These guides turn vague feature claims into questions you can actually test or calculate. They are written for readers who would rather investigate an AI companion than rely on a sponsored “top ten”.

Use the guides in this order

  1. How to choose — define one goal, one non-negotiable limitation and a realistic monthly ceiling before looking at a character gallery.
  2. Memory test — use invented names and events across seven days, score recall and distinguish persistent notes from ordinary chat context.
  3. Voice guide — verify the actual supported spoken language, permissions, latency and metered calls on your intended device.
  4. Pricing explained — calculate subscription, tokens, media packs and renewal rather than copying promotional prices.
  5. Privacy checklist — check retention, memory deletion, export and the provider’s published policy before sharing identifying details.
  6. Character consistency — set up a coherent fictional brief and test whether traits remain stable over multiple chats.

Two different kinds of evidence

Provider documentation can establish whether an option exists or how billing is described. Only genuine repeatable testing can establish whether a companion consistently remembers, responds naturally or follows a character brief. We provide procedures for readers to conduct those tests; we do not assert results we have not measured.

Before purchasing from the UK

Where a price is quoted in USD, calculate a personal GBP budget using your payment provider’s actual exchange rate and fee. Check the renewal date, whether any trial turns paid automatically, and how a user cancels. If extras are purchased in credits, the credit pack size affects the cash you must spend—not only the nominal cost of one image.

After reading the guides, visit our individual dossiers, compare two candidates in head-to-head analysis, and model spend in the calculator.