To answer directly: Ask a question. Watch the “thinking” process. Estimate the likelihood that all of the generated text is in the training data.

Do you disagree with my claim?

Or perhaps you were hoping for a very rigorous set of experiments?