Three competitors, one evidence rule. Every claim says where it came from, or says plainly that nobody has checked it.
Three pages, each comparing Teylo against one named competitor: Replika, Character.AI, and Nomi. Nobody has paid for any of these three products and run the multi-week tests that would answer questions like whether memory actually holds up after months, or whether a companion actually disagrees with you rather than just agreeing. That testing hasn't happened yet, so every page marks those rows as untested instead of guessing.
Teylo wrote these pages. We are not a neutral party in any of the three comparisons, and you shouldn't take our word for them. Every documented row names where it came from and when it was checked, dated, so you can reproduce most of it yourself in an afternoon.
The most established name in the category, and the most credible marketing we found in it.
Millions of user-made characters, against four fixed people with names.
The category's clearest example of romance and roleplay marketed outright.
A documented row names where it came from: usually the competitor's own site, or, where that isn't available, an archived capture of it or a named third-party source, plus the date it was checked. An untested row says so in the row itself, not just in a footnote. Every page also carries a real "Where they're the better choice" section with three honest concessions, not hedged ones, in a panel set apart from the rest of the page.
Same numbers as everywhere else on the site.