Authorised Release Certificate OCR turns a scanned or PDF EASA Form 1 into fields a receiving desk or an MRO system can use. Before picking a tool, check three things: which of the 14 boxes it returns, whether each value points back to the page it came from, and what it does with a box it cannot read.
TL;DR. Generic OCR returns text. Form 1 extraction returns boxes: tracking number (3), part number (8), quantity (9), serials (10), status/work (11), remarks (12) and the release block that was signed (13 or 14). Nanonets' ARC OCR page lists tracking number, invoice number, item, description, serial number and quantity. Doana returns boxes 1 to 14 with a confidence score and source page on each value, and leaves an unreadable box empty. Neither tool certifies airworthiness.
What is an Authorised Release Certificate?
It is the name printed in Block 2 of EASA Form 1: AUTHORISED RELEASE CERTIFICATE EASA FORM 1. EASA describes Form 1 as the certificate a production organisation (POA) issues for a new part, and one a maintenance organisation can also use after work on a component (FAQ 19466). The US counterpart is FAA Form 8130-3. Canada's is TCCA Form One. The same words, "Authorized Release Certificate", appear on all three, which is why "ARC OCR" searches mix them.
Block-by-block rules for issuers are in How to fill EASA Form 1. This page is about reading a certificate someone else issued.
What should ARC OCR extract?
The boxes a receiving inspector checks, in the structure the form uses:
| Box | Field | Why a reviewer needs it |
|---|---|---|
| 1 | Competent authority and country | Tells you which rule set issued it |
| 3 | Form tracking number | Unique reference; a correction carries a new one |
| 4 | Organisation name and address | Matches the approval number in 13 or 14 |
| 5 | Work order, contract or invoice | Links the certificate to your PO and invoice |
| 6 to 10 | Item, description, part number, quantity, serial | One record per line, serials as a list |
| 11 | Status/work | NEW, PROTOTYPE, or the maintenance terms (Overhauled, Repaired, Inspected/Tested, Modified) |
| 12 | Remarks | Dual release statements, AD/SB, LLP or shelf-life notes |
| 13a to 13e | Production release | Filled on new parts |
| 14a to 14e | Maintenance release | Filled after maintenance |
A tool that returns only part number, serial and date leaves out Blocks 11, 12, 13 and 14. Those are the boxes that decide whether the part can be accepted.
How do the options compare?
Three kinds of software read ARCs today. The table describes each category from public documentation reviewed in September 2026. Check each vendor for current capabilities.
| General document OCR / IDP | Aviation document AI | Doana | |
|---|---|---|---|
| Example | Nanonets ARC OCR: tracking number, invoice number, item, description, serial, quantity; other fields on request (Nanonets) | ePlaneAI Document AI: part number, serial, date and certifying agency from 8130 certificates, plus indexing for RAG pipelines (ePlaneAI) | EASA Form 1 boxes 1 to 14, one record per item line |
| Release blocks 13 and 14 | Not in the listed fields | Not in the listed fields | Returned as two separate objects, sub-boxes A to E |
| Evidence per value | Varies | Varies | Confidence (0 to 1) and page number on every value |
| Unreadable box | Varies | Varies | Returned empty, never filled by a guess |
| Where it runs | Vendor web app and API | Vendor platform and API | ChatGPT, Claude or your own agent over MCP |
ePlaneAI also covers 8130-3, which Doana does not extract box by box. If your inbound flow is mostly US tags, weigh that. If it is EASA Form 1, compare the release blocks.
What happens when a box cannot be read?
That decides whether the output is safe to import. A plausible wrong serial is worse than an empty box, because nobody checks a value that looks right.
In Doana's easa-form-1-v1 result, every field has value, confidence and page. A box that is blank, N/A or unreadable returns an empty value. Any non-empty value under 0.7 confidence is listed in uncertainFields, so the reviewer opens only those. If the file is not an EASA Form 1 (an 8130-3, a TCCA Form One, a UK CAA Form 1, an invoice) the result is documentType: unknown instead of boxes forced into the wrong layout.
Which files work?
PDF up to 50 MB and 20 pages, or a JPEG or PNG scan up to 20 MB. TIFF is rejected. A scan too degraded to identify the form comes back as unknown. Scan-quality detail: How to extract an EASA Form 1 with ChatGPT.
Is ARC OCR a check of authenticity?
No. OCR reads what is printed. It does not confirm that the organisation in Block 4 holds the approval in Block 13 or 14, that the signature is genuine, or that the part is eligible for installation. Those stay with a qualified inspector. Extraction removes the retyping before that check.
Looking for a Nanonets alternative for Form 1?
Nanonets is a general document AI platform: its ARC page lists six fields, and other fields are added on request. Pick it if you process many document types and your team is ready to configure templates. Pick an aviation-specific extractor if the release blocks, Block 11 status and Block 12 remarks are what your receiving decision depends on. Doana returns those boxes self-serve for EASA Form 1, and builds custom pipelines when the file also holds 8130-3 tags, AWBs or T1 documents.
Next
If you receive EASA Form 1 and need every box, try the Form 1 extraction page or connect the EASA Form 1 MCP server to ChatGPT or Claude. If most of your certificates are US tags, read How to read FAA 8130-3 with ChatGPT first.
Sources
- https://www.easa.europa.eu/en/faq/19466
- https://nanonets.com/document-ocr/authorised-release-certificate
- https://www.eplaneai.com/solutions/document-ai
- https://www.legislation.gov.uk/eur/2012/748/annex/I/appendix/I
- https://doana.ca/en/easa-form-1-automation