01

Reuse one fictional scene

Use the archive postcard prompt from the roleplay guide. Keep the same instruction about leaving your actions to you. Record the first reply, one correction, and the reply after correction. This produces inspectable observations rather than a vague “feels better” rating.

02

Fill an evidence matrix

For each service, record: turn-taking result; correction result; same-conversation recall; new-conversation recall if available; cost for your chosen usage; and documented removal and cancellation controls. Each cell should contain an observation, a policy reference, or “unknown.” Do not convert unknown to a passing score.

03

Make the trade-off explicit

For short writing exercises, turn-taking may matter more to you than voice. For recurring stories, continuity may be your first criterion. State your chosen priority before reading the results, then keep the conclusion tied to that use case.

04

Bound the conclusion

Record the date, plan, settings, and device for each side. Different plans or test conditions limit direct comparison. This page provides a protocol, not tested rankings, endorsements, or evidence that WantYou outperforms another product.