Am I correct in assuming that after an answer to a novel prompt is posted, it doesn't work as a reasoning test of LLM deployments that search the web?

Edit: an incorrect answer could degrade its performance too.