The co-founder of Cluely, an Andreessen Horowitz-backed startup whose motto is "Cheat on everything," admitted this month to lying to a reporter about his company's annual recurring revenue — a confession that has cracked open a wider debate about the integrity of Silicon Valley's most cited AI performance metric.

Annual recurring revenue, or ARR, has long served as the headline number investors, journalists, and founders use to gauge a software company's health. It was designed for subscription businesses with predictable, contracted income. But as AI startups proliferate and compete for attention and capital, the metric is being stretched, bent, and in at least one documented case, fabricated.

The metric was designed for predictable, contracted income — not the volatile, consumption-based reality of most AI businesses.

Why ARR No Longer Means What It Used to Mean

Traditionally, ARR represents the annualised value of a company's recurring subscription contracts — a reliable proxy for business stability. A company with $10 million ARR in the classic SaaS sense has $10 million worth of signed, renewing contracts providing predictable cash flow.

AI-era startups, however, frequently operate on consumption-based or usage-based pricing models, where revenue fluctuates month to month depending on how much customers actually use the product. Annualising a single strong month — multiplying it by twelve — produces an ARR figure that may never be repeated, let alone sustained. According to multiple investors and analysts cited by Bloomberg Technology, this practice is widespread and rarely disclosed.

The Cluely incident makes this implicit problem explicit. The startup's co-founder acknowledged misrepresenting the figure to a journalist, apparently treating ARR as a marketing tool rather than a financial disclosure. The company's motto — "Cheat on everything" — was intended as provocateur branding, but the admission gave that tagline an uncomfortably literal dimension.

A Metric Under Pressure Across the Industry

Cluely is unlikely to be alone. The incentive structure surrounding ARR in private markets creates systematic pressure to inflate or misrepresent the figure. Private AI startups face no regulatory obligation to report standardised financials. Investors negotiating term sheets, journalists writing profiles, and talent evaluating job offers often have no independent means of verifying the numbers they are given.

Andreessen Horowitz, one of Silicon Valley's most prominent venture firms, backed Cluely — a reminder that prominent investment does not guarantee the integrity of a portfolio company's self-reported metrics. The firm has not commented publicly on the ARR misrepresentation.

The broader pattern matters for anyone allocating capital or covering the sector. A 2023 analysis by Bessemer Venture Partners of cloud company benchmarks found that consumption-based businesses show revenue volatility up to three times higher than traditional subscription SaaS firms — making point-in-time ARR snapshots particularly unreliable as forward indicators.

The Human Cost of Misleading Numbers

The distortion is not purely academic. Employees join startups partly based on growth metrics that signal company health and job security. Candidates evaluating offers from AI startups frequently encounter ARR figures in pitch decks and press coverage, with no way to assess whether the number reflects contracted revenue, an annualised single-month spike, or something more creative still.

Enterprise customers signing multi-year agreements also factor a vendor's apparent financial health into procurement decisions. A customer choosing an AI infrastructure provider based on inflated ARR signals faces real operational risk if that provider's actual revenue base is a fraction of what was implied.

Small investors in the public markets face a version of the same problem when AI companies use ARR-adjacent language in roadshow materials before an IPO. The Securities and Exchange Commission has previously flagged non-standard revenue metrics as a disclosure concern, but enforcement in private markets remains limited.

What Comes After the Confession

The Cluely incident may accelerate calls for industry-wide standardisation of how AI companies define and report ARR. Some investors are already pushing portfolio companies to distinguish clearly between contracted ARR — revenue locked in by signed agreements — and run-rate ARR, the annualised projection of recent consumption. The two numbers can differ by an order of magnitude.

Journalists covering the sector face a parallel reckoning. ARR figures sourced directly from founders or their PR teams, without independent verification or definitional clarity, risk functioning as amplified press releases rather than meaningful data points. Bloomberg's reporting on the Cluely case is itself a corrective signal — but one data point does not reform an industry norm.

Regulators have shown increasing interest in AI company disclosures, particularly as more of these startups approach public markets or seek government contracts. The Federal Trade Commission and SEC have both signalled scrutiny of AI-related marketing claims, though neither has yet moved specifically on ARR methodology.

What This Means

For anyone relying on ARR to evaluate an AI company — as an investor, employee, customer, or reader — the Cluely admission is a practical warning: always ask how the number was calculated, and treat unverified ARR from private AI startups as a starting point for questions, not a statement of fact.