Updated July 2026
AI Citation Data & Benchmarks
12,430 prompts. Four AI platforms. One citation map.
This AI citation data hub tracks which sources AI search platforms actually cite — every benchmark below comes from the CiteMP source analysis, July 2026, the same dataset that powers our AI citation marketplace.
- 12,430prompts analysed
- 4AI platforms
- 132sources per prompt avg
- Monthlyrefresh cadence
Source types AI cites most
CiteMP source analysis, July 2026 · n = 12,430 prompts
Top source type
Comparison articles — 31% of citations
Share of all cited URLs by format, averaged across ChatGPT, Google AI Overviews, Perplexity and Claude.
Citation mix by platform
Every platform weights formats differently — which is why one generic backlink plan never fits. Browse format-matched inventory in the AI link building marketplace.
ChatGPT
Leans listiclesShare of platform citations · July 2026
Google AI Overviews
Leans comparisonsShare of platform citations · July 2026
Perplexity
Leans listiclesShare of platform citations · July 2026
Claude
Leans listiclesShare of platform citations · July 2026
Source format benchmarks
Demo benchmark values · CiteMP source analysis, July 2026
| Source format | Avg citation share | Median DR of cited pages | Avg content age | Refresh frequency |
|---|---|---|---|---|
| Comparison articles | 31% | 72 | 4.2 mo | Monthly |
| Industry publishers | 24% | 68 | 7.5 mo | Quarterly |
| Listicles | 18% | 64 | 3.1 mo | Monthly |
| Review platforms | 13% | 81 | 5.8 mo | Weekly |
| Original research | 9% | 76 | 11.2 mo | Annually |
| Other | 5% | 52 | 14+ mo | Rarely |
Formats map one-to-one to the placement types quoted in campaign pricing — the mix a package buys mirrors the mix the data shows.
What the data means for your campaign
Comparison + listicle share
49%
Two formats carry almost half of all AI search citations — the first placements a GEO link building campaign should buy.
Median authority of cited pages
DR 64+
Cited pages clear DR 64 in every format, so filter publisher placements by authority before price.
Avg age of cited comparisons
4.2 mo
AI platforms favour fresh pages, which is why AI citation building runs as a refresh cycle, not a one-off.
Methodology
- Sample. 12,430 commercial-intent prompts across 41 industries, collected January–July 2026.
- Collection. Each prompt run on ChatGPT, Google AI Overviews, Perplexity and Claude; an average of 132 cited sources per prompt classified by format.
- Cadence. The dataset refreshes monthly; figures on this page are demo benchmarks from the July 2026 run.
Get the citation map for your prompt.
We run your prompt through the same pipeline behind these benchmarks and return the exact sources — and placements — to target.