Short answerMost misses come from unreadable files (scans), very large mixed documents, or a question that does not match the wording in the file. Convert to text, split by topic, and test with the exact words people use.
Check in this order
- Is the text selectable? Scanned PDFs may have no text layer: fix scanned PDFs.
- Was the file actually added? Some platforms need a retrain or reindex step after changes.
- Is the file too big or mixed? Split by topic: splitting files.
- Does the question match the wording? Retrieval finds passages that resemble the question. Add a short FAQ file using real user phrasing.
- Do the instructions tell it to use the files? Add "Answer from the provided reference material; say you do not know if it is not there."
Verify
Ask a question whose answer is in a single paragraph of one file. If it still fails, the problem is upload or format, not wording. Background: how retrieval works.
Check your GPT before you move it
Paste your instructions, file names and Actions schema into the free checker. It runs in your browser and builds a portable bundle. Nothing is uploaded.
Open the free checker