Entities
i
62
Runs
i
126
Mentions
i
190
Citations
i
14
This is a read-only demo. Sign in with Google to track your own brand.
Sign in with Google
Grounding Inspector i
Query-centric grounding: what does the model retrieve for a query, what does it keep, and what does each snippet look like.
Runs OpenAI's native grounded search: the model answers your query, and we
capture every source it received, which ones it cited (with the answer
passage each one backs), and the retrieved content per source. OpenAI returns
each snippet as a sliding window it read — tagged with its size
(wordlim: ~100 words for news, ~200 for web/reddit) — so the
content column shows exactly that window, not the full page. Reasoning effort
is off; a run takes ~20s and returns inline.