How to Check an AI Before-and-After Photo

1. Match the view and scale

Put the original and generated image beside each other at a similar face size. Check eye level, head tilt and the position of the mouth. If the generated face is rotated or enlarged, a feature may appear different simply because the view changed.

Use the original source, not a screenshot of an earlier preview. A screenshot can introduce resizing or compression that makes fine edges harder to compare. Do not rely only on a small thumbnail: it can hide a blurry mouth, a doubled outline or changes around the nostrils.

2. Point to the requested difference

Read the brief again and name the exact feature it asks to change. For lips, check the upper and lower lip separately. For a side-profile nose request, follow the bridge and tip silhouette. For a chin request, examine the chin outline while keeping the jaw and mouth in view.

If you cannot point to a visible difference at the requested feature, classify the result as insufficient. A brighter face, smoother skin or a new expression is not a substitute. This matters especially when the requested edit is subtle: subtle should still be distinguishable from no change.

3. Check what should have stayed unchanged

Inspect the eyes, eyebrows, teeth, smile, hairline and background. A model may produce a plausible face that is no longer a faithful edit of your photograph. Check these areas even when the target feature looks good.

• Lips: Did the smile, teeth or lip color change?
• Nose: Did the face angle, upper lip or eye spacing change?
• Chin: Did the neck, jawline or mouth move as well?

Do not attribute unintended changes to a treatment. If several unrelated features moved, the image is a poor basis for comparison.

4. Give the result an honest label

Use a simple three-way decision. Useful concept: the intended change is visible and unrelated details remain reasonably consistent. Insufficient edit: the requested difference is missing or too unclear to judge. Unreliable comparison: artifacts, identity changes or a different pose undermine the result.

This is a practical review checklist, not a validated medical score or an automatic quality guarantee. A useful concept can still show a result that is not achievable through treatment. No image-only checklist establishes medical feasibility.

5. Revise the cause rather than repeating blindly

If the target feature was tiny or out of focus in the original, choose a clearer source. If the brief was vague, specify one feature, a degree of visible change and the details to preserve. If the model changed unrelated areas, explain that error in the next brief.

For example: “The previous version changed my smile but barely changed my upper lip. Keep the original smile and teeth; only add modest upper-lip fullness.” Review any revised plan and the displayed credit cost before creating another image. Repeated generations are not guaranteed to improve the result.

Use the nose preview guide or lip photo guide to prepare the next attempt. Save the source, brief and concept together if you plan to discuss a preference with a clinician.

Related guides

• AI Preview vs Clinical Before-and-After: How to Tell Them Apart
• How to Use Reference Photos for a Plastic Surgery Consultation
• How to Use an AI Photo Simulator on Your Phone