Manually typing passwords is a burden and is slowly being abstracted or going away. Disabled buttons for sign up, login, password resets are often a lazy solution with giant possibility to block users out. Do not use disabled buttons please.
Deliberate typo in the title, you know, with all the AI hype out there. I guess you have already seen AI powered accessibility compliance solution promises and let us take a step back and have a reality check.
Some countries have already begun to 'show their teeth' regarding accessibility ignorance, but many others are still waiting. It is time for harmonization on this level as well.
Google Lightouse measure accessibility scores for quite some time. New agentic browsing scoring partially re-uses the same audit. Will this encourage real accessibility? It has the potential, but only when human understands accessibility.
Canvas element in HTML enables tremendous possibilities - from two dimensional graphic interfaces to amazing three dimensional experiences, directly in the browser. It will soon be simpler to sync the visuals with the semantics.
If you use AI agents - you still need to be able to take over manually. WebMCP is a standard proposal for a new in-browser AI control and it still requires accessibility, it can make us more efficient, but all users need to have the possibility to take control at any time.
Staying curious and learning more about AI and where it can help with accessibility, I see a lot of work done with bolting it on instead of shifting it left. Seems that shift left needs us once again.
Perhaps well-intended, the auto-scrolling content can be a real nightmare for multiple groups of users. I've done a mobile app audit and turning on screen reader exposed quite a messy experience - and yes, web can suffer from the same issue.
Regressions happen, but proper processes can prevent a lot of it. With accessibility it may be more difficult, but systems can help. Fixing one bug and introducing another one reveals poor system processes - consider accessibility governance.
Thanks to Michael Fairchild and Microsoft, we now have some accessibility benchmarks for a selected range of AI models - where they were tested with automatic accessibility tools after generating sample code. Project is called A11y LLM Eval, and it provides quite some insights, both directly and indirectly.
Model Context Protocol (MCP) is not perfect, but it is a good start - as it can provide reality for Artificial Intelligence (AI). Less hallucinations help with inclusion.
Long story short - all political party websites have issues, some are really obvious and all are excluding some groups of people. They should do better if they want to get votes from as many as possible. And shift accessibility left does not mean political left!