Appellate Ethics & AI Fla. 4th DCA · No. 4D2026-2262 September 2026 Briefing

“AI Slop” Is Not a Legal Standard:
When Courts Infer Sanctions from Prose Style

In Lisandrillo v. Palozzi, a Florida appellate court ordered counsel to show cause—without using AI—why she shouldn’t be sanctioned after labeling her filings “AI slop.” Her 84-page response pushes back hard, setting up the most critical professional judgment test of the generative AI era.

Part I. The Charge, the Counts, and the Standard Response of Jaclyn R. Soroka, Esq. (Sept. 28, 2026)
“Suspicion can be formed from the way something looks; authenticity can only be taken from the source. In this Court the source is the record, and the authenticity of my filings is established by it, not by suspicion. The opinion was written, by its own account, solely to address my filing as it appears to be 'AI slop.' (Op. at 1.) The operative word, appears.”

The Core Tension: Earlier AI disciplinary decisions penalized lawyers for hallucinated case law—fake citations to nonexistent opinions. In Lisandrillo, every cited authority exists and the cases are real. The court’s sanction threat rests on a fundamentally different premise: that verbose, repetitive, or unfocused prose creates an inference that an advocate failed to exercise independent judgment.

84 Pages Filed Drafted by counsel without AI per court order
237 Citations Audited Line-by-line inspection across appellate and trial files
0 Fabricated Cases Zero nonexistent authorities or hallucinated holdings
2 Human Errors Conceded One missing rule subdivision & one typo ("chill" for "child")
The Structural Clash

The Judicial Order vs. The Practitioner’s Pushback

The Judicial Charge

The "AI Slop" Inference

The Fourth District Court of Appeal denied the petition and issued an Order to Show Cause why sanctions should not be imposed, explicitly requiring a response prepared without the use of artificial intelligence.

  • Stylistic Critique: Labeled filings lengthy, unfocused, and containing gibberish.
  • Presumed Delegation: Suspected reasoning was "assembled by an AI tool rather than developed through independent advocate judgment."
  • No Hallucinations: Expressly acknowledged the problem had nothing to do with fake cases or fake citations.
Counsel’s Pushback

“AI Slop” Is Not a Standard

Counsel answered with an exhaustive 84-page response, conducting a full citation audit and mounting a rigorous legal defense of independent professional judgment under Florida ethics rules.

  • No Legal Definition: "AI slop" appears nowhere in any Florida statute, rule of procedure, or ethics decision.
  • Delegation vs. Assistance: Fla. Bar Ethics Op. 24-1 permits AI assistance so long as the lawyer directs strategy and reviews the work.
  • Human Errors != Sanctionable Deceit: A 237-citation audit revealed 2 minor human drafting errors, but zero false law.
Procedural Evolution

The Shift: From Hallucinated Citations to Stylistic Policing

Until now, legal AI sanctions jurisprudence followed an unambiguous line: fabricating cases, inventing judicial quotes, or citing phantom statutes violated Rule 11, Rule 2.515, and Rule 3.3 (Candor to the Tribunal).

From Mata v. Avianca in the Southern District of New York to Florida’s own recent disciplinary matters in Eclectic Synergy v. Seredin and Gouldy v. Chiasson, every reported sanction involved lawyers submitting non-existent legal authorities produced by hallucinating large language models.

Lisandrillo v. Palozzi breaks entirely new ground. The Fourth District Court of Appeal specifically noted in its footnote 1 that the issue:

“...has nothing to do with hallucinated law or false legal citations. Counsel’s citations exist and the cases are real. The problem is that the filings appear to be AI slop—lengthy, unfocused, and filled with gibberish and convoluted arguments.” — Fourth District Court of Appeal, Order to Show Cause (Sept. 16, 2026)

By moving from objective falsity (did the case exist?) to subjective stylistic appraisal (does this writing feel like AI?), the court opened a Pandora's box for appellate advocacy.

Forensic Record Audit

The 237-Citation Forensic Audit

In her response, counsel took on the court's charge by performing a forensic line-by-line audit of every filing submitted across both the Fourth District and the Seventeenth Judicial Circuit (Broward County) trial court:

  • 71 Cases, 9 Statutes, and 16 Rules cited in the appellate petition, emergency application, and appendix.
  • 45 Cases, 6 Statutes, and 7 Rules cited in the underlying trial court Omnibus Application and Motion to Disqualify Judge.
  • Approximately 237 individual citation instances, audited pin-cite by pin-cite.

What did the forensic audit find? Every authority exists.

Counsel conceded two errors: an outdated rule subdivision reference in a citation to Rule 2.515, and an inadvertent transcription typo in paragraph 109 of her trial court motion where the word "child-related" was mistakenly printed as "chill-related."

As counsel pointed out, these are classic human drafting errors that have plagued legal practice for centuries. Conflating human typo error with algorithmic deceit treats ordinary advocacy imperfections as evidence of bad-faith automation.

Professional Implications

The Core Question: Can Courts Infer Sanctionable Conduct from Prose Style?

The crux of the Lisandrillo pushback goes to the heart of modern litigation:

The question is no longer merely whether lawyers can use AI. It is whether a court can infer that an attorney failed to exercise independent professional judgment simply because the work product "looks like AI"—even where the cases are real, the record is grounded, and the lawyer stands behind every word.

Under Florida Bar Ethics Opinion 24-1, lawyers are explicitly permitted to use generative AI tools to conduct legal research, organize thoughts, and draft pleadings, provided they review the output, verify the citations, and maintain non-delegable supervisory responsibility.

If judicial panels are empowered to label filings "AI slop" based on tone, word choice, or repetition, litigators face an unmanageable subjective standard. Dense brief writing, complex statutory recitations, and boilerplate contractual analysis all bear hallmarks of what skeptics call "AI-esque" prose.

Sanctions cannot be grounded in aesthetic displeasure. They must be tethered to the record, to truth, and to verifiable legal authorities.

Litigator Defense Playbook

Guarding Against the "Looks Like AI" Trap: A 4-Step Protocol

As judicial scrutiny intensifies, trial lawyers and appellate counsel must implement concrete safeguards to ensure filings are unassailable on both facts and appearance:

Protocol 1: Record Provenance
Pin-Cite the Record

Anchor every factual statement directly to transcript page and line numbers (e.g. Tr. Vol. II, at 47:12-18) and appendix pages (App. at A30). Generic factual narrative invites judicial suspicion.

Protocol 2: Authority Verification
Zero Trust Verification

Never rely on an AI assistant’s assurance that a quote is verbatim. Every judicial quotation must be visually cross-referenced against the primary court reporter before filing.

Protocol 3: Stylistic Pruning
Eliminate AI Prose Clichés

Prune wordy introductory clauses, generic synthetic transitions ("Furthermore," "Moreover," "It is axiomatic that"), and repetitive restatements that trigger algorithmic suspicion.

Protocol 4: Audit Trail
Document Independent Judgment

Maintain clear internal version history proving that the legal strategy, selection of authorities, and final editorial decisions originated with and were reviewed by licensed counsel.