Lesson 3 of 5
Understanding AI similarity
Read the score as an AI comparison of your submission and the target, with limits.
The AI evaluator compares the material you submitted with the revealed target using its current instructions. Its percentage summarises that comparison. Read the accompanying feedback and check whether the named matches and mismatches are visible in the original material.
The percentage is not a probability that you perceived hidden information, and it is not a measured advantage over chance. Free Response’s historical 25% reference is a product constant, not an experimentally established chance baseline. Different targets, hints and evaluator versions can affect comparability.
An illustrative example
If an evaluation reports 60% AI similarity, that does not mean “a 60% chance of a real ability” or “60% of the photo predicted correctly.” Look at which details support the comparison and which details contradict it.
Try this in your next session
After your next evaluation, check one claimed match and one mismatch against the sketch and photo. Keep your own review beside the AI feedback, including any disagreement. Do not change the original submission.
Try a blind-photo sessionKeep in mind: Use AI feedback as a review aid. A score alone is neither scientific evidence nor a verdict about you.