# Performance audits

> Run Google PageSpeed audits on your published pages — scores, Core Web Vitals, and one-click AI fixes for what's slowing you down.

Source: https://knowledge.estage.com/workspace/performance/
Part of the ESTAGE knowledge base (https://knowledge.estage.com). Full corpus: https://knowledge.estage.com/llms-full.txt

---
The **Performance** panel runs your live pages through **Google PageSpeed** — the same engine
behind pagespeed.web.dev — and turns the findings into fixes you can hand to the AI with one
click.

Open it with the **Performance** button in the toolbar.

:::note[Publish first]
Audits run against your **live URL**, so your site must be
[published](/workspace/publishing/). If it isn't, the panel asks you to publish and come back.
The same applies to changes: fixed something? **Publish, then re-run** — unpublished edits
won't show up in the score.
:::

## Running an audit

1. Pick a device: **Mobile** or **Desktop** (mobile is the default — it's also what Google
   weighs more).
2. Click **Run audit**.
3. Wait 30–90 seconds while it works through its stages — fetching the page, running
   Lighthouse, analyzing metrics, compiling the report.

The audit measures **the page you're currently viewing** in the builder, so switch pages to
audit each one. Results are kept separately per page and per device.

## Reading the results

### The four scores

Each score is 0–100: **green** (90+) is good, **amber** (50–89) needs work, **red** (below 50)
needs attention.

| Score | What it measures |
| --- | --- |
| **PERF** | Loading and rendering speed |
| **A11Y** | Accessibility — can everyone use the page? |
| **BEST** | Web best practices |
| **SEO** | Technical search-engine readiness |

### Core Web Vitals & metrics

Below the scores, individual metrics are color-coded against Google's thresholds — among them
**LCP** (how fast the main content appears), **CLS** (how much the layout jumps around while
loading), and **TBT** (how long the page stays unresponsive). Green is good; amber and red
show where the problem lives.

## Fixing what it finds — "Ask AI to fix"

The **Opportunities** list shows the top issues, each with a severity dot and the potential
savings (e.g. *−1.2s* or *−340KB*). Every opportunity has an **Ask AI to fix** button:

1. Click it — Genesis builds a focused request from the audit finding.
2. It lands in your chat and the AI applies a fix scoped to that issue.
3. **Publish**, then **Re-run audit** to confirm the score moved.

If the list is empty, you'll see *"No major opportunities"* — the page is in good shape on
that run.

## Cached results and re-running

When you open the panel, it shows your **most recent result** with a note like *"Showing a
cached result from 2 h ago."* Click **Re-run audit** for live numbers — especially after
publishing changes.

## Limits

Audits are rate-limited (PageSpeed is a shared Google service):

- **Per-project hourly limit** — the panel shows a countdown until you can run again.
- **Google's daily quota** — occasionally the upstream service itself is exhausted for the
  day.

Either way you don't have to wait: the panel links to **pagespeed.web.dev** where you can run
the same audit on your URL manually, no limits on our side. The footer's **Full report** link
opens the complete Google report for your page any time.

:::tip[A realistic workflow]
Run mobile first → fix the top 1–2 high-severity opportunities via **Ask AI to fix** → publish
→ re-run. Big scores come from a few big items (usually images and layout shift), not from
chasing every last suggestion.
:::