We rated eleven AI app builders. Here is what the panel found

App Builder Index publishes today with eleven reviewed builders, ten scoring axes and a weighting that adds up to one hundred and is printed in full on our methodology page.

What we did

Five reviewers ran six fixed briefs through every tool, three times each. The briefs range from a marketing page to a scheduled importer with a queue and a webhook receiver. We recorded wall-clock time, prompts per finished feature, real spend including failed attempts, and whether the result worked.

Every score in the index traces back to that harness. Where our reviewers disagreed, and they disagreed often, we publish the dissent underneath the verdict rather than averaging it into silence.

The headline results

Lovable takes the highest overall rating at 4.2 stars, on the strength of the best-looking default output in the index and the highest first-attempt completion rate. v0 follows at 4.2, with Totalum and Replit just behind on very different profiles: v0 for code quality, Totalum for administrative depth, Replit for what happens when your application has a back of house.

The most interesting finding was not at the top. Two site builders beat several full-stack platforms on our SEO and GEO axis, because architecture decides that axis and server rendering is not optional. Three builders produced pages whose main content was entirely absent from the initial HTML response.

What we are disclosing

Launch ratings include editorially seeded scores from our own testing panel. Those rows carry an Editor panel badge wherever they appear, community ratings accumulate on top of them, and the methodology page explains exactly how the two interact. We would rather say this loudly at launch than be asked about it later.

What is next

Quarterly re-verification of every builder, with the last verified date and the next scheduled check printed on every review page and driven by the database rather than a hardcoded string. Buyer guides for seven specific use cases. And a machine-readable dataset endpoint under a CC BY licence, because we would rather be cited accurately than paraphrased.

The full ranked table is on our rankings page and the raw dataset is at our public JSON endpoint.