Answers are grounded in this paper and cite their sources.
Businesses are being given specific advice about how to improve their visibility in AI recommendations, but identifying characteristics of recommended organizations does not establish why one organization was selected over another. Using recommendation outcomes previously collected across ChatGPT, Claude, Gemini, and Perplexity, we compared home-improvement companies within the service and geographic contexts where they could reasonably have been recommended. Public evidence was collected and, in several tests, frozen before recommendation outcomes were revealed. Explanations that emerged from the comparisons were tested against additional organizations where possible. Recommendation visibility was better described at the organization × service × geography level than as a single company-level characteristic, but contextual representation did not reliably explain selection. Companies could appear well represented for a recommendation context and still remain absent. A supplemental analysis of Perplexity's exposed search-result metadata provided evidence beyond the final recommendation. Among 37 instances in which one of six examined companies appeared in the observable retrieval footprint, 13 did not appear in the resulting recommendation. Being visible enough to be found was not sufficient for selection. The findings do not show that the characteristics examined are irrelevant to AI recommendation visibility. They show that observing those characteristics among recommended companies is insufficient evidence that they explain recommendation selection. Claims about what makes an organization more likely to be recommended therefore require evidence capable of accounting for plausible competitors that possess the same characteristics and still do not appear.
No takes yet. Share an insight, caveat, or question.
Christopher Britton (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: