Synthetic I-9 Employment Eligibility Verification Data

What this document is

Form I-9 is the employment eligibility verification form every U.S. employer must complete for every person they hire, without exception. Section 1 is the employee's own attestation of citizenship or work-authorization status; Section 2 is the employer representative's record of the physical identity and work-authorization documents they examined; Supplement A covers a preparer or translator who helped the employee; Supplement B covers rehire and reverification. It is the single most universally required onboarding document in the United States and the anchor of every E-Verify integration.

Why generate synthetically

A completed I-9 carries a Social Security number, a date of birth, a home address, a citizenship attestation, and one or more government identity-document numbers on the same sheet — and unlike most HR paperwork it is subject to a statutory retention period and to anti-discrimination rules about what an employer may even record. That combination makes real I-9 files effectively unshareable, so I-9 extraction models are routinely trained on a few dozen hand-filled samples and then deployed against every document combination in the country. Synthetic I-9s remove the constraint and let you weight the rare paths — permanent residents, employment-authorization documents, preparer-assisted forms, reverifications — at whatever frequency your evaluation needs.

What makes synthetic data useful

The citizenship attestation drives the rest of the document. A simulated employee who attests to U.S. citizenship gets a passport, a passport card, or a driver's-license-plus-Social-Security-card pair; a lawful permanent resident gets a Permanent Resident Card with a USCIS number that also appears in Section 1; an alien authorized to work gets an Employment Authorization Document with an expiration date that matches the Section 1 work-authorization expiry. Section 2 is completed by a named employer representative from the same simulated employer that appears on the employee's payroll documents, and the first-day-of-employment date is consistent with the attestation and signature dates.

Training challenges

Section 2 is a three-column grid — List A, List B, List C — where the correct answer is always exactly one column filled and the other one or two entirely blank, and the columns sit side by side with identical sub-labels (Document Title, Issuing Authority, Document Number, Expiration Date). An extractor that reads the grid positionally rather than as mutually exclusive lists will happily emit a List A title against a List B document number. The List C expiration cell carries the literal string "N/A" rather than a date on essentially every document, which breaks any date parser applied blindly to the column. Section 1 contains four mutually exclusive citizenship checkboxes whose selection gates whether the USCIS-number, I-94 and foreign-passport fields below carry values at all. Supplements A and B are full pages that are blank on the overwhelming majority of documents, so a model must learn that an empty page is a valid, expected page rather than a failed scan.

Who uses this data

Employment-eligibility verification platforms and E-Verify integrators, HR onboarding suites that ingest scanned new-hire packets, staffing and PEO document-automation teams, and I-9 audit and remediation vendors who have to grade thousands of historical forms for completeness. It is also the standard demonstration document for any IDP vendor selling into HR, because every prospect already knows what a correct one looks like.

Document complexity profile

67 fields laid out across four pages, of which three carry fields: 57 text, 4 checkbox, 4 signature and 2 date targets. Binding logic is deliberately thin — a single arithmetic binding and two function calls at a maximum expression depth of 1 — because the document's coherence comes from a registered computed-field module that resolves the citizenship attestation into a matching document set before any field is evaluated, rather than from expressions on the page. The extraction difficulty is structural: mutually exclusive List columns, attestation-gated regions, and two mostly blank supplement pages.

Key stats from our synthetic corpus

Quantitative characteristics of the I-9 Employment Eligibility Verification documents our generator produces.

How this document co-occurs with others

Rates at which identities in our corpus that produce a I-9 Employment Eligibility Verification also produce other documents.

Distribution figures above are corpus-derived: they were computed over the 641 employment-eligible identities inside a local synthetic corpus of 1,000 identities generated by SymageDocs' World Simulation Engine at seed 20260421. Field, page and type counts come from the shipped I-9 definition in the SymageDocs form library. No real employee, employer, or immigration data was used at any stage.

Frequently asked questions

What data format do synthetic Form I-9 documents include?

Each generated identity produces a rendered PDF plus a structured JSON annotation file with bounding boxes, field types, and ground-truth values for all 67 fields across the four pages, including the List columns and Supplement pages that render blank. COCO, YOLO, FUNSD, and BIO/NER exports come from the same job.

Do the Section 2 documents actually match the Section 1 attestation?

Yes. The document set is derived from the attested status rather than sampled independently, so a lawful permanent resident is never presented with a birth certificate as List C evidence, and an alien authorized to work carries an Employment Authorization Document whose expiration date agrees with the work-authorization expiry recorded in Section 1. That coherence is what makes the corpus usable for compliance-checking models, not just field extraction.

Are the Supplement A and Supplement B pages ever populated?

Yes, on a realistic minority of documents. Supplement A (preparer or translator certification) is completed on roughly one in forty generated I-9s, and Supplement B (reverification and rehire) on roughly one in twenty-five. Both pages are annotated on every document whether populated or not, so you can train and score the blank-page case that dominates real I-9 files.

Can I use this data commercially?

Yes. Every name, Social Security number, address, USCIS number, and document number is synthetic, contains no real personal data, and is licensed for commercial use including model training, benchmarking, and redistribution inside your own products. No consent, no redaction, no data processing agreement.

Can I generate a full onboarding packet around the I-9?

Yes. Every I-9-eligible identity also generates a W-4, a direct deposit authorization, a personal information sheet, and the other onboarding forms in this catalog, all for the same simulated employee and employer. That gives you a multi-document packet with a shared identity — the classification-plus-extraction task an onboarding platform actually faces when a new hire uploads a scanned bundle.