Most “can you clone it?” takes are guesses from a glance. oneprompt doesn't guess. Detailed analysis agents read the actual code, uncover the hidden backend, and reason — with live market research — about whether a site is genuinely worth building with a prompt.
We fetch the page HTML, follow a handful of internal subpages, and pull the site's own first-party JavaScript bundles. The judgment is grounded in the actual code that ships to the browser, not in how the page looks.
Every third-party SDK, API/GraphQL/WebSocket endpoint, auth provider, payment integration and CMS leaves a trace in the code. We detect those signals and translate them into the real backend work behind the façade — the part a pretty clone quietly skips.
A reasoning model weighs façade vs. hidden system and returns two honest estimates: prompts to rebuild just the visible page, and prompts to rebuild the entire project including every backend piece the evidence revealed. Some products (years of backend) it marks as not realistically promptable at all.
For signed-in analyses the agent runs live web research: demand and trend, named competitors, a rough monthly cost to run a clone, and how one could actually make money. Cheap to build but impossible to monetise is a real answer we give.
Everything collapses into a single EASY→HARD difficulty score, the effort in plain terms, and a one-sentence bottom line. No hype — if a prompt can genuinely rebuild it, we say so; if the clone is a mirage, we say that too.
The result is a verdict you can trust because it's built on evidence, not vibes. The crawled page content is always treated as untrusted data — instructions hidden inside a page never steer the analysis.
Analyze a site →