# Heading Order as a Metric: The Skipped H2 That Breaks Screen Reader Navigation

Source: https://complymetric.com/blog/2026-11-11-heading-order-metric-skipped-levels-screen-reader.html

*A screen reader user opens a page and asks for the headings, and the screen reader reads out the map: H1, H2, H2, H4. The H3 is missing, and the user has to guess what lives between the second H2 and the H4. The heading order is the navigation, and a skipped level is a missing road on the map. The scanner counts it, and the count is the metric.*

## Headings are the map, not the decoration

Sighted users scan a page by its visual hierarchy: the big text, the medium text, the small text. Screen reader users scan the same page by its heading structure, using the headings to jump from section to section without listening to everything in between. The heading order is the table of contents, and it only works when the levels follow each other: H1, then H2s, then H3s under them, in order, with no level skipped.

A skipped level breaks the map in a specific way. The screen reader user hears H2, then H4, and has to decide whether the H4 is a subsection of the H2 or a new section entirely. The answer is in the visual layout, which they cannot see. The heading order is the only information they have, and it is lying to them. WCAG 2.4.6, Headings and Labels, requires headings to describe the topic or purpose, and 1.3.1 requires the structure to be programmatically determinable. A skipped level fails the spirit of both, and the scanner counts it as a failure of heading order.

The [heading order pass rate post](2026-11-06-heading-order-pass-rate-skipped-levels.md) measures this across a site: the percentage of pages with no skipped levels, and the percentage with exactly one H1. The two numbers together are the metric, and they are the numbers a compliance lead should ask for before any other heading statistic.

## What the scanner counts

The scanner counts three things. First, skipped levels: an H3 that follows an H1 with no H2 between them, or an H4 that follows an H2. Second, multiple H1s: a page with two H1s has two maps, and the screen reader user does not know which one is the page. Third, empty headings: a heading element with no text in it, which the screen reader announces as a heading and then says nothing, and the user wonders what they missed.

The most common cause is the template. A content management system ships a heading for the page title, a heading for the sidebar, and a heading for the footer, and the content author writes their own headings without knowing the template already used H1 and H2. The result is a page that looks fine and reads as a maze. The [document language metric](2026-11-10-document-language-metric-screen-reader-pronunciation.md) is the same kind of template failure: a one-line setting the template got wrong, and every page on the site inherits it.

The fix is in the template, not in the content. Set the page title to the only H1, the section headings to H2, and the subsections to H3, and stop the sidebar and footer from using headings at all. Then the content authors cannot break the order, because the order is in the structure, not in their choices. The [skip link coverage metric](2026-11-09-skip-link-coverage-metric.md) is the same discipline: the navigation aid belongs in the template, and the template either provides it or it does not.

## The pass rate that actually means something

The number to track is the percentage of pages with zero skipped levels and exactly one H1. Not the number of heading errors, which grows with the number of pages, but the percentage of pages that pass. A site with a thousand pages and a 95 percent pass rate has fifty broken pages, and the fifty are the ones to fix. A site with a hundred pages and a 60 percent pass rate has a template problem, and the fix is one change, not forty.

Report the pass rate weekly, and report the trend, not the snapshot. A pass rate that climbs from 60 to 95 percent over a month is a template fix working. A pass rate that sits at 95 percent for three months is fifty pages nobody has looked at, and the next fix is a content audit, not a template change. The metric is the map, and the map is only useful if someone reads it.

## Frequently asked questions

### Why does heading order matter for screen reader users?

Screen reader users navigate by headings, jumping from section to section without listening to everything in between. The heading order is the map. A skipped level is a missing road on the map, and the user has to guess what lives between the sections they can see.

### What does a heading order scanner count?

Three things: skipped levels, such as an H4 following an H2 with no H3 between them; multiple H1s on one page; and empty headings with no text. Each one breaks the navigation map in a different way.

### Is a skipped heading level a WCAG failure?

It fails the spirit of WCAG 1.3.1, which requires structure to be programmatically determinable, and 2.4.6, which requires headings to describe topic or purpose. Most scanners flag skipped levels as failures, and the fix is in the template, not the content.

### What heading pass rate should I aim for?

Track the percentage of pages with zero skipped levels and exactly one H1. A pass rate climbing from 60 to 95 percent over a month is a template fix working. A rate stuck at 95 percent for months means the remaining broken pages need a content audit.