Research-based is a specific status
A research-based article summarizes identifiable public sources and adds editorial analysis. It does not imply that we purchased a plan, chatted with a character, generated media or tested support. All launch tool reviews have this status.
What qualifies as a hands-on review
A personally tested review must identify the test date, plan and model where available, the tasks performed, the limits of the assessment and actual observations. Screenshots or examples must come from that test and avoid exposing private information. We do not manufacture time spent, subscription receipts or product experiences.
How we handle features and prices
Feature records distinguish documented availability from unverified details. “Not verified” is not the same as “No”. Prices state the billing period; an annual equivalent must also make the upfront payment clear. Promotions, separate credits and add-ons require their own explanation.
Why there are no launch scores
Numerical ratings suggest comparable measurements. We do not have repeatable hands-on records for the launch products, so we do not publish star ratings or call a platform the proven winner. Buying guides offer a use-case shortlist and describe the evidence needed for a stronger conclusion.
Comparisons and updates
We compare similar plans and tasks when evidence permits. An article’s updated date changes when its substance changes. The source-check date identifies when a provider page was inspected; it is not a testing date. Unavailable or contradictory evidence is described rather than filled with a guess.
Commercial relationships
Potential affiliate relationships do not justify inventing results or hiding limitations. Commission-bearing links must be disclosed and technically marked as sponsored. Read the affiliate disclosure for the current launch state.
The evidence record behind a product claim
For a documented claim, record the provider, source URL, inspection date and the plan or feature it concerns. Preserve the distinction between a help article, a promotional page and an account-level checkout. When those sources conflict, identify the conflict and avoid silently choosing the most attractive claim.
For a calculation, show the inputs and assumptions. Multiplying a published monthly price by twelve is an editorial calculation. It is not a provider quotation or a prediction of the user’s actual spend. Optional purchases, tax and promotions may require separate treatment.
A repeatable testing record
A future hands-on assessment should preserve the original setup, first responses and relevant settings. It should count unsuccessful attempts and record the effect of corrections. A selection of successful screenshots without the underlying attempts cannot support a general reliability claim.
Tests should use original fictional adult characters and non-sensitive, non-explicit material. Memory, style, media quality, cost and interface usability should be reported separately. A good result in one category should not be used to fill an untested category.
Update and correction rules
Substantive changes include a revised price, an announced free-tier transition, a corrected feature claim or new evidence that changes a recommendation. A source can be rechecked without pretending that the product was tested on that date. A future change should remain clearly described as announced until it takes effect and is verified.
Translations require their own review when the source changes. A translated title and interface are not evidence that a full localized article has been completed. Published language versions should link to the corresponding article when that version exists, while unfinished material remains outside the public library.