THE PUBLICATION / OUR STANDARDS
Evidence before a verdict.
Every article identifies what its conclusions are based on. Reading a product page does not prove how well the product performs.
Source-based research
Research articles compare current official documentation, plan descriptions, help pages and terms. We link the sources and show when they were checked. We distinguish vendor statements from our interpretation and call out conflicting or inaccessible information.
Research can explain a feature, a documented limit or a buying tradeoff. It cannot establish that one tool produces better results, runs faster or fails less often. Our current launch articles are in this category.
Documented testing
A testing claim needs a retained record. For each test, record the product, plan, model, date, settings, input rights, prompt, output duration and resolution, attempts, failures, elapsed time, credits and spend. Preserve the output and the reason it was accepted or rejected.
Use the same brief and acceptance criteria across comparable tools. Report the number of attempts and all material failures. Keep a vendor demonstration separate from our own result. A small test is an observation under stated conditions, not proof of average performance.
The first three briefs
- Product advertisement: evaluate a short clip against a fixed brief, including product consistency, motion and editing needs.
- Narrated tutorial: evaluate script accuracy, intelligibility, timing and pronunciation without cloning anyone's voice.
- Short-form repurposing: evaluate whether an edit preserves meaning, captions and useful framing from an owned source clip.
These are planned procedures. No completed results are implied by this list.
A usable result is the unit that matters
Record what “usable” means before the run. Include rejected attempts when calculating cost. The cost calculator divides the entered spend by accepted outputs; it does not estimate sales or predict earnings.
Recommendations and corrections
Explain who a tool fits, who should consider an alternative, and which uncertainties remain. Do not let a commission rate alter those conclusions. Refresh a checked date only after a substantive source review. Material corrections belong in the affected article.