AI Diagnosis

A Chrome Lighthouse audit tells you everything that's wrong with a page, but not what matters most. It presents you with estimated savings that are overly optimistic, and it doesn't help you prioritize a fix.

Our AI diagnosis feature is the judgement that Lighthouse reports are missing: it suggests that one change that will actually move the needle, the evidence from your audit that backs it, and the fix that works for your stack.

Running a diagnosis

Your organization's first successful audit is diagnosed automatically.

For any other successful audit, open its detail page from the dashboard and click Diagnose.

What it looks at

The diagnosis reads the audit like a seasoned performance engineer:

• The network waterfall as a sequence: what blocked first paint, what was discovered late, and whether a critical request was actually starved by the other bytes in flight alongside it.
• Server response times across the audit's own runs and across your audit history. A server that answers in 260ms most of the time and 2 seconds when it starts cold has a different root cause than one that is slow on every load.
• Which element actually won Largest Contentful Paint, and whether that is even settled. A page whose hero image and heading trade places between loads has an LCP race, and fixing the image alone won't end it.
• Your audit history, so one bad load doesn't get diagnosed as a chronic problem.

Next steps

• Getting Started — add a page and run your first audit
• Audit Stability — how xcelera produces the stable measurements a diagnosis relies on
• Performance Monitoring — reading trends once the root cause is fixed