Was this website vibe coded?

A lovable.app address is a strong public association with Lovable. A reference to its gptengineer.js script or an unsigned public trust file is supporting evidence that can be copied. If the markers have been removed or the code is hosted elsewhere, attribution becomes less certain.

Bolt and Replit can expose hosting or platform markers. v0 often produces code that is hosted independently, so a finished site may have no reliable provider fingerprint. The checker reports no attribution rather than inventing a provider.

It can identify explicit public traces and statements about AI use. A generator tag, executable build comment or corroborated public project artefact can support a possible code agent association. A first party statement can support the overall AI build assessment without identifying a specific tool. None of these proves who wrote every line.

It can report selected language, formatting, placeholder and citation patterns in copy present in the public HTML response. Copy patterns can support the overall assessment only as one capped evidence group. They cannot identify a builder or an author on their own. Short or non English copy returns an insufficient evidence result.

No. The overall assessment may still find declared use or several groups of supporting indicators. A result with no material public evidence still cannot prove how the site was built. Limited collection is reported as an incomplete inspection.

Not by itself. The build method does not establish quality or security. The useful follow up is whether the application was reviewed, configured and tested appropriately. Vibe Scan addresses selected public security signals rather than builder attribution.

There is no published accuracy percentage. Provider controlled build platform fingerprints are stronger than generic styling. A pattern result combines capped evidence groups and shows what contributed. Common libraries, gradients, rounded cards, emoji and dash use cannot produce a likely result on their own. The result remains indicative rather than a definitive development history.

It checks public page structure and matching base styles for combinations such as gradient headings, glass panels, decorative glows, varied card grids, repeated icon cards, statistics strips, numbered step rails and highlighted pricing tiers. It also notes scaffolding conventions such as a familiar landing page section order, and reports those separately from appearance. It does not render a screenshot or inspect interactive states. Human designers and templates use the same patterns, so they remain supporting context in one capped interface group.