# Methodology and evidence standards

> How Crowdangels distinguishes editorial criteria, provider claims and documented tests when covering AI image and video tools.

Źródło: https://www.crowdangels.pl/methodology/

---

# Methodology and evidence standards

English policy edition. The main guide is available in Polish.

This page applies to the editorial publication Crowdangels. It concerns reading this website and following its links. It does not replace the policies of a generator, payment provider or review publication that you visit separately.

## What a guide establishes

Our main guide explains how to evaluate a type of tool. Its questions are a decision framework, not a report that every named provider was tested. A recommendation to check retention does not mean that we verified a deletion mechanism. A suggestion to inspect video continuity does not imply a measured error rate. We keep those distinctions visible so that a reader can understand the practical limits of the information.

## Separate kinds of evidence

A provider’s documentation can establish what it publicly states about its product. A dated checkout page can establish the displayed price and conditions at that moment. A documented hands-on evaluation can support observations about a particular plan, model version and test material. These sources answer different questions. We do not turn an advertised feature into a verified performance claim merely because the description is precise or widely repeated.

## Requirements for an actual comparison

An evaluation should record the selected plan, operation, model information when available and the material used with appropriate permission. Compare the exported file rather than relying only on a preview. Keep rejected attempts when assessing repeatability. For video, review the complete sequence and record the duration used. For costs, identify whether credits were consumed by failures or additional exports. Without that record, a numerical ranking can imply more certainty than the evidence supports.

## Scope of current coverage

These rebuilt guides do not present a proprietary benchmark, claim hands-on testing or publish a table of invented scores. They provide topic-specific advice and link to the [Clothoff review publication](https://clothoff.pl/) as a related reading destination. That destination is operated by the same team and has its own methodology. Its statements should be assessed against its evidence rather than treated as independently verified merely because we link to them.

## Updates and corrections

We change an article date when the content changes meaningfully. We do not refresh all dates simply because a build runs. A useful correction cites the page and the exact statement, explains the issue and provides a public source. Private or intimate files should not be submitted to establish a correction. The [contact page](/contact/) explains the currently available channels. Broader privacy questions can be directed using the official [EDPB authority directory](https://www.edpb.europa.eu/about-edpb/our-members_en). This process describes our editorial standard; it does not claim certification or external accreditation.