An empirical study comparing OpenAI, Gemini, and SBERT embeddings for GEO.
ChatGPT, Gemini, and Perplexity share only 21% semantic overlap on brands.
Only 4.57% of sources overlap across ChatGPT, Gemini, and Perplexity.
ChatGPT lists about 8 brands per answer, Gemini and Perplexity about 5.
