Patterns in how four engines cite media, and the pitching decisions we changed because of them.
The exercise
We ran a large batch of buyer-relevant prompts across ChatGPT, Gemini, Perplexity, and Claude and read the citations behind every answer, rather than just the answer text itself — which outlets got pulled in, how often, and whether the same sources kept reappearing across different phrasings of a similar question. The point wasn't to grade any one engine; it was to see whether the sources an engine trusts are stable enough to pitch toward on purpose.
The engines don't agree with each other as much as you'd expect
The most consistent pattern was inconsistency between engines. An outlet cited constantly by one engine for a given topic was sometimes barely retrieved by another, even when both were answering close variants of the same question. That single finding is why tracking "AI visibility" as one undifferentiated number is less useful than tracking it per engine — a brand can be well cited on Perplexity and nearly invisible on Gemini for the exact same category, and a single blended score would hide that entirely.
Recency beat authority more often than expected
A frequently-updated, well-structured page from a mid-tier outlet regularly outranked a stale but prestigious source on the same topic. This matched what we already believed about retrieval favoring freshness and structure, but seeing how often it actually decided the citation — not just contributed to it — changed how we weight "a placement from three years ago" in a brand's existing AI Citation Score.
What changed in the product
Two things, directly from this read-through. First, citation tracking now breaks out by engine by default instead of defaulting to a blended view, because the blended number was hiding the exact gaps that matter most for deciding where to pitch next. Second, the AI Citation Dashboard now surfaces recency alongside citation rate, since a high score built on old placements behaves differently going forward than one built on a source that's still being actively cited.
hello@newpr.ai — questions, tips, or a story we should be covering.