Core Web Vitals & site speed, the AI-native way
Google grades every page on three field metrics, measured at the 75th percentile of real visitors. The "good" thresholds (per Google's web.dev): LCP ≤ 2.5s (how fast the main content paints), INP ≤ 200ms (how fast the page responds to taps, INP replaced FID in March 2024), CLS ≤ 0.1 (nothing jumps around while loading).
Passing is only a ranking tie-breaker, worth a position or two between similar pages. But the conversion upside is real: Farfetch's published engineering data showed conversion dropping ~1.3% for every extra 100ms of LCP.
You read the report; AI turns it into a fix list you can actually follow.
Fastest path: one prompt, end to end
Run PageSpeed Insights (pagespeed.web.dev) on your homepage first, then paste the result into this:
🤖 AI prompt — paste into ChatGPT / Claude
You are a web performance engineer explaining fixes to a non-technical store owner.
My platform and theme: [Shopify / WooCommerce / other, plus theme name]
Here is my PageSpeed Insights result for [URL]: [paste the metric values and the whole "Diagnostics" list, or the report JSON]
Do this:
1. Tell me pass/fail against Google's "good" thresholds — LCP ≤ 2.5s, INP ≤ 200ms, CLS ≤ 0.1 — using only the numbers I pasted.
2. Turn every failed audit into ONE markdown table: Fix | What's causing it in my theme | Exact action in plain words (menu clicks before code) | Metric it improves | Effort (S/M/L).
3. Sort by impact and flag anything that truly needs a developer.
Use only the data I pasted — do not guess or estimate any number, and do not invent an audit that isn't in the report. If you're missing something, name exactly what I should re-run in PageSpeed Insights (pagespeed.web.dev) and paste in.
Or do it in 5 steps
- Measure two pages. In PageSpeed Insights (pagespeed.web.dev), test your homepage and your best-selling product page, Mobile tab first. Enough traffic and you'll see a "Core Web Vitals Assessment" from real Chrome users; low-traffic stores only get lab numbers — use those.
- Read the three scores against LCP ≤ 2.5s / INP ≤ 200ms / CLS ≤ 0.1, and note which page fails what. Product pages usually fail harder — more apps, more scripts.
- Fix the usual Shopify/theme culprits, in payoff order. (a) Images: right-sized hero and product images (Shopify's CDN already serves WebP/AVIF, but oversized source files still hurt) and lazy-load everything below the fold. (b) App bloat: every installed app injects scripts — uninstall the ones you don't use and remove leftover code; on most stores this is the single biggest INP win. (c) CLS: set width/height on images and reserve space for banners so nothing shifts.
- Paste the report into the prompt above and work through the prioritized list — do the S-effort items yourself today, queue the developer items.
- Re-test, then wait for field data. Re-run PageSpeed after each change. The field numbers and Search Console's Core Web Vitals report use a rolling 28-day window, so confirm the win there about a month later.
Done looks like: both pages green in the lab test, and Search Console's CWV report trending "Good" a month later.
My rule: fix speed for the customer, not the algorithm — the conversion lift shows up whether or not the ranking boost does.
Cadence: re-test after every theme change or app install; glance at Search Console's CWV report monthly.