What AI habit verification can—and can’t—verify

Life Leveling compares a submitted photo with the selected habit and its description. The AI judges whether the visible subject reasonably supports that habit, and returns a result with an explanation. For broad activities, relevant equipment, a setting, or the aftermath can count as context.

A workout check-in, for example, may show a gym or exercise equipment. That is evidence associated with a workout, but it is not a measurement of your entire session. A habit naming a particular object requires a clearer match to that object than a broad activity does.

This makes the feature an accountability aid. Its job is to assess the image you provide; the meaning of your completion record still depends on the action you actually took.

Before treating a photo as proof, describe what someone can directly see. Then describe what you are asking them to infer. The gap between those two statements is the limit of the evidence.

The image may be blurry, dark, tightly cropped, or full of distracting objects. The habit may also ask for something the picture cannot show. Even a clear image can receive an incorrect interpretation from an AI system.

There are two useful cases to distinguish. A rejected image can follow a real completed habit when the evidence is unclear. An accepted image can show relevant context even when the full activity did not occur. Neither outcome changes what you actually did.

Do not treat acceptance as a guarantee that a photo is authentic, current, or a complete account of the action. A photo check should not be presented to a partner as an infallible judgment about their honesty.

Read the explanation first. If the subject is not visible, take a clearer photo of the relevant evidence while it is available. If the habit is ambiguous, clarify what you intend to track for future check-ins. Avoid repeatedly changing the wording just to get an unrelated image accepted.

If you did not complete the activity, do not count an accepted image as completion. If you did complete it but cannot supply useful evidence, keep the distinction clear in your own notes or in a conversation with your partner. Changing the record to look better defeats the purpose of accountability.

For a recurring mismatch, use the support link and describe the issue. Share only the information needed to explain it; a sensitive photo is not automatically necessary for a useful support report.

Define the behavior independently of the app’s response. “Practice drawing for ten minutes” is a commitment you can evaluate yourself. “Get a photo accepted” measures success at submitting an image, which is a different task.

When using Duo, agree on that distinction together. Use proof as context for a conversation, and use accurate self-reporting for details the camera cannot capture. That leaves room for useful evidence without expecting a photograph to answer every question.