/notes/measuring-the-right-pixels/

0
critical
0
serious
3
moderate
0
minor
42
checks passed
No automated violations found on this page for the selected standard. Manual review still recommended.

Keyboard and focus order (20 stops)

This is the only lane that drives the page instead of reading it: we pressed Tab and recorded where focus actually landed, until it wrapped. Roles and names come from Chrome's own accessibility tree, the same computation it hands a screen reader. A keyboard user passes 10 stops before reaching main content, and the page has a working skip link.

Nothing in this lane is a failure. The keyboard order was walked end to end and the deterministic checks came back clean.

The links list, as a screen reader user gets it 20 links, 16 distinct names

Screen reader users navigate by pulling up every link on the page at once (Insert+F7 in NVDA and JAWS), stripped of the layout that surrounds them. This is that list. Read in place a link can borrow meaning from the heading above it; read here it has only its own name.

LinksAll announce as
3bekee gibson
2Say hello: contact page
2what a free scan can and can’t tell you

Names appearing more than once, most repeated first. A repeated name is not automatically wrong: a "Home" link in the header and the footer is fine. It matters when the destinations differ and the name does not.

Heading outline, as announced

What this looks like on the page
The page with every keyboard stop numbered in the order focus visits it. Red badges mark stops that are invisible when focused, have no accessible name, or show no focus indicator.
Every keyboard stop, numbered in the order focus actually visits it. Red marks a stop that is itself a problem. Open full size (new tab)
The same page with its layout preserved but its content replaced by what a screen reader announces. Hatched areas carry no information for a screen reader user.
The same layout, showing what is announced instead of what is rendered. Hatched areas carry nothing for a screen reader user: background images, decorative graphics, and media with no text alternative. Interactive elements show the name Chrome computed; everything else is an illustrative approximation, not a finding. Open full size (new tab)

AI-assisted findings (needs human confirmation)

These are judgment calls about link purpose and heading clarity that automated tools cannot make. A model flagged them; confirm before acting.

moderate Heading: The pile.AcceptedDeliberate editorial voice; the section headings on the notes are written to read, not to label.

SC 2.4.6 (AA) AI-assisted

This heading does not tell readers what the section covers. 'The pile' is vague and gives no indication of the subject matter, whether that is a collection of color swatches, measurements, tools, or something else entirely.

Suggested: Use a heading that names what was piled up, for example: 'The pile of color samples' or 'The stack of background textures I was testing'.

moderate Heading: So I measured them.AcceptedDeliberate editorial voice; the section headings on the notes are written to read, not to label.

SC 2.4.6 (AA) AI-assisted

This heading begins mid-narrative with 'So', reading as a continuation of a thought rather than a label for the section. A reader scanning headings cannot tell what was measured or why, making the section's subject unknown without reading the surrounding body text.

Suggested: Name what was measured, for example: 'So I measured the background pixels directly' or 'Measuring the pixels beneath each letter'.

moderate Heading: What an audit is.AcceptedDeliberate editorial voice.

SC 2.4.6 (AA) AI-assisted

This heading is too generic to describe the specific section that follows it. 'What an audit is' could appear on countless accessibility pages and tells a scanning reader nothing about what this particular section says about audits or how it relates to the article's topic of pixel-level contrast measurement.

Suggested: Make the heading specific to this article's angle, for example: 'What an audit catches that a scanner misses' or 'Why pixel-level checking is part of a real audit'.