AI App Builder Lab Results: Timed Builds We Actually Ran

By the PrototypeVerdict Test Team

← All free AI app builder tools

Every row is a build we actually timed. Times, fix counts and storage notes are copied from the lab logs — we do not invent or average a missing figure. If the log did not record something, the cell says “not measured” or “not reached.” Sort any column and filter by builder, test app or whether the app reached a working state. Download the visible rows as CSV.

Lab results explorer

Loading the table… If nothing appears, JavaScript is off; a static copy of every row is in the provenance notes we keep with this page, and the related reviews still have the write-ups.

How to read the table

  • Time to first preview is when we first saw a generated UI, as written in the log (including “about” and ranges).
  • Time to working is when the prompt’s required features worked, or “not reached” if they did not.
  • Sorting uses the first number written in the log (so “~2-3 min” sorts as 2 minutes, and “about 11-12 min” sorts as 11). Blank “not measured” cells sink to the bottom. That is a sort key, not a new measurement.
  • Softr’s public review is labeled as needing Pro for a score. This table reports what happened on Softr Basic; it does not publish an overall Softr score.
  • Google AI Studio rows come from the Nov 9 free-tier session. Private app URLs are kept out of this page.

The change-repair / security probes from Nov 8–9 are not listed here. Those sessions were access and repair checks, not a fresh timed run of the three standard apps, so there is nothing honest to put in a “time to first preview” cell.

Related

FAQ

Where do these numbers come from?

From our lab logs for Lovable, Base44, Bolt, Softr, Replit and Google AI Studio (the Oct 29–30 timed builds and the Nov 9 Google AI Studio free-tier session), with one Base44 storage note taken from our scoring notes. Each row’s source note is included in the CSV.

Why do some cells say “not measured”?

Because the log never recorded that figure. We would rather show a gap than fill it with an average or a guess.

Can I reuse the table?

Yes. Download the CSV. If you publish the numbers, please cite PrototypeVerdict and link to this page or the Usage Index. The underlying usage dataset on the Usage Index is CC BY 4.0; these timing notes are our lab records.

Did every builder finish every app?

No. Softr Basic did not reach a working client portal or invoice tracker. Google AI Studio’s portal remained a browser-local demo without a secure backend. Replit’s invoice tracker paused overnight when the free daily quota ran out, then finished the next day.

Is Google AI Studio scored like the others?

No. That session ran on the free tier on Nov 9, 2026, and Lovable was tested earlier on paid Pro (Oct 29). The comparison page leaves AI Studio unscored.

Not a vendor benchmark you can rerun with one click; one run per app per tool.