Loading...
The playbook Moonrank's SEO Agent uses to measure page speed the way Google does — on mobile, against real thresholds — and separate the platform problems from the page problems.
Download the methodology, see how the agent reads an LCP breakdown, or run it live inside the SEO Agent.
The Core Web Vitals skill is how Moonrank's SEO Agent measures whether a site is fast enough to rank and to convert. It samples pages from the sitemap, measures LCP, CLS and TBT on mobile, and reads the LCP breakdown to say where the time actually goes — server, image discovery, image weight, or render blocking.
Its most valuable output is usually not the per-page table but the site-wide pattern: an opportunity that appears on a third of the pages is one platform fix, not thirty content tickets. The playbook calls out the trailing-slash redirect mismatch by name, because it is common, invisible, and the cheapest win on most sites.
It's written as a SKILL.md, a plain-language instruction format that people and AI agents can both follow. Read it to run your own audit, hand it to a developer as a brief, or let the SEO Agent run it for you.
“Our Lighthouse score is bad but I don't know where to start. What actually matters?”
Of 50 pages sampled, 31 are Poor and 28 report a redirect before loading — your sitemap uses trailing slashes and your server doesn't, so every crawl pays a redirect first. Fix that one line and the median score moves before you touch a single image. After that, the LCP element on your worst five is a hero image discovered late by the browser: preload it and set fetchpriority high.
The site's sitemap URL, or a sample size if you want more than the default 50 pages.
LCP, CLS, TBT and the performance score per URL, against Google's own thresholds.
The LCP breakdown separates a server problem from an image problem from a render-blocking problem.
Site-wide patterns first, then a numbered fix list for the five worst pages.
Google indexes mobile-first, so the audit does too — no desktop vanity scores.
LCP, CLS, TBT and verdict per page, with failures marked ERROR rather than zero.
TTFB, load delay, load duration and render delay — so the fix targets the real bottleneck.
Opportunities appearing on more than 30% of pages, promoted above per-page work.
The trailing-slash disagreement between sitemap and server, named explicitly when it appears.
How many pages were measured out of how many exist — no extrapolating a site from 2.5% of it.