🏷️ Meta Fields

Meta fields extract generic document information — things like the filename, sender email address, email subject, or full document content — rather than data from the document body itself.

How they differ from schema fields

Regular extraction schema fields extract structured data from the document content (e.g. invoice number, total amount, line items). Meta fields are defined once at the inbox level and are automatically populated for every document that arrives — no AI extraction involved.

Where to find them

Meta fields are managed on the Extraction Schema page, under the Meta Fields tab.

Notes

• Meta fields appear alongside your regular schema fields in exports, webhooks, and integrations

• Their values come directly from document metadata — they are not re-extracted by AI on reprocessing

Confidence Scores

When enabled, Airparser adds a _confidence_ field to every parsed document containing a 0–1 score for each extracted field — for example {"invoice_number": 0.97, "total": 0.84}. A score of 1.0 means the model was certain; 0.0 means the field was not found or the model was very uncertain.

If you enable Confidence Scores after documents have already been parsed, you will need to reprocess them to get the scores.