Methodology
How we research and rank alternatives
Our process combines public primary sources, structured eligibility rules, independent feature-alignment checks, and transparent ranking signals.
1. Eligibility
A project must have verifiable open-source code, document a usable graphical application or platform, support self-hosting, and solve the target product’s defining primary job. Libraries, plugins, wrappers, directories, abandoned variants, and projects supported only by superficial feature overlap are excluded.
2. Sources and verification
Research starts with the project’s repository, README, repository description and topics, official website, license metadata, releases, commit activity, and other linked official documentation. Each published comparison exposes its available evidence links so readers can check the basis for the classification.
3. Feature comparison
The candidate’s primary workflow and output are mapped against the original product’s primary features. A separate alignment pass must agree with that mapping and provide evidence. Shared features and missing capabilities are recorded separately so partial alternatives are not presented as complete replacements.
How rankings are calculated
- Alternatives for a product are ordered by reviewed match confidence, then GitHub stars as a tie-breaker.
- Tool and category directories default to GitHub stars, then alphabetical order; directory controls offer additional sorting.
- Similar tools must share a category and are then ordered by GitHub stars.
- Product directory ordering is based on the number of approved alternatives.
- Paid placement does not influence the current rankings.
Stars measure repository popularity, not product quality. Confidence measures documented workflow alignment, not a universal score for security, usability, support, or suitability.
How automation and AI assist research
Automation discovers candidates, collects repository and project metadata, refreshes activity metrics, and detects duplicates. Structured AI analysis classifies candidates and extracts feature evidence only from supplied public material. A second AI-assisted check independently tests the primary-feature match, while deterministic gates reject entries that fail evidence, confidence, self-hosting, product-type, or workflow requirements.
AI can misunderstand documentation or miss recent changes. Kris owns the editorial rules and correction process, and automation is disclosed on comparison pages rather than presented as first-hand use. See the editorial policy for limitations and corrections.
Review and update dates
“Last reviewed” records the latest classification review. “Materially updated” records the latest change to the comparison record, such as its summary, feature mapping, evidence, status, or review data. Repository activity may refresh independently from editorial text.