Level
AA
Since
WCAG 2.0
/ WCAG criterion

2.4.5 Multiple Ways

Every page of a site has to be reachable in more than one way, for example through the menu and a search, unless it is a step in a process such as a checkout.

/ What it requires

People look for things in different ways. This criterion asks that every page in a set of web pages can be found by more than one route: the main navigation and a search, a site map, a table of contents, links between related pages, or a home page that links to every page. One route rarely suits everyone. A screen reader or magnifier user may type into a search rather than work through a large menu, and a person with a cognitive disability may prefer a site map that shows the whole structure at once.

A set of web pages is a collection with a common purpose from the same author or organisation, and W3C treats each language version as a separate set. The exception covers a page that is the result of, or a step in, a process: a checkout step, a funds transfer confirmation or a list of search results can only be reached by going through the process, and that is allowed. On a site of three or four pages, W3C says links from and to the home page may be enough, because the home page then also works as a site map.

/ How pages fail it

  • The main menu is the only route to most pages: there is no search, no site map and no index.
  • A site map page exists, but it was last updated two redesigns ago and leaves out whole sections, so those sections have one route left.
  • The site search indexes only the blog, so a product, help or policy page can be reached through the menu alone.
  • Landing pages built for a campaign are linked from nowhere on the site and can be reached only from an advert or an email.
  • A help centre lets people reach an article only by clicking down a tree of categories, with no search and no list of articles.

/ How to check it yourself

  1. Decide what the set of web pages is: pages with one purpose and one owner. Treat each language version separately, and set aside checkout steps, confirmations and search results, which fall under the process exception.
  2. List the routes the pages offer: main navigation, search, a site map page, tables of contents, related links, an index. Two of them have to lead to each page.
  3. Pick a sample of deep pages, such as an old article, a privacy page and a product in a subcategory, and reach each one by two different routes without typing its address.
  4. Search for the exact title of each sample page and check that it comes back in the results.
  5. Open the site map and compare it with the navigation. Look for sections it leaves out and for links that end in a 404.

/ How to fix it

  • Put a search form on every page: a form element with role="search" or the HTML search element, a labelled input of type search and a submit button, and let the search index the whole set of pages.
  • Publish a site map as an ordinary page of links grouped by section, link to it from the footer of every page, and generate it from the same source as the navigation so it cannot fall behind.
  • Add links between related pages: breadcrumbs, a list of related articles, or links in the text to the pages it mentions.
  • Give long documents and publications split across several pages a table of contents that links to every chapter.
  • Link every landing page that stays live after its campaign from at least a section page and the site map.

/ What our scanner checks

No automated check in our scan covers this criterion. Each rule judges a single page, and this criterion is about how the pages of a site connect. The HTML_CodeSniffer sniff for 2.4.5 only raises a notice asking a person to check, and the scan keeps errors alone, so nothing is reported. In a manual audit we follow two routes to a sample of deep pages.

/ Who fails it, in our own measurements

The weekly index has no figures for this criterion, because none of the checks the site scan runs can test it. A person judges it, which is what a manual audit is for.

One page per body, measured by automated checks on a weekly schedule. Any body that has fixed something since the last sweep will show it at the next one, and every entity page carries a right of reply.

/ Questions about this criterion

Does a site need a search box to meet WCAG 2.4.5?

No. Search is one of several ways W3C lists. Any two will do, for example the main navigation and a site map page, or the navigation and links between related pages. On a large site search is often the easiest second route, but the criterion does not name it.

Do checkout pages and search results need multiple ways?

No. The criterion excepts a page that is the result of, or a step in, a process. A checkout step, a payment confirmation or a results page can be reached only by going through the process, and that is allowed. The product pages that lead into the checkout still need two routes.

Does an XML sitemap count as one of the multiple ways?

Not on its own. The W3C site map technique describes a web page of links to the sections of a site, linked from the pages it lists. A sitemap.xml file is written for search engine crawlers and is not offered to people on the site. Keep it for search engines, and publish an HTML site map as well if a site map is meant to be one of the two routes.