What counts as a citation in Perplexity vs ChatGPT vs AI Overviews?

The definition of "citation" differs by engine, and conflating them produces misleading dashboards. The clean taxonomy across the major engines:

Perplexity: A citation is a numbered link source ([1], [2], ...) that appears beside the synthesized answer, with a clickable card showing title, domain, and snippet. Perplexity also surfaces "Sources" and "Related" sections — those count as citations too. A brand mention inside the answer text without an associated source link is not a Perplexity citation; it's a "mention." Track both separately.

ChatGPT (with Search): A citation is an inline footnote or "Sources" panel link added by the Search tool. ChatGPT base-model responses without Search enabled have no formal citations — brand mentions are pure model-knowledge mentions. The distinction matters: Search citations are recoverable via fresh content; base-model mentions require training-data presence and take 6–18 months to influence.

Google AI Overviews: A citation is a link card displayed beside or below the AI Overview, with site icon, title, and snippet. Google AI Overviews also blends "carousel" sources for shopping and local queries. Position matters — top citation slot drives meaningfully more click-through than slot 3+.

Claude (with web search): A citation is a link card in the response footer with title and domain, appearing when Claude's web tool retrieves the source.

Microsoft Copilot: Citations are inline footnote-style numbered references with a sources panel, drawing from Bing's index.

For consistent cross-engine measurement, score citations on three axes: (1) presence (cited yes/no), (2) position (slot number among cited sources), and (3) sentiment/context (is your brand recommended, neutral, or warned against in the surrounding text). Surfaced normalizes all five engines into a single share-of-voice metric while preserving the per-engine breakdown.