Media performance
Image and font performance, measured in a real render
Four image rules and three font rules measured in the rendered page: whether images reserve their space, ship a responsive source, stay under the weight line and hint their loading; whether fonts avoid invisible text, reflow when they swap, and cache for at least a day. Each warning is a risk signal, and an absent measurement stays na.
Last reviewed 2026-11-18
The free page scan measures one URL; scheduled audits and the monitoring suite are plan features — see pricing.
What we measure
Seven rules cover the media a page cannot render without: four for images and three for webfonts. Each is a warning-level signal measured in the local render lane, and each keeps its own evidence instead of a composite score.
Image rules judge layout reservation, responsive sources, transfer weight and loading hints. Font rules judge the display strategy, the fallback metric difference and the cache lifetime of the files. The table states what each signal can and cannot tell you.
| Signal | What we measure | What it cannot tell you |
|---|---|---|
| Image layout reservation | In-flow images with width and height attributes, or a computed aspect-ratio that reserves space. | Whether the reserved box matches the final paint: real layout shift is not measured. |
| Responsive sources | Displayed width, natural size, device pixel ratio and whether a non-empty srcset is declared. | Whether the declared candidates are well chosen or cover every viewport. |
| Image weight | Compressed transfer bytes per image, and the total across the first 100 sampled images. | What the file costs after CDN re-encoding or at a different viewport. |
| Loading hints | Document position against the viewport plus the loading and decoding attributes. | Actual LCP timing, or whether the browser honoured the hint on a real connection. |
| Font display | font-display on loaded webfonts whose text is used above the fold. | How long text stayed invisible: the invisible period is not timed. |
| Fallback metrics | Width, x-height and line-box difference between webfont text and its rendered fallback. | The real CLS score, or the reflow on a device whose fallback font differs. |
| Font cache | Effective freshness from Cache-Control on each 2xx font file: missing or zero directives read 0. | CDN and browser behaviour beyond the response header: revalidation is not observed. |
The full rule reference — criteria, failure examples and fix steps for every audit check — is in the check reference.
Image checks: four rules
The audit measures images in the rendered page, not in the HTML source: layout reservation, a responsive source, transfer weight and loading hints. The image performance explainer gives each rule a failure example and a three-step fix.
Layout reservation
Checks: In-flow <img> elements: width and height attributes, or a computed aspect-ratio, must reserve layout space. Absolutely positioned, zero-sized and sub-16 px images are skipped.
Warns when: An in-flow image has no width or height attribute and its computed aspect-ratio is auto.
Fix: Set the intrinsic width and height attributes, or an explicit aspect-ratio, so the box is held before the bytes arrive.
Responsive source
Checks: A bitmap displayed at 200 CSS px or wider with no srcset. Its natural width or height is compared with the displayed size × DPR × 1.5; SVG and images with a non-empty srcset are skipped.
Warns when: Natural width or height is at or above 1.5× the displayed size × DPR.
Fix: Add srcset and sizes so the browser can pick a candidate near the rendered size and pixel density.
Transfer weight
Checks: Compressed transfer bytes for each sampled image, plus the total across the first 100 sampled images.
Warns when: One image is at or above 500 KiB (512,000 bytes), or the sample total is at or above 2 MiB (2,097,152 bytes).
Fix: Compress, convert to a modern format and resize to the largest rendered size; lazy-load only below the fold.
Loading hints
Checks: Document position against the viewport: documentTop < viewportHeight counts as above the fold. Above the fold, loading=lazy warns; below the fold, neither loading=lazy nor decoding=async warns.
Warns when: An above-the-fold image declares loading=lazy, or a below-the-fold image has neither lazy loading nor async decoding.
Fix: Keep first-viewport images eager and give below-the-fold images loading=lazy with decoding=async.
Font checks: three rules
Webfonts decide whether first-screen text is visible and whether it moves when the file arrives. The audit reads the loaded faces and their responses; the font performance explainer covers each failure with its fix steps.
