Direct answer
Run one invented case that looks like your usual filing through a live demo, and judge the software on what it does with that case. You want to see whether intake answers land in the right fields on every related form, and whether each form is on an edition USCIS accepts on the filing date. You also want to see what happens to a draft when an edition changes, which forms the product files electronically, and whether you can get back the exact copy you filed.
Suppose a firm files family-based adjustment packets every week. Its test case is an invented principal applicant in a family preference category, with a derivative child. Assume both are eligible to adjust and the firm files an I-485, I-765 and I-131 for each. Real client files stay out of a vendor's demo environment. The vendor drives, and the firm's staff record what each form shows.
A clean sample proves little, because copying a name and a birth date onto three forms is the easy part. The test case should carry the facts that are awkward to map: an earlier surname, a long address history with a gap in it, a child who needs a packet of their own, and a correction made after the forms exist.
The firm decides before the demo what counts as a pass on each fact. It scores each one during the demo and keeps the generated PDFs. A second vendor run against the same case then produces a scorecard that can be compared line by line with the first.
Six facts and what a pass looks like
| Fact in the test case | Pass | Fail |
|---|---|---|
| A surname used before marriage | Entered once and printed on every form that asks for other names, spelled the same way | Retyped on the second form, or missing from one |
| More addresses than the I-485's history rows hold, with a three-month gap between two of them | All of them print in date order, the extra entries continue on the additional-information page, and the gap is flagged for staff | Entries past the last row are dropped, or the gap prints with no warning |
| A derivative child with a separate I-485 | The child's packet is generated from the family record, with the shared address and the parent's name carried over | The child is keyed in as an unrelated client |
| An address corrected after the forms are generated | The change reaches every form that carries the address, automatically or through a regeneration step that shows what changed | One form keeps the old address with no warning |
| A required answer left blank at intake | The blank is flagged on screen with the form and field named | The form prints with an empty required field |
| A narrative answer written in Spanish | The answer appears in English for staff to check, and the client's original stays on the record | The narrative prints untranslated, or the original is overwritten |
A product page gives the demo a baseline. eImmigration's form management page says client data entered once generates the I-485, I-765 and I-131 together, and that an address changed in one place updates everywhere. It says validation checks flag formatting issues. It also says the questionnaires run in 13 languages, with translated answers returned in English and the originals kept.
Those statements speak to the surname, the corrected address and the Spanish narrative. They don't describe what happens to address entries that overflow the form, to a gap in the dates or to a derivative's packet. A formatting check also isn't necessarily a check for a blank required field. So those are the rows where the demo adds the most, and the rest confirm the page against your own facts.
If questionnaires supply the case data, start the test at client intake and follow each answer to the printed page.
USCIS tells paper filers to use only the current form edition, to keep the edition date and page numbers visible at the bottom of each page, and to complete every page with the same edition. If pages are missing or come from a different edition, USCIS may reject the filing.
The rule works form by form. The I-485, I-765 and I-131 in one packet each carry their own edition date, and a form started on one edition and finished after a replacement can end up with mixed pages. So read the footer of every page on the finished packet as well as on the first draft.
Two editions of a form can be acceptable at once, which makes the cutoff date more useful than the newest PDF. The USCIS Forms Updates page lists a 09/09/26 edition of Form I-129. It says that starting November 9, 2026, USCIS will accept only that edition, and that the 02/27/26 edition can be used until then. Suppose an H-1B petition is drafted on October 20 on the 02/27/26 edition and the filing slips to November 12. The draft was on an accepted edition the day it was written and isn't on the day it goes out.
Software that handles this well shows the edition a draft was built on. When the draft moves to the new edition, it lists any answer that has no matching field. eImmigration lists Form Version Control, Government Instructions, Blank Forms and Automatic Data Population among the 48 standard features in every plan. Its public pages don't describe what version control does to a draft in progress, so open a draft that was started on a replaced edition and watch it happen.
Speed counts too. Ask for the date the 09/09/26 I-129 became available in the product. A dated release note can be compared with the date USCIS posted the edition. And because a count of 300+ forms doesn't say which forms are in the library, open the ones your firm files most.
Generating a form produces a completed PDF. E-filing sends the data or the PDF to an agency system, and the agency decides which forms it takes. USCIS's list of forms available to file online separates forms completed in a guided online workflow from forms uploaded as a completed PDF.
So a product can generate a form that it doesn't submit, and eImmigration's plan table shows the split. It puts form generation in its standard features and lists e-Filing with CBP Data Import, described as submitting forms to USCIS, DOS and DOL, among its 17 premium features. Starter includes one premium feature of the firm's choice, Essentials four and Complete all 17. A Starter account that wants portal submission spends its one choice there.
The public pages don't name which forms go to which portal, so that list is what the demo has to produce. A pass is a named set of forms for each portal, the steps staff still do by hand, and a place on the case for the agency's receipt. The product marking a form as sent is an earlier event than the agency confirming it arrived, and the record should keep the two apart. Run this test as a simulation. Have the vendor walk through the submission screens for the form your firm files most and stop before anything is sent, because an invented case shouldn't reach an agency.
Two more tests cover what happens around the form. For attorney review, have a second user open the packet and send one field back for correction. It's a pass if the change is recorded with a name and a time, and if the form can't be marked final until the correction is cleared.
The last test is the filed copy. When a filing is questioned months later, the firm needs the PDF as it was filed, with its edition date. A form regenerated from current data is a different document, because the client's address may have changed and USCIS may have replaced the edition. eImmigration's case management page describes a caseworker portal that holds cases, clients, forms and billing together. It doesn't say whether the submitted PDF is stored as a fixed copy.
The test is short. Have the vendor close a sample case in the demo account, change the client's address, and then retrieve the filed PDF. The copy passes if it still shows the old address and its original edition date. Where that record lives is part of the difference between forms software and case management software.