The AI Washing Index

In Q4 2025, 331 S&P 500 companies cited "AI" on their earnings calls — the highest number in the past decade, and more than triple the ten-year average of 94. For most of 2025, saying "AI" on earnings calls correlated with stock outperformance — in Q3, AI-citing companies averaged 13.9% gains versus 5.7% for non-citers. But by Q4, the premium had reversed: AI-citers averaged just +1.5% versus +5.6% for companies that didn't mention AI.¹

The market rewarded the word "AI" for most of 2025. By Q4, it stopped. The premium evaporated — possibly because investors started asking what the word actually meant. But 331 companies are still saying it.

The SEC has noticed.

The Fines

The SEC's crackdown on AI washing began in March 2024 with simultaneous enforcement actions against two investment advisory firms — Delphia and Global Predictions — for making false and misleading statements about their use of AI. Delphia paid $225,000. Global Predictions paid $175,000. The amounts were small. The message was not.²

In February 2024, the SEC settled a fraud case against Rockwell Capital Management and its founder Brian Sewell for raising $1.2 million for a cryptocurrency fund purportedly powered by AI and machine learning. The AI never existed. Sewell agreed to pay $1.6 million in disgorgement plus a $223,000 civil penalty.²

In January 2025, the SEC charged Presto Automation — the first AI washing action against a public company. Presto marketed "AI speech recognition" for restaurant drive-throughs. The reality: the technology was owned and operated by a third party and relied on significant human intervention. What was sold as automated AI was humans behind a curtain.³

In April 2025, the SEC and DOJ jointly charged the founder of Nate, Inc. He raised $42 million claiming his shopping app used AI to process transactions. The actual automation rate: essentially zero. Manual workers in foreign countries completed the purchases. He told investors the rate was above 90%.³

Four enforcement actions. Four cases of companies claiming AI capabilities that didn't exist. $42 million raised on a lie about a shopping app.

The Lawsuits

The SEC's enforcement is just the beginning. Private lawsuits are proliferating.

Apple delayed its Siri AI upgrades from 2025 to 2026. By mid-2025, it became clear the company's AI progress was more modest than advertised. Apple's stock lost nearly a quarter of its value — approximately $900 billion in market capitalization. A lawsuit now alleges that Apple's AI claims amounted to securities fraud.⁴

DocGo, Inc. misled investors about its "proprietary central AI system." A March 2025 ruling found the company had misrepresented the technology. The CEO had also falsely claimed a master's degree in computational learning theory — a credential tied directly to his purported AI expertise.⁴

C3.ai faced a class action in August 2025 alleging it misled investors about AI adoption and product performance. Elastic N.V. was sued in early 2025 after executives allegedly overstated AI integration in its products.⁴

The pattern: company claims AI on earnings calls. Stock rises on the claim. AI doesn't materialize. Stock falls. Lawsuit follows. The 8.2-point premium becomes an 8.2-point liability.

The Trajectory

FactSet has tracked AI mentions on S&P 500 earnings calls for a decade. The trajectory is vertical.¹

Ten-year average (2015-2024): 86 companies per quarter citing AI. Five-year average: 149. Then ChatGPT launched. Q2 2023: 180 companies. Q1 through Q3 2024: roughly 210-214 — a plateau. Then the second wave. Q3 2025: 306. Q4 2025: 331.¹

In two years, the number of S&P 500 companies invoking AI on earnings calls tripled. Not the number of companies shipping AI products. Not the number of companies with measurable AI revenue. The number of companies saying the word.

The question the SEC is asking, and that investors should be asking, is how many of the 331 can back it up.

The Gap

We know the answer isn't 331. McKinsey's global AI survey found that 88% of organizations use AI in at least one function, but only 39% see any EBIT impact.⁵ S&P Global Market Intelligence found that 42% of companies abandoned most AI initiatives in 2025.⁵ As we documented in The 14% Problem, only 14% of workers using AI report consistently positive outcomes.

Eighty-eight percent adopted. Thirty-nine percent saw results. Fourteen percent of workers say it's working. And 331 companies told investors it was part of their story.

The gap between "we mention AI" and "AI makes us money" is where the washing happens. And the SEC's four enforcement actions, against a backdrop of 331 companies claiming AI, represent a coverage rate of approximately 1.2%.

The Enforcement Problem

The SEC has made AI washing an "immediate priority." Senior officials from the Enforcement Division and the newly constituted Cybersecurity and Emerging Technologies Unit (CETU) have said explicitly that "rooting out" AI fraud is at the top of their agenda.⁶

But enforcement faces a definitional problem. Nate, Inc. was easy — automation rate zero, claimed 90%. Presto was easy — humans, not AI. What about a company that has an AI product that sort of works, claims it's "transforming" the business, and cites it on every earnings call while the actual revenue impact is negligible? Where is the line between optimism and fraud? Between aspirational marketing and material misrepresentation?

The SEC's early cases targeted obvious fabrication. The harder cases — the ones that will define the field — involve companies that have real AI teams, real AI budgets, and real AI products that simply don't do what the earnings call suggests. Those cases are coming. The legal framework for them is still being written.

The New York State Bar Association has published on the need for aggressive enforcement. Holland & Knight documented the SEC and DOJ "warming up" to broader enforcement. The pipeline of cases is growing.⁶⁷

The Index

What would an honest AI Washing Index look like?

Start with the 331 companies that cited AI on Q4 2025 earnings calls. Cross-reference with disclosed AI-specific revenue — not "AI-influenced" revenue, not "AI-enhanced" products, but revenue attributable to AI capabilities that didn't exist two years ago. The number that survives this filter would be the real AI index.

Nobody has built this index. FactSet counts mentions. The SEC investigates fraud. Nobody is systematically measuring the gap between what companies claim about AI and what AI actually contributes to their business.

Until someone does, the 8.2-point stock premium for saying "AI" will continue. Companies will keep saying it. Investors will keep rewarding it. And the SEC will keep fining the most egregious offenders while the systemic overstatement continues unchecked.

331 companies said the word. Four got fined. The other 327 are still talking.