Geddes v. LoanCare: Attorney-Plaintiff Sanctioned for Fabricated Quotation Marks Around AI-Generated Text | Advocate Prakhar

Case at a Glance

Background

Krista Geddes, a California-licensed attorney, brought a pro se action against LoanCare, LLC and the Federal National Mortgage Association (FNMA) relating to a wrongful foreclosure on her property. The case raised the question of whether Trustee Corps, the foreclosure trustee, was a diverse party for purposes of establishing federal jurisdiction, or whether it was a nominal party that should be disregarded.

At a motion to dismiss hearing in February 2026, the court directed supplemental briefing on the jurisdictional question. Geddes filed supplemental briefing citing two California cases to argue that Trustee Corps could not be treated as a nominal non-diverse party. Defendants filed a reply alleging that the quotations Geddes attributed to these cases were fabricated. The court ordered Geddes to respond, and she acknowledged the quotes were not verbatim — but characterised them as “paraphrased summaries.”

The AI Issue

The court’s analysis of the AI issue is precise and methodologically significant. In her supplemental brief, Geddes placed quotation marks around two passages. One purported to quote Kachlon v. Markowitz for the proposition that a trustee’s immunity “may be lost where the trustee acts with malice or exceeds the scope of its statutory duties.” The other purported to quote Barrionuevo v. Chase Bank for the proposition that “[b]ecause Plaintiff’s challenge the validity of the foreclosure process itself, the trustee defendants cannot be considered nominal parties.” Neither passage appears in the cited cases. The Barrionuevo case does not even address nominal party status.

The court’s critical factual finding: the same brief contained other sentences that paraphrased cases — and those sentences did not have quotation marks around them. The selective use of quotation marks demonstrated that Geddes knew the difference between paraphrase and direct quotation. By choosing to use quotation marks for the fabricated passages, she represented to the court that those were direct quotes. That was a deliberate misrepresentation, not an innocent error.

What the Court Decided

• Sanctions imposed under Rule 11 and California Rule of Professional Conduct 3.3(a)(2), which prohibits knowingly misquoting an authority to a tribunal.
• Geddes was ordered to pay monetary sanctions.
• A copy of the sanctions order was directed to be served on the California State Bar for review.
• Court rejected Geddes’ defence that the errors were unintentional “paraphrased summaries” — the deliberate use of quotation marks around fabricated text precluded that characterisation.
• The court distinguished between pro se non-lawyers (treated more leniently) and licensed attorneys acting pro se (held to full professional standards).

“Plaintiff’s choice to put quotation marks around specific sentences, indicating a direct quote, but those sentences were not found in the cited opinions. The difference in how these sentences were formatted demonstrates intent to misrepresent authorities.”

— Magistrate Judge Dennis M. Cota, E.D. California, April 22, 2026

The India Angle

Indian Law Equivalent

In India, the equivalent professional obligation is found in Bar Council of India Rule 52 (Chapter II, Part VI of the BCI Rules), which makes it a duty for advocates not to knowingly misquote cases or misstate legal principles to courts. India’s higher courts have inherent jurisdiction under Section 151 of the Code of Civil Procedure and Articles 129 and 215 of the Constitution to punish for contempt, which extends to wilful misstatement of the law. An advocate who places fabricated quotation marks around AI-generated text and submits it to the Supreme Court or a High Court could face contempt proceedings in addition to disciplinary action before the Bar Council.

Bar Council Rules

BCI Rule 49 (prohibition on misstatements of fact or law) and Rule 52 (accurate citation of authorities) directly apply. The use of deliberate quotation marks to represent fabricated text as direct quotes would likely be characterised as dishonest conduct under Section 35 of the Advocates Act, 1961, which empowers the Bar Council to suspend or debar an advocate for professional misconduct.

Practical Advice for Indian Advocates

• Never use quotation marks in court submissions unless you have verified the exact text against a primary source — AI’s reproduction of case language is unreliable and frequently paraphrased or invented.
• When using AI to assist with research, structure your workflow so that AI identifies potentially relevant cases and you then retrieve and read the actual text from authenticated databases (SCC Online, Manupatra, or official court websites).
• A licensed attorney who represents themselves in court is subject to full professional obligations — the pro se status provides no shield against disciplinary action for misrepresenting the law.

Quick Takeaways

• Placing quotation marks around AI-fabricated text is evidence of intent to deceive — courts will not accept a “paraphrase” defence when quotation marks were deliberately chosen.
• Licensed attorneys acting pro se are held to the same professional standards as when representing clients, and sanctions orders may be referred to the relevant state bar.
• The distinction between paraphrase (acceptable without quotation marks) and direct quotation (requiring verified accuracy) is a simple rule that AI makes dangerously easy to violate.

Deep Dive: The Quotation-Mark Problem — How AI Encourages the Most Dangerous Citation Error

The Geddes ruling cuts to the heart of one of the most dangerous AI-era litigation errors: the fabricated direct quotation. When AI generates a response summarising a case, it typically writes in fluent prose that resembles the style of legal opinions. Users who copy this prose and place quotation marks around it — or who ask AI to “quote” from a case — are highly likely to receive fabricated text, because AI language models do not retrieve stored case text; they predict what plausible case text would look like based on patterns in their training data.

This is categorically different from citing the wrong case or mischaracterising a holding. A fabricated direct quotation is a lie about what a judge said. It invents judicial language. Courts regard this as a more serious violation than mere citation error, because it directly deceives the tribunal about what the law is and who said it. As the Geddes court found, where a lawyer deliberately uses quotation marks — indicating an intention to represent the text as directly quoted — the falsity cannot be characterised as innocent mistake.

The practical implication for every advocate who uses AI in legal drafting is stark: never put quotation marks around text you received from an AI tool unless you have verified every word of that text against an authenticated primary source. The verification must be character-level: even a single word inserted or changed in an otherwise accurate quotation is a misquotation. Legal quotations that are being presented to courts as direct quotes must be copy-pasted from verified, authenticated sources — not reproduced from AI output.

The referral to the California State Bar adds a further dimension. An AI-fabricated quotation sanction is not simply a case-level consequence. It initiates a professional disciplinary process that can result in suspension or disbarment. For Indian advocates, the parallel trajectory — from court sanction to Bar Council of India disciplinary proceedings — is equally real. A single careless AI-generated quotation can threaten a lawyer’s entire career.