What changed in the November 2026 Index

The November 2026 edition of the Silktide Index is the biggest change to the Index since 2023. Three things are new:

Every website was tested by our new testing engine, the same one that tests every Silktide customer's website. The October edition was the last one tested by the previous engine.
The Index shows two accessibility scores side by side. The V4 score is calculated the way Index scores have been for years. The V5 score is calculated the new way. See V4 and V5 scores.
The Index covers more than accessibility. It now shows how brands appear in AI answers and how people use the web, and it will keep growing.

If your website's score changed between October and November, this page explains why. We expect most V4 scores to stay close to October's, but some will move, and a few will move a lot, even when nothing on the website changed. V5 scores are new, so they have no October score to compare with.

If your score changed

Start with these questions. They cover most of the changes we expect.

Did your website change? A redesign, a new template, a new cookie banner or a new chat widget changes the result on every page that uses it. Fixing issues raises your score; new issues lower it.
Were different pages tested? The Index tests a sample of each website's pages: up to 125, and 25 for most websites, the same number as in previous editions. The pages in the sample can differ from one month to the next, and a different sample can find different issues.
Does your website block or slow down automated visitors? The new engine handles this very differently, and it is the most likely reason a score moves sharply. See Websites that limit our visits.
Was your website affected by a fix to one of our checks? Some checks now find problems they missed before, and others no longer report problems that were not real. See Fixes to our checks.
Are you comparing the same kind of score? Compare V4 with V4. A V5 score is calculated differently, so comparing it with an earlier V4 score tells you nothing about your website.

V4 scores: the same rules, a new engine

The V4 score is calculated exactly as before. The checks, the weight each check carries, and the way those weights combine into a score are unchanged, so a website that did not change, tested on the same pages, should score about the same as in October.

What changed is how pages are loaded and tested. Every website in the November edition was tested again from scratch by the new testing engine, which has been testing every Silktide customer's website since October 2026. In recent months that engine has had hundreds of fixes and reliability improvements, and most of them make no difference on most websites. The ones below are the most likely to move a score.

Websites that limit our visits

Many websites limit how quickly any one visitor can request pages. When a visitor asks for too much, too fast, the website refuses with an error (an HTTP 429, "too many requests"), or shows a security challenge instead of the page. See .

The previous engine did not slow down when a website refused it. Pages that were refused could fail, or the error page could be tested in place of the real one. On some websites that meant too few pages to give a score at all. On others the score was based partly on error pages.

The new engine backs off politely:

When a website refuses, we slow down. We halve the number of pages we request at the same time, wait, and try again. We only give up on a website after it refuses us many times in a row.
A page we still cannot load is recorded as rate limited. It is left out of the score instead of being scored as an error.
Security challenges are recognized. A or a "checking your browser" page from a website's security service is treated as a refusal, not tested as if it were your page. We back off and try again, as we do for any other refusal.
Slow pages are less likely to fail. We no longer wait for decorative background images that are slow to arrive before testing a page, and when a page still cannot be tested, we record why.

For most websites this changes nothing. For websites that limit automated visitors, it means the November score is based on more of their real pages than before. When we tested a sample of 500 Index websites with the new engine, 37 that had too few pages to score in the previous Index now had enough. A smaller number, 13, went the other way: they had a score before, but we could not test enough of their pages this time.

Our tests also now come from different servers than before. A security service that let our previous servers through may challenge the new ones. If your website has no score this month and had one before, check whether a security service or firewall rule is blocking automated visitors.

Fixes to our checks

These fixes change what a check finds, so they change both the V4 and the V5 score. They apply on every page they are relevant to.

4.1.1 Parsing no longer counts toward WCAG 2.2. WCAG 2.2 removed success criterion 4.1.1 Parsing because modern browsers made it obsolete, but our WCAG 2.2 results still included its two checks: duplicate IDs, and lists that do not follow list structure. They no longer count, so websites where those checks found issues score slightly higher.
Contrast is measured correctly for modern CSS colors. Colors written with newer CSS features, such as colors mixed with color-mix() or colors in wide-gamut color spaces, were sometimes misread. Text could be reported as white on white, or not checked at all. These colors are now read correctly, so false contrast failures disappear and real ones are reported with the right colors.
Web components are tested inside. Contrast and target size are now checked inside web components (content in an open shadow root), and links inside them are followed. Websites built with web components may see new issues reported.
Content revealed with CSS is tested. Content that is hidden by a parent element but made visible again with CSS is now tested like any other visible content.
Target size is measured as WCAG defines it. The check for controls that are too small or too close together now uses the 24-pixel circle that WCAG 2.2 describes. Fewer controls fail that were in fact far enough apart.
Focus indicators are recognized in more forms. A focus ring drawn with a CSS pseudo-element (::after) is now recognized as a visible focus indicator.
The Silktide cookie banner no longer hides focus. On websites that use Silktide's own cookie banner, links underneath the banner were reported as hidden behind it. The banner keeps keyboard focus inside it while it is open, so those links cannot be reached until it closes, and they are no longer reported.
One PDF check no longer runs. The check that PDFs give alternative text for every image no longer runs, so it no longer counts toward V4 scores. Websites whose PDFs failed it may score slightly higher.

V5 scores: new, and shown beside V4

The V5 score is new in the November edition. It uses the scoring method of the latest version of Silktide: every score starts at 100 and each check takes off up to a fixed number of points, depending on how badly a website fails it. The V4 score spreads points across a tree of categories. The two methods weigh the same problems differently, so the same website can score higher in one than the other.

For the Index, both scores come from the same test of the same pages, with the same checks. Only the scoring differs. V4 and V5 scores explains how each one is calculated, and how they differ.

Some things to know about V5 scores:

They start with this edition. There are no V5 scores for October or earlier, so trends for V5 begin in November.
A website without a V5 score has no V5 rank. It is never shown as scoring 0.
Rankings can differ. Because V4 and V5 weigh problems differently, a website can rank differently under each.

What happens to V4 scores

For now, the Index shows V4 and V5 scores side by side. V4 is shown first, so that you can keep comparing with earlier editions. We plan to retire V4 scores in 2027, and we will announce when well before it happens.

New in the Index

The Index is no longer only about accessibility.

AI visibility. We ask AI assistants, such as ChatGPT, Gemini and Perplexity, the kind of questions people ask when choosing between brands, such as "What are the best athletic clothing brands?", and search Google for them too. The Index shows which brands appear in the answers, each brand's share of voice over time, and how visible it is on each platform.
How people use the web. Real figures from real visits, measured by Silktide's analytics, that answer questions such as how many people use dark mode, how many ask websites to reduce motion, and how many visits come from AI assistants.

We will keep expanding what the Index covers, edition by edition.

V4 and V5 scores - how each score is calculated for the Index.
How scoring changed in Silktide 5 - the full explanation of the V5 scoring method.
- what levels A, AA and AAA mean.