Definition
Citations Per Response is the average number of source citations an AI search engine includes in a single answer. It varies by platform and model: Perplexity tends to cite more sources per answer than ChatGPT, and the count shifts with model rollouts. It is a key context metric for interpreting citation share and citation diversity.
Citations per response matters because it sets the denominator for citation share. A platform that cites five sources per answer offers more citation slots than one that cites two, so a brand's citation share must be read against the platform's baseline. Promptwatch's average sources per response report tracks this across ChatGPT, Claude, Perplexity, and Gemini, and the ChatGPT citation drop after GPT-5.3 report shows how a model rollout can change the count overnight.
For GEO teams, track citations per response per platform and watch for sudden changes that signal model or retrieval shifts. A drop in citations per response compresses every brand's citation share, so distinguish a platform-wide change from a brand-specific decline before reacting.
Pair citations per response with multi-source synthesis and source aggregation to understand how platforms combine sources, not just how many they cite.
Examples of Citations Per Response
- A brand compares [average sources per response](/data/average-sources-per-response) and adjusts its citation-share targets per platform accordingly.
- A GEO team uses the [ChatGPT citation drop](/data/chatgpt-citation-drop) report to confirm a platform-wide decline, not a brand-specific loss.
- A team finds Perplexity cites more sources per answer and prioritizes it for citation-share growth given the larger denominator.
- A marketer tracks citations per response alongside [citation diversity](/glossary/citation-diversity) to see whether platforms broaden or narrow their sources.
