When AI Becomes the Auditor: The Carvana Case and the Future of Financial Fraud Detection
A Conversation That Changed My Perspective
Two days ago, on January 28, I had a fascinating discussion with an attorney who co-founded a specialized firm focused on deception identification and behavioral analysis. This organization, established in 2009 by experienced FBI agents and business leaders, helps clients whose outcomes depend on accurate analysis of verbal and written information.
Our conversation turned to an unexpected development in financial markets: AI systems are now analyzing SEC filings and corporate disclosures—and they’re finding troubling patterns that traditional analysis might miss. According to this securities law expert, AI-assisted analysis is sometimes revealing that corporate statements are, to paraphrase, “dubious at best and sometimes untruthful.”
Based on recent investigations, legal actions, and reports, Lululemon comes to mind as a prominent example of a company accused of using “puffery” and "greenwashing" for image management rather than transparent environmental action, but I had not considered the fraud topic to be worthy of investigating as a financial journalist. Until now.
The timing of our conversation was remarkable. That same day this associate and I were speaking, Carvana (NYSE: CVNA) became what may be a landmark case study in AI-assisted fraud detection.
The Carvana Earthquake: When Numbers Don’t Add Up
On a January day that should have been routine, Carvana’s stock plummeted 14% to around $408, wiping out a significant portion of its one-year gains immediately after its corporate filing was publicly reported. The catalyst wasn’t poor earnings—in fact, Carvana had just released record Q3 financials with revenue of $5.65 billion (up 55% year-over-year) and adjusted EBITDA of $637 million.
The problem was something far more serious: Gotham City Research published a report alleging systematic accounting irregularities involving over $1 billion in overstated earnings.
For those who don’t know, Gotham City Research is a controversial financial research and investment firm, founded by Daniel Yu, operating from a "secret location in Manhattan," that specializes in forensic accounting and short-selling.
The Allegations: A Web of Related-Party Transactions
Gotham City Research’s claims center on complex relationships between Carvana and entities controlled by the Garcia family (Ernest Garcia II and III):
The GoFi Mystery: A little-known entity called GoFi, LLC allegedly exists primarily to shuttle loans between related parties, deriving essentially 100% of its $7.1 million in 2024 revenue from selling finance receivables.
The DriveTime Question: Rather than being the financial backstop Carvana suggests, DriveTime allegedly burned through more than $900 million in operating cash from 2022-2024 and had to raise debt to fund itself.
The Bridgecrest Profit Shuffle: The most serious allegation involves Carvana selling loans to related-party Bridgecrest at allegedly inflated values, booking immediate profits. Bridgecrest then marks down these same loans by up to 15%—about $900 million in 2024—effectively moving losses off Carvana’s books.
The Audit Concern: Adding to the complexity, the same auditor (Grant Thornton) audits all three entities—Carvana, DriveTime, and GoFi.
The SEC issued a subpoena in June 2025 regarding these related-party transactions.
How AI Enters the Picture
This is where the story becomes particularly relevant to my conversation about AI and financial analysis. The patterns Gotham City Research identified—the circular flow of receivables, the timing of markdowns, the concentration of revenue in shell entities—are exactly the kind of complex, multi-entity relationships that AI excels at mapping.
Traditional analysis might review each entity’s financials separately. AI can simultaneously process:
Transaction flows across multiple related entities
Timing patterns in asset transfers and markdowns
Discrepancies between stated asset values and subsequent write-downs
Revenue concentration patterns that suggest entities exist primarily for accounting purposes
Historical patterns in similar alleged schemes
The sophistication required to track loans being sold at one value, then marked down by a related party, across multiple quarters and multiple entities, is substantial. This is computational detective work—exactly what AI systems are increasingly capable of performing.
The Market’s Split Decision
Institutional investors appear divided. Despite the accounting allegations:
JPMorgan maintains an Overweight rating with a $510 price target
Wells Fargo set a $525 price target
The underlying business metrics remain strong: $2.1 billion in cash, $1.2 billion in debt reduction, and operational improvements including same/next-day delivery for 40% of sales
Yet the stock trades around $414—below both recent highs of $486 and analyst targets above $500.
What This Means for Investors
The Carvana situation illustrates a new dynamic in financial markets. We’re entering an era where:
AI as Watchdog: Sophisticated analysis tools can now process the complex web of related-party transactions, timing patterns, and accounting flows that might indicate irregularities.
Speed vs. Depth: Markets receive allegations faster than they can be fully investigated. The SEC subpoena confirms regulatory scrutiny, but resolution will take time.
The Valuation Paradox: Strong operational performance (55% revenue growth, improved EBITDA margins) can coexist with serious questions about how that performance is being accounted for.
Investment Implications
For risk-tolerant investors, Carvana presents an unusually binary outcome:
If the accounting clears: The stock could quickly recover toward analyst targets above $500, representing 20%+ upside.
If irregularities are confirmed: Substantial downside remains, with potential delisting risk depending on severity.
At approximately 90 times trailing earnings and with unresolved SEC scrutiny, this is emphatically not a position for conservative fiduciary-managed portfolios.
The Broader Message
The more important story transcends Carvana. My conversation with the deception analysis expert highlighted something profound: AI is becoming an increasingly powerful tool for analyzing corporate disclosures, and the results are sometimes revealing uncomfortable truths.
As these AI systems become more sophisticated and widely deployed, we may see:
Earlier detection of accounting irregularities
More frequent challenges to complex related-party arrangements
Increased scrutiny of companies with intricate corporate structures
A new arms race between creative accounting and AI-assisted detection
For investors with five decades of market experience like myself, this represents a genuinely new development. The tools for analyzing corporate truthfulness have fundamentally changed.
Conclusion
Whether Carvana ultimately proves to be a case of aggressive but legal accounting or something more serious remains to be determined. The SEC investigation will provide answers—eventually.
But the broader lesson is clear: AI is changing how markets police themselves. The complex web of transactions that might have gone unnoticed for years can now be mapped, analyzed, and questioned in real time.
For investors, this means heightened vigilance around companies with complex related-party structures—and recognition that the tools for uncovering deception are evolving faster than the methods for creating it.
Disclosure: This analysis is for informational purposes only and should not be considered investment advice. The author writes the Weekly Global Market Navigator and other investment publications. Investors should conduct their own due diligence and consult with financial advisors before making investment decisions.