AI search for a Sydney marine services company
Its main search moved from position 20.7 to 3.0 on the same page.
The starting point
The business had real, checkable authority in its founder's record, and AI systems couldn't see it. An audit in May scored the site 30/100 for AI-search readiness: no structured data on its main pages, no llms.txt, and thin service pages. A separate check found 44 broken URLs.
The diagnosis
The site was competing with itself. Its main search, "yacht cleaning", was split between two pages, averaging positions 15.1 and 20.7. "Yacht riggers" had the same split, with a second page sitting near 50. Google couldn't tell which page to trust, so neither ranked well.
What I changed
- Gave each page one job: one page for local searches, one for Sydney-wide searches. Rewrote titles and meta descriptions and fixed a misdirected internal link.
- Opened every service page with a direct, self-contained answer, the kind AI systems quote, and added FAQ sections with FAQ markup.
- Added structured data for the business and its founder, and published an llms.txt file.
- Traced the 44 broken URLs and put an 18-redirect map live.
Results
- yacht cleaning: the local page went from 20.7 to 3.0 (changes live 29–30 Sep), and Google now treats it as the page for this term. The other page does its own job: "boat cleaning services sydney" moved from 22.3 to 11.5, on a small sample.
- yacht riggers: 10.9 → 5.6 (changes live 26 Sep), and the competing page dropped out for this term. It is a small search, so read this as "around 5 to 6".
- AI-search readiness: 30/100 in May → 52/100 on 23 Oct 2026, scored against the same five-part rubric.
Positions are Google Search Console page averages: 27 Aug – 24 Sep before, 24 Sep – 21 Oct after. Readiness scores are AI-assisted audit judgements against a fixed rubric, not a tool's measurement; the current rubric no longer gives llms.txt any weight, so 52 is the conservative reading. Positions moved first; clicks had not followed yet in this window.
What I'd do next
A founder bio page tied to the structured data, named and dated authors on every page, and a presence where AI answers look, such as sailing communities and video. Then the next job: turning those positions into clicks.