AI Shopping Recommendations Are More Stable Than Dashboards Imply

The research analyzed approximately 3,300 AI shopping responses across 186 combinations of brands, prompts, and models. By querying requests like "best cordless vacuum for pet hair" repeatedly over time, Lantern found that brand presence remained unchanged in 80% of all tracked combinations. Even in unbranded category searches, where competition is fiercest, 71% of brand recommendations stayed consistent throughout the observation period.

Andrew Lissimore, CEO and co-founder of Lantern, noted that the inherent randomness of large language models creates a false sense of volatility. Because these models rarely produce identical text twice, minor stylistic shifts in an answer can be misinterpreted by tracking tools as a change in market position. In reality, the top-ranked brands maintained their lead in two-thirds of all cases. The study concluded that visibility scores oscillate around stable baselines, meaning a brand's performance from a week ago is often just as predictive of its status as yesterday's data. For brands, the takeaway is to ignore the daily dashboard noise and focus on the long-term factors that drive consistent product surfacing.