The Reasonable Inquiry · Issue 001 · 8 September 2026
The case was real. The lawyer still got the order.
Most lawyers check citations by asking one question: is this a real case?
That question is why they're getting sanctioned.
Here's the thing the sanctions orders keep proving. The dangerous citation isn't the invented case. It's the real case at the wrong address, or the real case with an invented quote. Those pass the "does it exist" test every time. And they're now the majority of what courts are catching.
This letter is about doing the inquiry Rule 11 actually asks for. Four parts, twice a month. No spectacle. Lessons.
The Three Questions
In late August, a federal judge in Colorado ordered a lawyer to show cause over a citation the court called "the nonexistent case." Ricks v. Starbucks.
The lawyer's response: the case exists. The quote exists. Here they are.
He was right. The case was real. What wasn't real was the citation — the volume, reporter, and page pointed at a different opinion entirely. Real case. Wrong address. And it still cost him a show-cause order, because from the bench a wrong address and a fabrication look identical until someone proves otherwise.
That's how AI builds a bad citation. It doesn't invent from nothing. It assembles plausible parts: a real caption, a real reporter, a page number in range. Sometimes every part is real and only the combination is false. "Does it exist?" clears it. Only one thing catches it: asking the questions in the right order.
Question 1 — Does it resolve? Take the volume, reporter, and page to the source. See what opinion is actually sitting there.
Question 2 — Is it the case named? Compare the caption, court, and year of what you found to what your brief claims. Short forms are fine. A different case is not.
Question 3 — Is the quote in it? If your brief has quotation marks, find those words in the opinion. Fabricated quotes inside real cases are the fastest-growing category in the orders.
Notice what's not on the list: does the holding support my argument? That's Question 4, it's yours, and no tool does it for you. But 1 through 3 are mechanical. They take minutes. And skipping them is the entire sanctions docket.
Rules Watch
Two states just changed what your signature means. Neither one asks whether you used AI.
Florida — Rule 2.515(d)(2), effective June 15, 2026. Your signature on a filing now represents that the authorities cited exist and are accurately cited. Not "I disclosed AI." Not "I was careful." Accurate. The signature is the certification.
New York — 22 NYCRR Part 161, effective June 1, 2026. System-wide: attorneys certify that filings contain no fabricated or fictitious AI-generated content. Individual judges' standing orders stack on top.
Read those together and the trend is obvious. The profession spent two years arguing about disclosure. The courts moved past it. They want accuracy, and they want it certified. Add the hundred-plus active standing orders on the trackers and a 2025 Federal Judicial Center survey where 62% of federal judges said disclosure rules should be mandatory — and "I checked" stops being something you say. It becomes something you have to be able to show.
By the Numbers
Before we sent this, we tested our own tool against the record. Not a demo. The record.
We pulled the fabricated citations named in 30 published sanctions orders, 2023 through last month — 136 citations, each with the court's own finding attached — and ran every one through the same code Cite11 uses in production. No tuning. One pass.
Of the fabricated citations the public record could test, zero were cleared as verified — across five runs. Where the record could check them, 97% were flagged: every one of the nonexistent cases, and nine in ten of the real cases cited at the wrong address. The rest were refused as unverifiable — which is the correct answer when a checker can't check, and the opposite of what most tools do.
Now the number you won't see in anyone else's marketing. On the real citations, the first run raised warnings it shouldn't have — and nearly all of them traced to two causes worth knowing even if you never use a tool:
- Short-form captions. "Bell Atl. Corp. v. Twombly" was being compared too strictly against the official caption. Every brief cites this way. Fixed.
- Westlaw-only citations. A "WL" cite doesn't exist in the public record, so it cannot be verified against it. That's not "wrong." That's "outside coverage," and the tool now says exactly that instead of guessing.
After the fixes, 5 of 59 distinct real citations still drew a warning — 8.5% — and three of those five are recent opinions the public record hasn't indexed yet. The tool says "can't verify" for those, not "wrong."
The full benchmark — every order, every citation, every miss — is at cite11.com/benchmark. Here's a rule for the whole category: if a verification tool won't show you its misses, don't trust its hits.
Case File: Adams v. Matrix Providers, D. Colo., Aug. 27, 2026
The citation: Ricks v. Starbucks, cited for a quoted proposition.
The problem: The address pointed to a different opinion. The court couldn't find the case and ordered counsel to show cause.
The defense: Case real, quote real, citation wrong.
The lesson: "The case is real" is what you say after the order. Question 1 is how you never get the order.
One honest note: this citation is Westlaw-numbered (2023 WL 8872476), so Cite11 reports it as outside its coverage rather than guessing — confirm a WL cite like this one in Westlaw directly.
Also on the docket this fortnight: Ruiz v. Magellan Financial (D. Ariz., Aug. 24) — fabricated quotation in a real case, ChatGPT-assisted. Douglas v. Deutsche Bank National Trust (D.C. Ct. App., Sept. 3) — AI-generated citations in an appellate filing. Capital Standard v. U.S. Bank (Fla. 2d DCA, Aug. 21). Booker v. The Kroger Co. (N.D. Ga., Aug. 28). Each gets a registry entry at cite11.com.
Every citation in this issue was run through Cite11 before we hit send. The single lookup is free at cite11.com/lookup. A full Verification Report — dated, numbered, yours to keep — is what the tool is for. Cite11 verifies that citations exist, resolve to the case named, and contain the language quoted. It does not tell you whether a case is still good law. That part is still your job.
Published by Legal Search Hub LLC, which also runs Cite11 and ClearStatute. Reply to this email. A person reads it.