95.9% of websites in the top million contain accessibility errors for people with disabilities — averaging 56.1 errors per homepage. This figure has grown by 10.1% over the year, breaking a six-year improvement trend for the first time. The reason is the mass adoption of AI tools in development: models were trained on existing code with the same barriers and now reproduce them in new projects.
The average homepage now contains 1,437 elements — 22.5% more than a year ago and nearly double what it was in 2019. Code is generated faster, volume grows, but quality control can't keep pace. When a marketing team uses AI to create landing pages or copywriting, they automatically inherit a problem that was long considered the responsibility of developers.
Why AI tools amplify accessibility barriers
Generative models that write text and generate code were trained on real web pages. The problem is that the internet was never designed to be accessible to a significant portion of users. The six most common error types — low text contrast, missing alt text for images, unlabeled form fields, empty links, empty buttons, and missing language declaration — have topped the ranking for seven consecutive years.
When a model analyzes billions of pages for training, it learns exactly these patterns. Most modern AI tools don't have enough examples of accessible code in their training data. Asking an AI to generate accessible markup is equivalent to asking it to solve a problem it has no foundation for.
The code looks correct, the page renders properly, but the problem only emerges when a screen reader user tries to navigate with the keyboard — by then, the negative experience is already formed.
The New York Bar Association's June 2025 report stated this directly: AI cannot ensure accessibility if it hasn't been trained to recognize accessibility requirements. Errors remain invisible until a real user encounters the barrier — or until a law firm sends a notice of violation.
How lawsuits are growing and why having a tool doesn't protect you
The number of digital accessibility lawsuits in the US has doubled since 2020. In 2025, 26,253 lawsuits were filed at the federal and state levels. 78% of lawsuits target e-commerce. Nearly eight out of ten cases are now being litigated in state courts, where damage awards can scale significantly higher.
Key fact: 38.5% of companies sued for accessibility violations were already using accessibility tools at the time the claim was filed. They believed the problem was solved, but in reality they had only partial coverage and didn't know it. Litigation exposes these gaps publicly, with serious reputational consequences.
This isn't just an American issue. On June 4, 2026, a French court ordered Carrefour to bring carrefour.fr and its mobile app into full compliance with accessibility requirements within six months. This was the first ruling under the European Accessibility Act that rejected partial compliance as a legal defense. Carrefour's own declarations claimed 50–70% compliance across different platforms. The court stated the position clearly: a website cannot be partially accessible — it either works for everyone or it doesn't.
Settling claims typically costs $15,000–$75,000. Legal fees add another $10,000–$30,000. These are the quick cases. One widely cited case against a major retailer started as a routine claim in 2020 and concluded five years later after more than 200 court filings with a total cost of $5.15 million. Defense often costs more than early settlement and significantly more than proactive problem-solving.
What this means for Russian brands and marketing
Russian legislation also establishes digital accessibility requirements. Federal Law No. 419-FZ requires organizations to ensure website accessibility for people with disabilities. While legal enforcement in Russia is less active than in the US, the trend toward stricter control is evident.
For marketing teams, this means accessibility responsibility is no longer purely a technical task. When a brand uses AI to create content, landing pages, or email campaigns, it automatically assumes the risk associated with the quality of the generated code. Missing alt text on images, low contrast on CTA buttons, unlabeled subscription form fields — all of this reduces conversion rates and creates barriers for part of your audience.
Research shows that improving accessibility has a positive impact on all audience segments: clear navigation, intuitive forms, and readable text enhance the overall user experience. This isn't a narrow social initiative — it's a question of product quality and reach to solvent audiences.
Marketer's checklist: how to verify AI-generated content accessibility
- Check text and interface contrast — minimum 4.5:1 ratio for regular text, 3:1 for large text.
- Ensure all images have a meaningful alt attribute. Decorative images should have an empty alt="".
- Test keyboard navigation: all interactive elements must be accessible via Tab, tab order must be logical.
- Check forms: each field must have an explicit label connected via a label element.
- Open the page with a screen reader (built-in VoiceOver on macOS or Narrator on Windows) and verify how content is read aloud.
- Use automated testing tools: WAVE, Axe DevTools, Lighthouse in Chrome — they'll identify basic errors.
- Include accessibility checks in the approval process before publishing materials.
How to measure results and control quality
Automated tools identify 30–40% of accessibility issues. The remainder requires manual testing or real user testing. For brands heavily using AI for content creation, regular audits make sense: monthly checks of key pages and new materials reduce the risk of critical errors accumulating.
Metrics to track:
- Number of automatically detected errors on key pages — target is fewer than 5 per page.
- Percentage of images with properly filled alt text — target is 100%.
- Percentage of forms with explicitly labeled fields — target is 100%.
- Time to complete key scenarios with keyboard — should be comparable to mouse navigation.
Including these parameters in your content publication checklist allows you to catch issues before they become public or legal problems.
Frequently asked questions
Why does AI create websites with accessibility errors
AI models were trained on real web pages, 95.9% of which contain accessibility errors. The models learned these patterns and reproduce them in new code. Without specific training on accessible code examples, AI cannot generate proper markup for users with disabilities.
How to check a website's accessibility before launch
Use a combination of automated tools — WAVE, Axe DevTools, Lighthouse — and manual testing with keyboard and screen reader. Automated tools identify 30–40% of issues; the rest requires testing by real users or accessibility specialists. Include checks in your content approval process.
Does having an accessibility tool protect you from lawsuits
No. 38.5% of companies sued for accessibility violations were already using specialized tools at the time the claim was filed. Partial coverage is not legal protection — a website is either accessible to all users or it isn't. Courts reject partial compliance as a defense argument.
In brief
- 95.9% of top million websites contain accessibility errors, their number has grown 10.1% over the year after six years of improvement.
- AI tools reproduce accessibility problems because they were trained on code with the same errors.
- The number of digital accessibility lawsuits in the US has doubled since 2020, with 78% targeting e-commerce.
- 38.5% of companies that received lawsuits were already using accessibility tools — partial coverage doesn't provide legal protection.
- European courts have begun rejecting partial compliance with accessibility requirements as a defense.
- For Russian brands, accessibility is a matter of legal compliance, audience reach, and product quality.
If your team uses AI to build landing pages or create content, ETC can audit your digital experience for accessibility and reduce reputational risks before your campaign launches.
