Use the intelligence. Keep the judgement.

Thinking Scan generates interpretations from visual evidence. Results can be incomplete, outdated, or wrong—even when phrased confidently.

Use a scan to explore and organize possibilities, not as proof of identity, authenticity, value, safety, legality, or condition.

The confidence indicator is directional guidance about the model’s read. It is not a calibrated probability, guarantee, certification, or professional opinion.

Low-quality images, occlusion, uncommon objects, reflections, text at small sizes, and missing context can all reduce reliability.

• Health, medication, food safety, or emergency decisions.
• Legal, financial, employment, insurance, or eligibility decisions.
• Electrical, structural, mechanical, or hazardous-material assessment.
• Authentication, appraisal, provenance, or purchasing decisions of consequence.

Use an appropriately qualified professional and primary sources for high-stakes decisions.

• Use an evenly lit, sharp image with the whole subject visible.
• Add another angle when a feature is hidden.
• Point to a detail for a quick local read; shade a precise area when nearby details could confuse the result.
• Ask for uncertainty, alternatives, and what evidence would confirm the answer.

See the complete operating guide for the full workflow.

A restyle is a generated concept, not a photograph of a finished product or a guaranteed feasible design. Materials, geometry, branding, text, and physical details may change.

Review permissions, intellectual-property rights, and practical constraints before publishing or manufacturing from a generated concept.

Do not use the app to infer sensitive traits, identify people without a lawful basis, surveil individuals, or make consequential decisions about them.

Only upload images you are permitted to process, and avoid private or sensitive material whenever a less sensitive alternative will do.