Evidence follows the claim
Official documentation establishes what a provider advertises. A dated interface check establishes what was visible in that check. A creator demonstration shows that creator’s selected example. These forms of evidence are useful for different questions and are not interchangeable.
We preserve source URLs, dates, authors where known, relevant limitations and the pages using each source in a central evidence register. An unknown field remains unknown.
Experience and scoring
We distinguish hands-on observations, provider specifications and creator demonstrations. A measured performance claim requires the original brief, settings, outputs and usage record. Example prompts are starting briefs; showcase images retain their source context.
Published ratings identify the source and the product being rated. We display scored platforms with customer reviews, and keep our editorial recommendations separate from those external ratings. Attributed review summaries retain relevant limitations as well as positive feedback.
Prices and updates
Prices record the source, checking date and billing period. Annual equivalents are separated from monthly payments. Different credit systems are not assumed to buy the same capacity.
We revisit mutable information before a material content update. Contradictory sources are resolved using the most direct relevant record, with unresolved points kept out of factual sales claims.
Editorial independence and corrections
Our recommendations follow project fit, creative controls and the requirements of a finished deliverable. Positive feedback and practical limitations both belong in a useful review.
A factual correction should identify the page, statement and a direct supporting source. Use the contact information published on the contact page when an editorial channel is available. Product-account requests belong with the provider.
Match the evidence to the statement
Our editorial policy starts with the kind of claim being made. A provider page can establish an advertised allowance or a documented control. A recorded operation can show how a workflow is arranged. A creator opinion can explain that person’s experience. None of those, by itself, establishes a repeatable performance result for another user’s account or project.
We use stronger language only when the underlying record supports it. A statement about measured output quality needs the original brief, relevant inputs, settings, attempts and accepted result. A statement about paid usage needs the actual purchase or operation record. An attractive demonstration is not a substitute for that record.
Keep examples and comparisons useful
A tutorial example should give the reader a concrete task, not an invented success story. We label prompts as examples and explain what to inspect. If a bottle, character or short ad is an editorial brief, the instructions describe the intended process rather than claiming that a particular paid run has already produced it.
Comparisons begin with requirements. We distinguish a native model control from a broader workspace, and we keep different usage units separate. A lower public customer-service rating is not a measured image-quality score. A selected shortlist does not become a comprehensive market ranking merely because it is displayed in a table.
Maintain prices, ratings and recurring facts consistently
Plan prices, allowances and repeated feature lists belong in shared data so a summary and a detailed buying page do not quietly disagree. Monthly payments and annual equivalents are different commitments. Advertised list-price discounts and savings against twelve monthly payments can also use different baselines; the wording should identify the relevant comparison.
External ratings keep the platform and product scope attached to them. A rating for a specific feature listing should not be presented as a universal score for every part of a service. Selected review summaries are kept distinct from the full rating and do not create a new aggregate of their own.
Represent people, recordings and commercial relationships accurately
A supplied author portrait and a stated work background identify the author; they do not establish a particular test duration, software bill or cancelled subscription. Third-party video remains third-party video. We do not replace its creator identity, imply that our team filmed it or fabricate a matching reviewer to make separate sources look like one person.
Eligible affiliate purchases may support our work. Product links and commercial disclosures should describe the actual relationship. A referral URL is not proof of a discount, unrestricted trial or refund guarantee. Ordinary source links are not called affiliate links. Our recommendations should remain understandable from the project fit and evidence, without depending on a commission.
Make corrections meaningful
When a material fact changes, revise the affected copy and the shared information it relies on. Check the pages that reuse that information, including comparison text, FAQs, calculators and structured data. A corrected headline is not enough if an old allowance remains in another visible module.
We prefer a specific correction over a vague assurance that content is current. Identify what changed and retain an internal source record for the revision. This editorial policy also applies to limitations: if an offer, test result or capability cannot be substantiated, preserve that uncertainty instead of filling the gap with a plausible number.