# Can Tessary's groundedness classifier check a very long answer?

Up to a point. Tessary's groundedness classifier reads an answer and its retrieved documents together in a single pass capped at 8,192 tokens. When the two don't both fit, the answer itself is cut at a fixed length first. If a long answer, even cut, still leaves no room for a single document, that answer isn't scored at all.

This matters most on a `summarize` call site, where a long answer competes with its own source document for the same budget, more directly than a typical `rag_answer` call citing a handful of short chunks. [Fitting more into one context window always means giving something else less room](/answers/retrieval-augmented-generation/why-cant-i-just-increase-the-context-window-instead-of-using-rag), whichever system is doing the fitting.

An unscored answer isn't the same as a passing one. It's simply absent from the call site's learned rate, so a call site that mostly produces long answers can look quieter than it really is.

---

Source: https://tessary.ai/answers/tessary-groundedness/can-groundedness-check-a-very-long-answer
More on Tessary groundedness: https://tessary.ai/answers/tessary-groundedness
From Tessary, agent reliability for AI agents in production: https://tessary.ai