Before → after
| Metric | Before | After |
|---|---|---|
| AI citation share (40 target queries) | 1.0× | 3.2× |
| AI Overviews presence | 4 of 40 | 18 of 40 |
| Referral traffic from AI assistants | baseline | 4× |
A B2B analytics company had strong organic rankings but was nearly invisible in AI answers. When prospects asked ChatGPT “best analytics tools for startups,” the model cited three competitors and never them.
The challenge
Their content ranked in classic search but wasn’t built for citation:
- No
llms.txt, minimal structured data. - Core claims and statistics buried in long paragraphs.
- Thin presence on the third-party hubs AI models actually read.
- No way to measure whether AI answers were improving.
What we built
We ran a GEO program focused on citation-worthiness:
- Technical foundation added
llms.txt, full JSON-LD entity markup, and made every key page fetchable by AI crawlers. - Citation engineering restructured core pages around direct, quotable answers: a clear definition, a named statistic, and a “why” within the first 100 words.
- Data-rich programmatic pages built comparison and “alternatives” pages that gave models specific, citable facts.
- Distribution strengthened presence on review sites and comparison hubs the models read.
- Monitoring a scheduled agent queried ChatGPT, Perplexity, and Gemini weekly to track citation share and alert on changes.
The results
Within 90 days, the brand’s AI citation share more than tripled, and the domain went from appearing in 4 to 18 of 40 AI Overviews. Referral traffic from AI assistants grew 4×. The monitoring agent now runs continuously, so the client gets a Slack alert the moment a competitor displaces them in an answer and we can respond in days, not quarters.
The takeaway
Ranking is no longer just about Google, it’s about being the answer an AI cites. Structure for citation and measure it like any other channel.
We went from never-mentioned to a default citation in our category in a single quarter.