Two structured GEO experiments reveal flaws in conventional AI visibility advice

Search Engine Land reports on two rigorous generative engine optimization experiments designed to test whether current GEO theory actually works in practice. The first, a months-long test with an existing brand across ChatGPT, Claude, Gemini, and Perplexity, achieved roughly 10–12 keyword appearances out of 15 tracked, with listicles and PR working together to drive 72.4% of citations—a combined effect stronger than either channel alone. The second, a 30-day cold-start test with a SaaS link building agency across six AI platforms, generated 298 appearances from zero baseline, with Gemini and Google AI Mode leading. Key findings include that source authority matters more than content quality alone, rapid source decay means one-time placement doesn't sustain visibility, the peer set surrounding a mention influences citation likelihood, exact-match keyword targeting still matters, and placement within the first 100 words of an answer significantly improves citation likelihood—though one original conclusion from experiment one did not replicate in the second test.