Aggregated, anonymised audit data from active engagements. No claims, no derivations from study summaries — these are the numbers I see in mandates myself.
The figures come from a rolling cohort of 500+ enterprise domains in DACH, the EU and the US. All data is anonymised at industry level; individual domains cannot be identified through the aggregation. The purpose: market-wide orientation for operators — not marketing.
Data as of April 15, 2026. The next update is in preparation; until then the values stay as they are and are not extrapolated. Text last reviewed: September 23, 2026.
The benchmarks show that AI answers have become a structural part of search. In our cohort of 500+ enterprise domains, the median AI Overview Absorption Rate (AAR) on informational queries rose from 0.05 in October 2024 to 0.41 in April 2026, and the curve has been flattening since Q4 2025. In the EU market, our weighting model assigns roughly 52 percent of LLM usage to ChatGPT, 22 percent to Gemini, 11 percent to Claude and 10 percent to Perplexity.
Citation rate in Perplexity climbs with Domain Authority, from below 1 percent (DA 0–20) to above 7 percent (DA 81+), with a disproportionate rise from DA 60. Among AI crawlers, GPTBot leads with a median of 1,240 requests per day on domains with more than 10,000 URLs; ClaudeBot reaches 610. On the Reputation Drift Index, e-commerce is the only industry in negative territory (−8, Q1 2026). All values reflect data as of April 15, 2026.
Median AAR across 500+ domains, informational queries. October 2024: 0.05 (shortly after EU rollout). April 2026: 0.41. The curve has been flattening since Q4 2025 — the structural shift is nearly complete.
Weights for EU markets. In the US: Perplexity higher (~18%), ChatGPT lower (~44%). In Turkey, ChatGPT dominates with >65%.
Citation rate in Perplexity rises disproportionately from DA 60 upwards. Threshold for relevant citation presence: ≥ DA 40. Our working hypothesis: authority signals carry heavy weight in retrieval. The correlation in our cohort supports this, but it is not proof of causation.
Median requests on enterprise domains (> 10k URLs), measured by user agent in server logs. Google-Extended and Applebot-Extended are robots.txt control tokens without a user agent of their own (Google, Apple) and cannot be counted in logs. The values previously shown for both tokens were removed on September 23, 2026.
Reputation Drift Index (median) by industry, Q1 2026. E-commerce shows a negative trend due to increased price- and service-related negative mentions.
Methodology repo in preparation · data aggregated in BigQuery · SVGs delivered statically · data as of April 15, 2026