The five engines disagree far more than they agree

The category talks about AI visibility as though it were a place. It is not a place. It is five retrieval systems, built by five companies, leaning on different corpora, and they return different vendors for the same question often enough that an average across them hides more than it shows.

For each prompt we took the set of vendors named by each engine and compared the sets pairwise. Full agreement would mean every engine named the same vendors. Zero would mean no vendor was named by more than one engine.

The pattern in that table is the practical finding. An engine that leans on documentation rewards a company that writes well about how its product works. An engine that leans on roundups rewards a company that appears in other people's comparisons. Those are different projects, and a team that does one and measures the other will conclude the work failed 1.

Overlap climbs sharply for the single most established vendor in a mature category. When one company has been the default answer for years, every engine finds the same evidence and returns the same name. Agreement is a symptom of an entrenched position rather than a route to one.

Overlap collapses for problem-first prompts that never name a category. Those are the prompts where a buyer has not decided what kind of product they need, and they are the ones where a challenger can actually be introduced 2.

Pick the engine your buyers actually use, prove movement there first, and treat the others as secondary until it holds. For most B2B SaaS panels we have run, that means starting where the buying research happens rather than where the traffic looks largest.