3cubed.ai Research
Research on how AI engines recommend brands
Studies, benchmarks, and methodology notes for teams measuring AI citation coverage across ChatGPT, Gemini, Claude, Perplexity, and the source domains that shape their answers.
Research ReportJune 2026
AI Recommendation Stability
A 3cubed.ai Research study of 2,580 AI runs across 26 unaided buyer questions, four AI engines, and three controlled conditions.
2,580
AI runs
48-68%
real-world consistency
27%
four-engine agreement
Put the research to work
Findabl turns these findings into weekly citation tracking, source-domain mapping, and prompt-level action plans.
Get Your Free GEO Score Download the white paper