Start with what is actually observed
The input is a two-dimensional photo. The output is an AI-edited image with a line and a label for each of the two main people. The system is not observing a conversation, measuring affection, or following a relationship over time.
A useful distinction is between observation and interpretation. “The torso appears tilted” describes the image. “This person cares more” adds an explanation that the line itself cannot establish.
Does body-language research validate the trend?
Research on nonverbal communication is a broad field. A 2021 editorial by Patterson and colleagues discusses its research challenges. It does not give a universal dictionary in which each pose has one meaning.
Barrett and colleagues' 2019 review considers facial movements and emotion inference. It is relevant background about interpretation, but it does not test the Green Line Test, couple compatibility, or our AI model. Citing a paper about faces would not establish a claim about who has power in a couple.
What would a useful validation study need?
The following is our explanation of the questions a study would have to answer, not a claim that such a study has been completed for this product.
- A defined claim. Does “works” mean repeatable line placement, agreement with a human annotation, or prediction of a relationship outcome? Those are different targets.
- A defensible reference. A relationship outcome cannot be defined by simply asking whether the green line looks convincing.
- Representative inputs. Include different camera angles, body proportions, clothing, cultures, poses, and photo quality.
- Independent evaluation. Assess images that were not used to create or tune the scoring rule.
- Transparent errors. Report failures, uncertainty, and how the method compares with a simple baseline.
Without that information, a number such as “95% accurate” has no clear meaning. Our product makes no such accuracy claim.
Confounding factors in a couple photo
A photographer can ask one person to lean toward the other. A narrow crop can encourage both people to move inward. A taller person may change their pose to fit in the frame. The camera may also be tilted.
Each of those factors can alter the drawing without changing the underlying relationship. The illustrated results guide shows why a single pattern has several possible explanations.
Why selected examples are not a prediction test
Finding an old photo after a celebrity breakup and drawing a line on it does not demonstrate that the breakup was predictable. You would also need to examine the photos that did not fit the story and decide the rule before looking at the outcome.
The same applies to personal photos: choosing the most dramatic frame can make a visual rule look more consistent than it is. A gallery of matching examples is not a substitute for an independent evaluation.
What we disclose about our own tool
Our method page explains the input, requested overlay, pricing model, and limitations. We ask a generative image model to add posture lines; we do not return a calibrated biomechanical measurement or a validated relationship score.
For a practical answer to “should I believe my result?”, read does the Green Line Test actually work?. For instructions, use the drawing guide. The scientific discussion here is about interpreting evidence, not diagnosing people in photos.