How We Test AI Builders
Every score on this site comes from the same brief, the same six axes, and the same verification process. Here is exactly how it works.
The standardized brief
Every builder is asked to ship the same product: a small SaaS feature with email/password authentication, a Postgres table with row-level security, a CRUD dashboard with optimistic UI, and a public marketing page. The brief is delivered as one prompt, with iteration allowed for clarifications only. No architectural hand-holding.
The 6 axes
| Axis | Weight | What it measures |
|---|---|---|
| Speed | 20% | Time from initial prompt to a working, deployable preview that satisfies the brief. |
| Developer experience | 20% | Quality of generated code, ease of editing, debugging tools, and how predictable iteration feels. |
| Output quality | 20% | Production-readiness: accessibility, responsive design, error handling, security defaults. |
| Pricing | 15% | Value relative to comparable stacks, message quotas, and what is bundled vs. add-on. |
| Support | 10% | Documentation depth, community activity, official support responsiveness. |
| Ecosystem | 15% | Integrations, code export, hosting flexibility, and risk of vendor lock-in. |
Verification
Every shipped app is deployed to a real domain and pen-tested by a second reviewer for accessibility (axe-core), CWV (Lighthouse), and security (basic OWASP top-10 checks). Scores get re-verified every quarter or whenever a major builder ships a release.
Conflicts of interest
We disclose all affiliate relationships in the footer of each review. No vendor sees scores before publication. Sponsored placements (if any) are clearly labeled and never alter editorial ranking.
How to flag an inaccuracy
Email corrections@lovableaireview.com with the URL, the claim, and the evidence. We publish corrections within 7 days with a visible "Updated" timestamp.