A study by Leapd based on the analysis of 680 million citations shows that three leading AI search systems — ChatGPT, Google AI Overviews, and Perplexity — use fundamentally different source selection strategies. ChatGPT relies on encyclopedic resources, Perplexity focuses on content freshness, and Google AI Overviews combines social content and structured data. A universal optimization strategy for all systems simultaneously does not work.

image
image

What happened

Leapd published an empirical study of citation patterns across three AI search systems on a sample of 680 million recorded citations. ChatGPT highlights Wikipedia with a share of 7.8 percent of all citations and relies on training data. Google AI Overviews prefers Reddit with a share of 2.2 percent and LinkedIn with 1.3 percent. Perplexity focuses primarily on Reddit — 6.6 percent of citations, while 82 percent of all cited materials in Perplexity are less than 30 days old. Domain overlap between ChatGPT and Perplexity is only 11 percent. Pages that appear in AI Overviews receive 35 percent more clicks, and traffic from Perplexity converts approximately 11 times higher than regular organic search.

Context

Before the publication of this study, the industry often considered "AI search" as a single new distribution channel, similar to traditional search. Leapd's data demonstrates that each platform uses a separate retrieval pipeline with non-matching ranking signals: ChatGPT is tied to static encyclopedic sources, Perplexity maximizes freshness, and Google AI Overviews applies a hybrid approach. This means the fragmentation of AI search into three independent ecosystems — and the emergence of a niche for AI citation monitoring products.

Why this matters for the industry

The study confirms that a universal SEO strategy no longer covers AI search. For ChatGPT, encyclopedic authority and Bing indexing are critical — the FAQ format increases citability in ChatGPT by approximately 40 percent. For Google AI Overviews, structured data and multimodal content are necessary. For Perplexity, material freshness and presence in Reddit communities are important. Developers can start building specialized tools: API wrappers for tracking citations, FAQ generators for ChatGPT, fresh content engines for Perplexity. For data researchers — a baseline for citation-diversity benchmarks and designing retrieval systems for specific use cases.

Why this matters for users

If your content does not appear in AI search answers, the reason may not be quality, but a format mismatch with the specifics of a particular platform. To get into ChatGPT, use FAQ schemas and ensure presence in Bing indexing. For Google AI Overviews, add structured data and multimodal elements. For Perplexity, publish fresh content and participate in Reddit communities. Each platform requires a separate approach to formatting and updating materials.

What is still unknown / limitations

Leapd's study methodology is proprietary and not publicly described: sample selection criteria, the definition of citation, and the methodology for measuring traffic conversion from Perplexity (11-fold excess) are not disclosed. Replicability of results is not guaranteed. Conversion data may depend on the control group and metrics that Leapd did not publish.

Sources

Author

Look at AI, editorial team