The AI site audit: what data-fed AI finds that prompts can't

Type “audit my website: mysite.com.au” into a chatbot and you'll get a professional-sounding report in thirty seconds. It's worth what it cost. The chatbot never crawled your site — it pattern-matched what's usually wrong with websites and presented the average as your diagnosis.

The teardown: same site, two audits

We ran both on a Cairns business site. The blank-prompt audit returned confident boilerplate: “improve page speed, add meta descriptions, build quality backlinks, create fresh content” — advice which is never exactly wrong and never actionable. The data-fed audit returned: 14 pages returning soft-404s from a retired booking system, a redirect chain eating crawl budget on the money pages, mobile LCP of 4.1s caused by one unoptimised hero video, and two service pages cannibalising the same keyword. Four findings, four URLs, four fixes. That's the whole argument in one paragraph.

What feeds a real audit

Five inputs, none optional: a full crawl (every URL, status, tag and internal link); Core Web Vitals field data — what real visitors on real FNQ connections experience, not lab scores; Search Console (what Google has indexed, excluded and why); analytics (where humans actually enter, engage and bail); and competitor deltas (their measured structure and coverage against yours). AI earns its place on top of this stack — triaging hundreds of findings by revenue impact and translating them into plain English — which is precisely where it sits in Content Flow's audit stage.

Why prioritisation is the real product

Any crawler can dump 400 'issues'. The expensive question is which six matter — and that requires joining datasets: a missing meta description on a page with zero impressions is nothing; a redirect chain on the page earning 40% of your leads is this week's job. Data-fed AI does that join in minutes and shows its working. A blank prompt can't, because it has nothing to join.

Run the one-question test

Whoever offers you an AI audit — including us — ask: which specific URLs are affected, and what data says so? Real audits answer instantly with addresses and numbers. Prompt theatre changes the subject. The test is free; it's also the fastest way to sort the data-fed from the guesswork before any money moves. Our version lives inside the Technical SEO service — crawl first, conclusions second, always.