How we test
Every answer on Flyder comes from runs we did ourselves. This page explains how a run works, what we publish and how to dispute a result.
Discovery and liveness
Our crawler finds apps on builder deployments, GitHub, launch sites and author submissions. An app counts as live when it answers, is not a builder placeholder and shows real content. Two failed checks in a row take it out of every answer.
Scenarios by task type
Each question has a task type (site widget, messenger bot, generator, integration). The scenario comes from the task type, never from the app or its author, and the test data is generated for every run.
The run
An agent opens the app in a clean browser, signs up with a generated address, does the task on our test site and records every step with a screenshot. A separate judgement step decides works, partial, fails or blocked against the scenario's criteria.
Publication and freshness
A question is published only when at least three apps work or partially work, tested within the last 90 days. Evidence expires after 60–120 days depending on traffic and is re-tested.
What money never buys
Payments buy speed of testing, a Pro profile or API volume. They never change a result or a position; a test switches off every paid flag and checks that the order is identical.
Corrections and disputes
Authors can dispute any run. A dispute starts a new run with new test data; the old run stays public and is marked as replaced. Report errors to corrections@flyder.com.