Hallucinated citations are prevented by design: the AI can only draft from cases confirmed in the court record and read in full. Then a rival AI checks every claim against those opinions, and anything unsupported is flagged before it reaches your brief.
Two briefs from the live system, unedited, each with the cases it read and the verdict on every claim. The live system runs on request: ask through the access form.
9 cases read · 35 of 36 claims grounded · 23 min
6 cases read · 6 of 7 claims grounded · 1 flagged · 10 min
Patent pending
Most legal AI bolts a citation checker onto a general-purpose model. The model writes from memory, and then something tries to clean up the mess. JuriScripta was engineered in the opposite order, and that order is the invention behind our pending U.S. patent.
General-purpose AI
JuriScripta
Every stage can reject what the stage before it produced, and every rejection is shown to you. Nothing is swept under the rug.
Both kinds of fake citation, handled
The headline-grabbing kind. Here it simply can’t happen: a citation that isn’t confirmed in the court record never makes it into the draft. It isn’t filtered out afterwards. It never gets in.
The sneaky kind, and the one citation checkers wave straight through, because the case is real. JuriScripta rereads the opinion and tests the claim against it. If the case doesn’t say it, you see it in red at the top of your brief.
Two gates, in order
What no chatbot can do
Full opinions, fetched and read before the first sentence is drafted. The AI isn’t remembering the law. It’s quoting it.
Immigration law changes faster than any AI can be retrained. Every model stops learning on a fixed date. JuriScripta follows the citation network forward to decisions issued after that date, reads them, and flags them as new.
A second model from a different company rereads each opinion and attacks every claim in the draft. Two competitors rarely share a blind spot. Whatever fails is flagged in red at the top of your brief.
The problem
Ask a chatbot a legal question and you get a confident answer with citations attached. Some are real. Some are invented. Some are real cases that say something completely different. Lawyers have been sanctioned for filing them. In immigration court, it is your client who pays for the mistake.
Checking whether a case exists catches the invented ones. It misses the real case that was misread, which is the more common error. Catching that takes the opinion itself and something willing to read it twice. That’s what we built.
One run, start to finish
What comes back
Every brief ships with the cases it used, the cases it tried to use and was refused, and a verdict on every single claim. Problems sort to the top. You never have to go digging for the one thing that matters.
Example
3 claims checked · 2 grounded · 1 not supported
| Claim | Authority | Verdict | Why |
|---|---|---|---|
| A notice to appear that omits the hearing time deprives the immigration court of jurisdiction. | Pereira v. Sessions, 585 U.S. 198 (2018) | Not supported | The opinion decides the stop-time rule and says its holding is narrow. It does not address jurisdiction. |
| A notice to appear that omits the time or place of the hearing does not trigger the stop-time rule. | Pereira v. Sessions, 585 U.S. 198 (2018) | Grounded | Stated as the holding. |
| The required information must arrive in a single document, not across several. | Niz-Chavez v. Garland, 593 U.S. 155 (2021) | Grounded | Stated as the holding. |
Who it is for
Motions to reopen, BIA appeals, petitions for review. A brief-ready research pass while you’re in court on something else.
Board of Immigration Appeals and federal court decisions, including the newest rulings, the ones no AI model has ever seen.
A first pass on every matter, with the weak spots already marked before an attorney or accredited representative picks it up.
Built for flat fees
About fifteen minutes for a researched, claim-checked first draft, because it reads every opinion it cites. You bring the judgment. It brings the hours of reading you didn’t have.
Limited beta
Every brief runs real research, so we’re opening access in small groups. Tell us who you are and what you’re working on, and we’ll get you in.