Include bullet points
86.8% ↓
Benchmark data from rocketblue's monitoring of AI-generated responses, brand mentions, citations, and source behavior across major AI search platforms.
Share of cited links from major global platforms over the last four months. Pick an AI model to see the weekly trend. ChatGPT is shown first.
Share of cited links, weekly
Current benchmark profile for pages cited by tracked AI platforms.
Include bullet points
86.8% ↓
Include visual data
42.9% ↓
Include author credentials
56.0% ↓
Include empirical evidence
23.1% ↓
Include FAQs
27.5% ↓
Include original research
1.1% ↓
Current benchmark values from the newest available reporting period, with change vs the immediately previous reporting period when that comparison exists.
Average response citation rate (across tracked platforms)
78.6% ↓
Average citations per cited response (across tracked platforms)
6.70 ↓
Average brands mentioned per response
5.13 ↓
ChatGPT responses with citations
98.5% ↑
| Platform | % responses with citations | Avg citations per cited response | Avg brands per response |
|---|---|---|---|
| ChatGPT | 98.5% | 6.38 | 5.97 |
| Gemini | 80.5% | 3.67 | 5.03 |
| Claude | 28.7% | 7.84 | 5.95 |
| Perplexity | 91.8% | 9.45 | 4.08 |
| Google AI Overview | 87.3% | 7.30 | 5.09 |
| AI Mode | 85.0% | 5.53 | 4.65 |
These findings show change based on the most recent monthly report (Aug 31, 2026 – Sep 6, 2026) .
ChatGPT citation rate: 98.5% (previous: 96.2%, change: +2.3 pp).
ChatGPT cited sources more often.
corporate website share among cited websites: 47.3% (previous: 50.7%, change: -3.4 pp).
corporate website accounted for a smaller share of cited websites.
Average brands mentioned per response: 5.13 (previous: 5.35, change: -4.2%).
Responses mentioned fewer brands on average.
Average response citation rate: 78.6% (previous: 79.4%, change: -0.8 pp).
Tracked platforms cited sources in a smaller share of responses.
rocketblue monitors prompts that map to brand discovery and purchase-intent questions, then collects AI-generated responses, citations, and brand mentions on a recurring reporting schedule.
The current benchmark includes tracked responses from ChatGPT, Gemini, Claude, Perplexity, Google AI Overview, and Google AI Mode, based on the prompts and sampling windows in each reporting period.
A citation is counted when the model response includes a referenced source URL or an inline source reference that maps to a source document captured in the dataset.
Reporting periods are defined by each monthly snapshot's period start and end dates. New monthly reports are added as new snapshots become available.
This dataset reflects rocketblue's monitoring methodology, reporting schedule, and tracked prompt mix. It should not be interpreted as AI platform market share, real-world usage share, or representative of all AI conversations on the internet.
Based on about 2.3 million response links from United States, United Kingdom, Israel, Germany, Australia, India, France, Italy, Switzerland, Spain, and 50 more.