inflating ARR – A legal AI founder says many startups overstate ARR by mixing it with contracted or projected revenue—distorting benchmarks for investors, journalists, and customers.
Thousands of AI startups are chasing enterprise dollars, and the metric they highlight—ARR—has become a marketing centerpiece. Misryoum spoke with the founder behind a public warning that suggests some companies are inflating that number to look faster and bigger than their “live” revenue.
The concern raised by Scott Stevenson. cofounder and CEO of the legal AI startup Spellbook. centers on how startups define and report annual recurring revenue.. Misryoum’s focus here is less on whether AI companies can grow. and more on how financial storytelling can drift away from what ARR is supposed to measure.
ARR is meant to reflect the annualized value of recurring subscription contracts.. The clean version is straightforward: take what a company invoices over a month. then project that rate over a full year. assuming similar billing continues.. Stevenson argues that the problem starts when startups blur ARR with “contracted annually recurring revenue. ” or CARR—figures that can include future payments tied to contracts that are not fully realized in the current billing cycle.
Misryoum sees why the distinction matters.. ARR is not just a scoreboard; it influences hiring plans. fundraising expectations. and how quickly a startup attracts partners who want proof of repeatable business.. When contracted or potential revenue is treated like invoiced revenue. it can turn “what’s confirmed” into “what might happen. ” and that shift can ripple outward through the entire startup ecosystem.
According to Stevenson, some AI startups in fundraising decks may effectively report CARR as if it were ARR.. In practice. the approach can vary: a startup might count the full value of a long contract even though customers can opt out after a short window; it might treat a free or low-cost pilot as if it already represents paid recurrence; or it might assume revenue during periods when features are still being built. despite no guarantee the customer will start paying once development ends.
Stevenson also points to a pattern emerging after accelerator or early-stage demos—where headline ARR numbers appear impressive. but a deeper look reveals that much of the “recurring” revenue is not yet converted into live. billable subscriptions.. Misryoum interprets this as a gap between fundraising language and operational reality.. For investors. that gap increases diligence costs and raises the risk of backing a growth story that looks stable on paper but isn’t fully monetized.
The issue extends beyond the data room.. Stevenson argues that journalists often lack access to the underlying contracts and therefore may report ARR figures at face value.. That limitation can make the public narrative about AI adoption more bullish than the underlying billing traction suggests.. Misryoum’s view is that this creates a feedback loop: stronger headlines can help startups attract additional attention and capital. which in turn can encourage even more companies to chase the same metric optics.
There are also behavioral consequences inside the ecosystem.. If one company appears to be “winning” based on an inflated ARR definition. competitors may feel pressured to match the visible performance—regardless of whether customers are actually paying for the same level of product usage.. Stevenson describes how these distortions can create a kind of momentum that pushes companies toward risky moves. employees away from clear visibility into real performance. and customers who are trying to compare providers on a level playing field.
From a market perspective, Misryoum sees this as part of a broader credibility challenge for AI businesses.. There is already skepticism about revenue durability across the sector—ranging from large-model spending and data center costs to enterprise applications built on top of those models.. If ARR benchmarks are further inflated. the “bubble” concern becomes harder to test and easier to believe. because everyone is pointing at numbers that may not be fully comparable.
What would accountability look like in practice?. Stevenson suggests that stakeholders—especially those writing or underwriting deals—should ask a sharper question: does the ARR number reflect live invoiced revenue. or is it contracted or projected revenue bundled into the same headline?. Misryoum adds that clarity on what is contracted versus billed can help align expectations across founders. investors. and customers—and may reduce the incentive to treat metric games as a substitute for product-market fit.
As AI startups continue competing for enterprise contracts. the next phase of differentiation may be less about model capability and more about financial transparency.. Investors and observers will likely push harder on “quality of revenue. ” and companies that can demonstrate stable. invoice-based recurring streams may find that credibility becomes its own growth advantage.