In August 2026, independent French developer Théo Ducreux presented the results of a large-scale web code quality audit in the ValidateHTML project. The study covered the 5,000 most popular domains in the world, and the results were shocking for the industry: nearly nine out of ten sites contain errors that violate basic World Wide Web standards. This points to a systemic crisis in development quality, where the speed of product release is often prioritized over technical correctness and accessibility.
The Scale of the Problem: 100,000 Violations
During the analysis of the sample of 5,000 domains, it was found that only 2,656 of them are full-fledged websites intended for humans. The remaining addresses were used as technical domains (for example, Google's "gstatic" infrastructure). Despite this, even among "human" sites, the situation is critical. In total, more than 100,305 violations of HTML code correctness were found on all checked resources, including HTML5 specifications, as well as nearly 19,000 CSS errors.
The statistics of perfection look even more alarming: only 12.8% of sites were deemed to have correct code. Completely "clean" resources that did not produce a single error or warning were only 2.6%. This means that the vast majority of internet users interact daily with technically incorrect pages, which can lead to unpredictable content rendering on different devices.
Accessibility Crisis for People with Disabilities
One of the most serious problems identified during the check was the inaccessibility of web resources for people with visual impairments. More than a third of sites failed the accessibility check, meaning they are incorrectly processed by screen reader software. W3C and WHATWG standards were created specifically to ensure identical interpretation of content by all browsers and assistive technologies, but in practice, these norms are ignored.
Key accessibility metrics are also at critically low levels. Alternative text (alt text) for images, which often contains vital information for blind users, is missing on 20.4% of sites. Moreover, on 41.6% of pages, ARIA labels are completely absent from the code — special attributes that identify page areas and help navigation for screen readers.
"Guilty" Browsers and Frameworks
Experts identify two main culprits in the current situation: developers and the browsers themselves. Invalid code is often a byproduct of working with modern frameworks. The most common error is incorrect tag nesting, which is technically impermissible but often found in generated code.
However, the main reason is the "unpretentiousness" of modern browsers. They are so smart that they can ignore code errors and automatically restore the desired page layout. This creates an illusion of safety for developers: if the site looks normal in Chrome or Safari, everything is fine. Unfortunately, screen reader software does not possess such "resilience" and requires strict adherence to standards to correctly voice content to the user.
