#target pool

On most of these tools the target ID and the feedback are the same object. The random number you are handed is a link, and one click on it opens the photo. That is convenient, and it is also a loaded spring. If you peek, hover, or let your eye drift to a preview before your notes are closed, you have front-loaded yourself, and everything you write after that is memory dressed up as signal. Treat the ID as a sealed envelope. Copy the number over to wherever your session notes live, and do not go back to that tab until the page is shut and your notes are final.

Some generators also show you a category before you start: location, person, object, event. It feels harmless. It is not. The moment you know it is a person, your descriptors quietly narrow to person-shaped things and you stop describing what is actually arriving. If the tool has a category toggle, turn it off. You want the cue to be nothing but an arbitrary handle with no meaning attached.

The second problem is the feedback itself. A lot of these pools are one photo per target, and often an iconic one: a famous tower, a landmark, a stock sunset. Two things go wrong there. A single frame lets you score a loose vibe match instead of checking descriptor by descriptor, and iconic targets tempt you to guess from a mental shortlist of famous places instead of describing raw sensory data. Both feel like hits. Neither one is.

So favor targets with concrete, verifiable feedback and real sensory contrast, and score the raw descriptors, not the label. Tall, hard, vertical, windswept is a clean hit on a bridge even if you wrote the word building. I felt like it was famous is not a hit on anything. And keep accuracy and confidence in separate columns while you are at it, because how sure you felt is not evidence of whether you were right.