The brief
Northwind had strong traditional SEO but was losing ground in AI answer surfaces. In a regulated category, being cited — or not cited — was the difference between inbound and silence. They wanted to own the primary-source definitions for three products before the next model generation locked them in.
What we did
- 01Mapped category queries across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews. Recorded citation sources and synthesized claims, query by query.
- 02Audited the existing entity graph and JSON-LD. Found three silent canonicalization bugs, a schema type mismatch, and a stale organization node. Shipped the fixes in week one.
- 03Wrote nine comparison and definition pages with Northwind’s domain experts. Each page was designed to be quoted — short declarative openers, named entities, numeric claims with citations.
- 04Rebuilt the internal linking graph to route authority toward the new answer pages. Added llms.txt and tuned crawler access for the five leading model operators.
What changed
By week four, Northwind was appearing in the top three sources cited for 61% of tracked queries. By week twelve, that figure hit 74%. Assisted conversions on organic channels rose 31%, driven entirely by branded-query lift from AI surfaces. Three competitors re-brief their vendors within the quarter.
What we would not repeat
We over-indexed on comparison pages in week two. A single explainer covering the category basics would have compounded faster — comparisons need a definition to link back to. We ship the explainer first now.