The State of AI Search 2025: How 10,000 Queries Were Answered by AI
Our analysis of 10,000 AI search queries across ChatGPT, Perplexity, Gemini, and Claude reveals citation patterns, industry breakdowns, and what the data means for your content strategy.
We analyzed 10,000 AI search queries across ChatGPT, Perplexity, Gemini, and Claude to understand which content characteristics predict citation. This is what we found.
Study methodology
We collected 10,000 queries across four AI platforms (ChatGPT with browsing, Perplexity, Google Gemini, and Claude) between January and June 2025. Queries were stratified across 12 industries and split between informational, commercial, and navigational intent. For each cited source, we analyzed 28 AEO signals to identify which signals appeared most frequently in cited vs uncited pages.
Citation rates by platform
Citation behavior varied significantly across platforms. Perplexity cited the most sources per query; Claude cited the fewest but showed the strongest preference for high-authority sources.
Which signals most strongly predict citation
Across all platforms, five signals showed the strongest statistical correlation with citation (p < 0.01):
FAQPage Schema
+41% citation rate vs pages without
Content freshness (dateModified < 90 days)
+38% citation rate
Question-format H2 headings
+34% citation rate
Author Person Schema with credentials
+29% citation rate
Answer-first paragraph structure
+27% citation rate
Citation rates by industry
Citation rates varied significantly by industry, largely tracking with the average AEO score of published content in that industry. Technology and finance had the highest citation rates; retail and hospitality had the lowest, reflecting lower average Schema adoption.
The measurable impact of Schema markup
Schema is the single highest-ROI change
Schema adoption among cited pages was 69% on average — versus only 31% for the general web. This gap represents significant opportunity for publishers who haven't yet invested in Schema implementation.