A teardown of why ChatGPT names some brands and ignores others, and what that means for content strategy in 2026.
The mechanic, in three sentences
An AI assistant retrieves passages from its training data plus a live search layer, ranks them by relevance and authority, then writes a synthesised answer. The "citation" you see at the end is the model deciding which passages most influenced its output. That is not the same thing as the highest-ranking pages on Google.
That's why a small, focused, well-cited industry blog can outrank a large publisher in AI answers. Authority signals are different here.
What the data shows
Across 200+ projects we've tracked, three patterns repeat:
- Specificity wins. A page titled "Best CRM for 50-person agency" gets cited more than "Top 10 CRM tools 2026", even if the latter ranks #1 on Google.
- Recency matters. Pages updated in the last 90 days are cited disproportionately. Old evergreen content fades faster than in classic SEO.
- Original numbers win. If your page contains a primary statistic ("we surveyed 1,247 buyers"), AI will quote it, often with attribution. Curated lists of others' stats: rarely cited.
What to do this week
Update your three highest-traffic pages with: a refreshed publish date, an original data point, and a tightened headline that matches the question your buyer actually asks AI. Then track for two weeks. Most clients see citation rate move within 14 days.
