just code, and it is not just an accessibility tree. It also includes flow design, content understanding, emotional experience, trust-building, and the real experience users have at every step.
So yes, AI may encourage more people to take accessibility seriously again. But the true core of accessibility is still helping people use websites more smoothly and more easily.