Observed performance per page — server response time, page weight, request count, compression and caching.
9 active of 15 published rules · each active rule is versioned and backed by evidence. Planned rules are on the roadmap.
The server took over a second to return the HTML, delaying everything that follows.
What it checks
Measures time-to-fetch the HTML; flags responses over 1000 ms.
Confidence
High — derived from evidence quality, not asserted.
The HTML document itself is large (over 500 KB), which slows first render.
What it checks
Measures the HTML document size; flags documents over 500 KB.
Confidence
High — derived from evidence quality, not asserted.
The page references a large number of scripts, stylesheets and images (over 80), each a separate request.
What it checks
Counts referenced scripts, stylesheets and images; flags pages over 80.
Confidence
High — derived from evidence quality, not asserted.
A script in the <head> has neither async nor defer, so the browser must download and run it before it can render the page.
What it checks
Counts <head> <script src> elements that are not async, defer, or type="module".
Confidence
High — derived from evidence quality, not asserted.
The <head> links a large number of stylesheets (over three), each of which blocks the first render.
What it checks
Counts <head> <link rel="stylesheet"> elements; flags more than three.
Confidence
High — derived from evidence quality, not asserted.
One or more images omit width/height attributes, which forces the browser to reflow and contributes to layout shift.
What it checks
Counts <img> elements with a src but no explicit width or height attribute.
Confidence
High — derived from evidence quality, not asserted.
The page carries many images but none use loading="lazy", so every image is fetched up front.
What it checks
On pages with five or more images, checks whether any image uses loading="lazy".
Confidence
Medium — derived from evidence quality, not asserted.
The page has a very large DOM (over 1,500 elements), which slows style, layout and rendering work.
What it checks
Counts all elements in the document; flags more than 1,500.
Confidence
High — derived from evidence quality, not asserted.
The page loads a large number of scripts from other origins (over five), each adding a third-party connection and execution cost.
What it checks
Counts <script src> whose host differs from the page host; flags more than five.
Confidence
High — derived from evidence quality, not asserted.
The main content takes too long to render (LCP over 2.5 s).
What it checks
Needs the headless-render service or CrUX/PSI field data — not measurable from server HTML.
Status
Published in the catalogue; not yet evaluated by the scanner, so it never produces a finding.
Visible elements move during load (CLS above 0.1). Image dimensions (PERF-006) are only a partial proxy.
What it checks
Needs the headless-render service to measure actual layout movement.
Status
Published in the catalogue; not yet evaluated by the scanner, so it never produces a finding.
The page is slow to respond to user interaction (INP above 200 ms).
What it checks
Needs field data or a headless interaction trace — not observable from static HTML.
Status
Published in the catalogue; not yet evaluated by the scanner, so it never produces a finding.
Long main-thread tasks block input during load (TBT above 200 ms / Speed Index high).
What it checks
Needs the headless-render service to trace main-thread work.
Status
Published in the catalogue; not yet evaluated by the scanner, so it never produces a finding.
The HTML is served without gzip/brotli compression, inflating transfer size.
What it checks
Needs the captured Content-Encoding response header, which is not yet stored on the snapshot.
Status
Published in the catalogue; not yet evaluated by the scanner, so it never produces a finding.
Static assets are served without Cache-Control/Expires, so browsers cannot reuse them.
What it checks
Needs captured per-asset response headers, which the bounded MVP crawler does not yet fetch.
Status
Published in the catalogue; not yet evaluated by the scanner, so it never produces a finding.