Understand image and mobile rendering checks
Interpret broken images, image-profile changes, viewport, and overflow regressions.
Last checked 21 August 2026
Primary page checks use a 390x844 mobile viewport. TechDash combines capped first-hop image probes with rendered image and layout measurements. A pass can mean no captured URLs, unchanged signals, absent baseline features, or no qualifying failure. It does not prove every image works, has appropriate alt text, or renders well on every device.
Image URL errors and redirects
Images with HTTP errors is Warning because broken assets harm users and can affect image discovery and loading performance. It fails in absolute state when a captured internal img[src] returns a concrete broken status (for example 404 or 500), or an external image returns 404 or 410. Rate limits (429), auth or WAF blocks, timeouts, and soft overload (503) are treated as crawl noise and do not open this incident. Images that redirect is Advisory and fails when an internal image returns any first-hop 3xx, or an external image redirects to a different host, excluding known social and share hosts.
Both checks pass when no image URLs were captured or none meets the trigger. Evidence lists the source URL with its error or first-hop destination. TechDash does not follow the redirect. It probes up to 100 unique URLs, reduced to 40 under backlog, and excludes unsafe, non-HTTP, blocks, rate limits, and ambiguous external results. A pass therefore does not cover every image on a large page or prove successful decoding, dimensions, format, visual quality, lazy loading, or Largest Contentful Paint.
Signed URLs, hotlink protection, consent variants, or an intentional CDN migration can explain a finding, but users still benefit from direct stable URLs. Restore files, permissions, and case-correct paths, or update src to the final destination. Run Check & Verify; recovery requires a later capture with no qualifying rows.
Image profile regression
Image profile collapsed is Advisory. It matters because a frontend regression can remove meaningful product, editorial, or accessibility signals without producing an HTTP error. A prior profile is eligible when it had at least three rendered images, a detected hero source, or at least 50% alt coverage. After that baseline, the check fails if any trigger is true:
- Image count falls to half or less from a prior count of at least three.
- A previously detected hero source becomes absent.
- Alt coverage falls by at least 40 percentage points from a prior coverage of at least 50%.
A pass means no trigger fired or the prior profile was not eligible. It does not mean every image is useful or every alt value is appropriate. Evidence shows only prior and current image counts and alt percentages. It does not show which images disappeared, either hero URL, the hero-detection reason, or individual alt text. Hero loss can therefore trigger an incident without being visible in the displayed evidence.
Lazy loading, personalization, consent, stock state, or an intentional visual redesign can be benign. Decorative images may correctly have empty alt text, so do not add meaningless labels to improve a percentage. Compare screenshots and rendered DOM, inspect image components and loading behavior, and restore only intended imagery and accurate alternatives. Check & Verify recovers when the profile no longer meets a collapse trigger or an intentional baseline is accepted.
Mobile rendering regression
Mobile viewport meta regression is Warning. TechDash treats viewport content as valid when it contains width=device-width after whitespace removal or a numeric width of two to four digits. It fails only when a prior valid viewport becomes missing or invalid. A first invalid viewport passes as baseline, and an unchanged invalid viewport can continue to pass.
Mobile horizontal overflow regression is Advisory. It fails when prior horizontal overflow was 10 pixels or less and current overflow is more than 10 pixels. Evidence shows prior and current pixel measurements. A pass means that boundary was not crossed, including pages that already overflowed by more than 10 pixels.
These narrow rendered signals do not identify the overflowing element or test every breakpoint, device, orientation, interaction, overlap, tap target, font size, accessibility behavior, or real-user Core Web Vitals. Intentional wide canvases or tables can be benign when usable, but accidental off-screen content can block mobile users and browser agents.
Restore a valid viewport meta when removed. Reproduce at 390 pixels, inspect fixed widths, tables, carousels, translated text, margins, and positioned elements, and fix the responsible layout rather than hiding useful content. Run Check & Verify. Recovery requires a valid viewport and overflow at 10 pixels or less relative to the accepted baseline; separately test real devices and interactions.