WebAIM - Web Accessibility In Mind

The WebAIM Million Update
A 6-month accessibility re-analysis of the top 1,000,000 home pages

Introduction

In August 2019, WebAIM conducted a re-analysis of the accessibility of the home pages for the top 1,000,000 web sites. The initial WebAIM Million analysis had been conducted in February - 6 months prior. This report documents changes in accessibility over that 6 month time period. The same sampling techniques and methods were used during this update as were used for the original WebAIM Million. The same caveats discussed in the original research remain.

Note!

The site lookup functionality provides updated metrics based on this August 2019 analysis. It also indicates changes in home page accessibility over the course of these 6 months.

Errors and Error Density

Errors are accessibility issues that are automatically detectable via WAVE, have notable end user impact, and are likely WCAG 2 conformance failures. There was an average of 59.1 errors per home page, a slight decrease from 59.7 errors per home page in February.

Home page complexity increased 4.3% in 6 months - from an average of 783 elements per page to 816 elements per page. 7.2% of home page elements had a detectable error in August 2019, compared to 7.6% of home page elements in February 2019. This decrease is more due to the significant increase in page complexity rather than a decrease in page errors.

WCAG Conformance

The WCAG 2 failure rate based on automatically detectable errors was 98.0% in August compared to 97.8% in February. While slightly more pages had detectable WCAG failures, the number of errors per page decreased.

Most common types of WCAG 2 failures
WCAG Failure Type% of home pages in February% of home pages in August
Low contrast text85.3%86.1%
Missing alternative text for images68.0%67.9%
Empty links58.1%58.9%
Missing form input labels52.8%53.2%
Missing document language33.1%30.5%
Empty buttons25.0%27.0%

The prevalence of pages with low contrast text, empty links, missing labels, and empty buttons increased slightly in 6 months, whereas the prevalence of pages with missing alternative text and missing document language decreased slightly. Simply addressing these few types of issues would have a significant positive impact on web accessibility.

Notable Changes

Changes in accessibility data in this sample were generally minor over the course of 6 months. Minor changes, however, can give us insight into what may become the trajectory and potential future of web accessibility. Below are changes of interest:

  • Use of the HTML5 doctype increased from 74.1% of home pages to 77% of home pages.
  • The prevalence of image alternative text errors decreased by 2.2%. However, 32.9% of all images (12 images per page on average) do not have proper alternative text.
  • The prevalence of form inputs increased from an average of 3.4 per home page in February to 3.6 per home page in August. However, the number of unlabeled form inputs decreased by 4% over that same time period. Pages have more inputs, but more of them are being labelled. Still, 56.8% of all inputs are not properly unlabeled.
  • Implementation of HTML regions and ARIA landmarks is on the rise. The majority of pages had at least one landmark. 25.5% of pages had a main region or landmark in August, compared to 23.5% in February.

ARIA Increases

The most notable change over this 6 month period is the significant increase in ARIA usage. Pages that include ARIA markup increased from 60.1% in February to 64.5% in August. The average number of ARIA attributes per home page with ARIA increased sharply (10.9%) from 37 to 41 in that same period.

Pages with ARIA present have 27 more detectable errors (65% more errors) on average than pages without ARIA. This is up from 26.7 errors per page (61.2% more errors) in February. The correspondence between ARIA usage and number of errors is getting stronger over time.

It is important to note that home pages with ARIA are also much more complex, averaging over 1000 elements per page - over double the number of elements than pages without ARIA present. Pages that utilize JavaScript frameworks and libraries tend to have many more errors, more page elements, and around twice as many ARIA attributes as pages that don't utilize such frameworks.

Technologies

Pages with the following common technologies (present on over 1% of home pages) exhibited decreases in accessibility errors compared to 6 months prior:

  • Angular
  • React
  • RequireJS
  • TweenMax
  • Hammer.js
  • Lodash
  • Modernizr
  • Slick
  • YUI
  • Laravel
  • Rocket CMS
  • Google Maps
  • ReCaptcha
  • ASP.net
  • Google AdSense
  • DoubleClick Ads

Pages with the following common technologies (present on over 1% of home pages) exhibited increases in accessibility errors compared to 6 months prior:

  • Mustache
  • Select2
  • Lazy.js
  • Underscore.js
  • Blogger
  • 1CBitrix CMS
  • Java
  • Criteo Ads

jQuery and WordPress usage decreased from February to August. The average number of accessibility errors per page for both jQuery and WordPress changed only slightly. Home pages utilizing React (35,935 home pages), Slick carousel (43,164), and Lodash (20,577) increased notably.

Conclusion

The number of detectable errors decreased slightly, but the number of home pages with WCAG failures increased slightly over 6 months. It is therefore difficult to say whether these data indicate an improvement in accessibility or not. The notable and alarming increase in page complexity—with increased ARIA implementation—would, if this trend continues, likely make the web more difficult for users. While it is tempting to extrapolate these findings into an increase or decrease in web accessibility more broadly, using these and future data to draw trend lines over time will be more accurate.

WebAIM will regularly analyze a wide swath of the web for accessibility and, as always, will share the results freely. Our next analysis (early 2020) will provide additional data to help provide insight into the current state of web accessibility and trends that develop over time. If you have questions about this research or would like us to analyze the database for something specific, please contact us.