Wren Beta — what that means

dpreview.com

readable

A Excellent overall on the measured subjects

Loading speed is diagnostic only; it does not affect this grade or search ranking.

SubjectWeightScore
Interruption 35.71% not scored
Readability 28.57% 98
Accessibility and keyboard 21.43% 92
Stability and weight 14.29% 87

This letter rests on 64% of the grade. The subjects marked “not measured” are left out rather than scored as zero (which would fail a site for something nobody looked at) or as full marks (which would be a claim we have not earned). How this is worked out.

dpreview.com
Largest paint
0.3s
Paywall
none
Ad requests
198
Tracker requests
43
Page weight
2.2 MB
Readability
18px type · 1.5 line height · 65 characters per line · 18.1:1 contrast
Accessibility
1 automated check failed (1 serious)
Green hosting
not measured
Measured
2026-09-01

Observed interruptions

Ad and tracker counts above are network requests, not counts of visible ads or separate companies. They do not affect the interruption score.

Interruption evidence is incomplete or needs review, so this subject is not scored. Missing evidence is not treated as an interruption-free page.

The type, and who can use it

Measured on 1 rendered page; latest measurement 2026-09-05. Every number is a mechanical fact about the page, not a judgment of its design.

WhatMeasuredThe bar
Body type size 18px, 16px at phone width 16px, the browser default
Characters per line 65 45–90
Line height 1.48 1.5 (WCAG 1.4.8)
Text contrast 18.1:1 4.5:1 (WCAG AA), 7:1 (AAA)
Text column 51% of the window at least 45%
Reflows at phone width yes WCAG 1.4.10
Dark mode no noted, not scored
Automated accessibility checks 1 automated check failed (1 serious): color-contrast WCAG 2.1 A and AA, via axe-core
Keyboard no trap; focus visible on 92% of stops never trapped, always visible
Structure main landmark, 1 h1, no skipped heading levels a main landmark, one h1, no skips
Alt text 100% of 1 content images described every content image
Reader mode parses parses (Firefox’s Readability)

Automated accessibility checks catch an estimated 30–40% of real problems, so “no automated violations detected” is a floor, not a certificate. We never say “accessible”. How these are scored.

Tripped on some pages, but not most

GateWhat tripped
G8Detected ad area over 30% of a sampled viewport

At least one page we rendered failed this outright, but most did not, so the site is indexed. Which page you land on may well decide whether you can read it.

Measurement dates

Oldest evidence 50 days old. At least one subject is due for rechecking (30 days or older). Dates refer to the evidence, never the card rebuild.

Measurement history

Newest 4 saved observations. Each describes a sampled page at that time; older, failed or held observations do not override the current reviewed card. Raw automated findings may have been superseded by later review or detector replay. No previous grade is inferred.

Read dated observations
  1. — Readability and accessibility observation; prose source p; desktop type 17.6px; phone overflow not observed
    https://www.dpreview.com/articles/dpreview-mini-challenge-begins-on-september-1-for-save-your-photos-month/
    Evidence source: readability-top5000.jsonl · observation 21a1777bada1
  2. — Access observation: excluded; detected gates G8; review required
    https://www.dpreview.com/opinion/how-visiting-a-car-show-made-me-feel-better-about-the-future-of-photography/
    Evidence source: verdicts-top5000.jsonl · observation 3e288990e4b2
  3. — Access observation: pass; review required
    https://www.dpreview.com/articles/dpreview-mini-challenge-begins-on-september-1-for-save-your-photos-month/
    Evidence source: verdicts-top5000.jsonl · observation b9d4f5bf79dc
  4. — Access observation: pass; review required
    https://www.dpreview.com/photography/6-photographers-share-the-mistakes-theyre-glad-they-made/
    Evidence source: verdicts-top5000.jsonl · observation 7c87416bd3b9

How we know

Pages rendered3
Site verdicterrors dropped; majority of remaining; ties to the stricter verdict; active reviewed exclusions override admission
Figures shownnumeric measurements generally use medians; interruption costs, overflow, keyboard traps and accessibility failures retain the strongest observed finding
Cited by1,941 encyclopedia articles
Flaggedat least one render was degraded, so this card is queued for re-measurement

Think this is wrong? It may well be — the measurements cover only a handful of pages. Any reviewed admission decision and its dated evidence are shown above. Re-measurement on request is not built yet, so there is no button here that would do anything. When it exists it will be free and it will be on this page.