Creating and publishing content that does not exclude disabled users is a complex and iterative task.

While there is no one-size-fits-all solution to maintaining accessible content, the PyData Sphinx Theme and this documentation site use some techniques to avoid common content shortcomings.


Issues and pull requests to identify or fix accessibility issues on this theme or site are heartily welcomed!

What We’ve Done#


Several of our documentation pages contain metadata (i.e., .. meta:: directives in reStructuredText) giving summaries of the page contents. If you notice a page that lacks metadata, please open a pull request to add it!


  • Our default code highlighting styles are a11y-high-contrast-light and a11y-high-contrast-dark from Quansight-Labs/accessible-pygments. These styles are designed to meet WCAG 2 AA or AAA contrast requirements. If you don’t like the look of our default code highlighting styles, there are several more to choose from at Quansight-Labs/accessible-pygments.

  • We recently revisited the PyData Sphinx theme color palette to ensure that the colors we use meet WCAG 2 AA or AAA contrast requirements.

  • We also re-defined our primary and secondary colors to be more accessible and distinct from semantic colors used to denote success, warning, info, and danger contexts or information.

  • We simplified the color palette and removed some colors that were problematic in meeting WCAG 2 AA or AAA contrast requirements and for certain types of colorblindness.

  • We have improved how we assign text colors to interactive elements such as buttons and dropdowns to ensure that they meet WCAG 2 AA or AAA contrast requirements.

What You Can Do#

Site configuration#

The following sections include recommendations for settings in the file that can positively impact the accessibility of content generated by this theme and Sphinx in general.

Natural Language#

If not using a more robust internationalization approach, specifying at least the baseline natural language will help assistive technology identify if the content is in a language the reader understands.


In your file, specifying the language your docs are written in will propagate to the top-level HTML tag.

  language = "en"

Add a Site Map#

Site maps, usually served from a file called sitemap.xml are a broadly-employed approach to telling programs like search engines and assistive technologies where different content appears on a website.

If using a service like ReadTheDocs, these files will be created for you automatically, but for some other approaches below, it’s handy to generate a sitemap.xml locally or in CI with a tool like sphinx-sitemap.


For a simple site (no extra languages or versions), ensure sphinx-sitemap is installed in your documentation environment, and modify your

    extensions += ["sphinx_sitemap"]

    html_baseurl = os.environ.get("SPHINX_HTML_BASE_URL", "")
    sitemap_locales = [None]
    sitemap_url_scheme = "{link}"

Logo best practices#

If you use both light and dark themes, it’s best to provide a logo that works well in both or to provide an alternative for the dark theme. If you have a logo, you can add alt-text to it by adding the following to your

  "logo": {
      "text": "PyData Theme",
      "image_dark": "_static/logo-dark.svg",
      "alt_text": "PyData Theme home",

Note the use of “home” in the alt text to indicate that the logo is also a link to the home page.

In the Browser#

Several in-browser tools exist for interactively debugging the accessibility of a single page at a time and can be useful during the content development cycle.

Built-in tools#

Most major browsers, including Firefox and Chrome, have accessibility tools built-in as part of their web developer tools. These tools can help to quickly identify accessibility issues and often include links to standards.


tota11y is an open source “bookmarklet” which modifies the currently-loaded page in place and highlights several accessibility issues.


WAVE is a proprietary (but gratis) browser extension which can highlight multiple issues.


Note that automated testing and extensions such as the ones mentioned above will at best catch 30-40% of accessibility issues. They are not a replacement for manual testing and having a perfect score on any of these tools does not mean that the site can be used by disabled users but instead signals that it follows some accessibility best practices.

In Continuous Integration#

Several automated tools are available for assessing glaring accessibility issues across some pages at once, usually with many configurable options.


Lighthouse provides an automated assessment of basic accessibility issues in addition to search engine automation, page performance, and other best practices.


Specifically, foo-software/lighthouse-check-action is run on selected pages from the generated documentation site.