Font display
Checks: The font-display value on each @font-face webfont whose status is loaded and whose text is used above the fold.
Warns when: font-display is auto or block. swap, fallback and optional pass.
Fix: Use font-display: swap (or fallback/optional) so text renders with a fallback instead of waiting invisibly.
Fallback metrics
Checks: For above-the-fold webfont text, the width, x-height and line-box difference against the fallback actually rendered. Text whose family is optional is skipped.
Warns when: Any difference crosses its threshold: width 5%, x-height 8% or line box 3 px.
Fix: Choose a closer fallback and apply size-adjust, ascent-override or descent-override so the fallback metrics match the webfont.
Font cache
Checks: Effective freshness from Cache-Control on every 2xx font file response.
Warns when: Freshness is under one day (86,400 s). no-store, no-cache and must-revalidate without a positive max-age all read as 0.
Fix: Serve font files with Cache-Control: public, max-age=31536000, immutable, and keep the URL stable across deploys.
When the checks run
All seven rules depend on a real render. The lane loads the page in headless Chrome, records image geometry and font metrics, then judges that sample. Outside the lane there is nothing to judge, so the rules say so instead of guessing.
- Local render lane: the rules run and return pass or warn.
- Static crawl or Browser Run: the local lane is off, so all seven record na with an insufficient signal reason.
- Lane disabled: the same result — na, never pass.
- Single-page re-crawl (M2) and MCP verifyPage: always na; they do not run the media measurements.
What these checks do not claim
The boundary is part of the feature, not a gap that later copy will quietly fill.
- A warning is not a ranking factor. The thresholds are community practice, not a Google ranking system, and the audit never turns them into an LCP, INP or CLS score.
- The audit renders on Linux headless. Its fallback fonts differ from a user's device, so a width, x-height or line-box difference is a proxy for reflow risk, not the jump a specific visitor sees.
- Only <img> elements in the main document are measured. CSS background images, <canvas>, inline SVG, iframes and Shadow DOM are out of scope.
- Image bytes are compressed transfer bytes, and the sample is the first 100 images. A page with more images is judged on that sample.
- No automatic fix. The rules name the risk and the fix direction; the page still has to be changed.
Frequently asked questions
How the checks run
When do these checks run?
Only when the audit runs in the local render lane. Static-HTML crawls, Browser Run audits and runs with the lane disabled record na with an insufficient signal reason for all seven rules — an absent measurement is never reported as pass. Single-page re-crawls and the MCP verifyPage tool are always na.
Is a warning a ranking penalty?
No. Every rule here is a warning-level signal measured against community thresholds, not a ranking factor in Google's systems. It flags layout and loading risk; it does not predict rank, and the audit does not turn it into a score.
Which images does the audit look at?
Only <img> elements in the main document, and only the first 100 sampled images. CSS background images, <canvas> output, inline SVG, images inside iframes and Shadow DOM are not measured, so a clean result does not cover them.
Fixing what it finds
How do I fix an image that reserves no space?
Add the intrinsic width and height attributes, or an explicit aspect-ratio, so the browser holds the box before the file arrives. The image explainer walks through the attribute and CSS variants in three steps.
How do I fix an oversized image?
Ship a responsive source: add srcset and sizes so a bitmap shown at 200 CSS px or wider is not downloaded at 1.5× or more of its displayed size × DPR, and compress it under 500 KiB per image and 2 MiB across the sample.
Why does a font warn when it looks fine on my machine?
The audit renders on Linux headless, whose fallback fonts differ from the ones on a user's device. A width difference of 5%, x-height difference of 8% or line-box difference of 3 px against that fallback warns; fix it with a closer fallback, size-adjust overrides, font-display swap/fallback/optional and at least one day of font cache.
Measure the media layer
Run the page scan, then read the image and font rules
Seven warning-level rules, measured in the local render lane.