Give the critique a task, not just a screen.
A screen alone cannot explain who is using it or what they need to accomplish. CritiqueAI pairs screen uploads with optional task, platform, context, and website URL inputs. This case study examines the public workflow form and the UX reasoning behind that input structure; the internal agent graph and generated outputs have not been inspected.
Designers and product teams seeking a second perspective on an interface
A UX reading of the supplied artifacts. The sequence below explains the design logic, not a verified chronology of research. Intended benefits and proposed tests are not measured outcomes.
The screen upload is the primary input. Local uploads and pasted file links give teams two ways to provide the same visual evidence.
The public form labels user_task, platform, context, and Website URL as optional. These fields can give the critique a purpose beyond visual taste, while keeping the initial interaction lightweight.
An explicit Execute action gives the user a review point before sending their screens to an external workflow. Run Once and Run Batch are visible modes; batch behavior and output quality remain untested.
A proposed next step is to compare feedback on the same screen with and without task context. Ask designers which observations are evidence-based, actionable, or incorrect.
A quick first run reduces setup friction; task context can help distinguish an actual usability problem from a stylistic preference.
Without context, feedback may assume the wrong audience or task. Suggested follow-up: explain what each optional field improves.
An AI critique should inform a design decision rather than automatically approve or reject an interface.
Human review adds effort, but unsupported or conflicting recommendations need to be challenged.
Screens can contain customer data or unreleased work. Use redacted examples and review the hosting service’s policies before uploading.
Redaction removes some context; a safe synthetic screen may be preferable for public demonstrations.