Scott Stevenson, CEO of legal AI startup Spellbook, lit a fuse last month when he called AI revenue reporting a “huge scam” on X. According to TechCrunch AI, his post drew over 200 reshares and forced a public reckoning across the AI startup community about how annual recurring revenue gets calculated, inflated, and waved around for PR coverage. TechCrunch AI spoke with more than a dozen founders, investors, and finance professionals, and the answer was uncomfortable: the inflation is widespread, and many investors know about it.
Here’s what stands out. The metric founders trumpet as “ARR” often isn’t ARR at all. It’s CARR, or contracted ARR, which counts signed deals before customers are onboarded, before product is deployed, and sometimes before a single dollar has been collected.
How the Numbers Get Stretched
The playbook is straightforward, and that’s what makes it spread. One investor told TechCrunch AI that some companies report CARR figures 70% higher than actual ARR. Tactics include:
- Counting yearlong free pilots as recurring revenue
- Including three-year contracts with steep year-one discounts and assuming customers stick around for full-price year three
- Reporting signed deals that haven’t been implemented (and might never be)
- Ignoring expected churn and downsell, which Bessemer Venture Partners explicitly warned against back in 2021
One former employee told TechCrunch AI their company booked a substantial yearlong free pilot as ARR, with the board (including a VC from a large fund) fully aware. Several investors said they personally know of a high-profile enterprise startup that publicly crossed $100 million in ARR when only a fraction came from paying customers.
Why This Matters Now
AI is the fastest-funded category in tech history. Valuations are getting set off these revenue claims. When a startup announces it “hit $50M ARR in 8 months,” the press writes it up, the next funding round prices off it, and competitors feel pressure to match the narrative.
As one investor put it to TechCrunch AI: “When one startup does it in a category, it is hard not to do it yourself just to keep up.” That’s the contagion mechanic. Honest founders look like they’re losing.
The second twist is the acronym itself. Some founders use “ARR” to mean annualized run-rate revenue, which takes a single strong month and multiplies by 12. That’s a different metric entirely, and lumping it under the same three letters muddies what investors and reporters are actually being told.
What Practitioners Should Do
For founders: pick a metric, define it in your pitch deck and your press releases, and stick to it. If you’re reporting CARR, call it CARR. The companies that survive the next correction will be the ones investors trust on the numbers.
For investors: ask for the breakdown. Live ARR vs contracted vs pilots vs annualized run rate. If a founder bristles at the question, that’s the signal.
For journalists covering AI: stop treating “ARR” as a clean number. Ask which definition. Ask how much is paying, deployed, and live today. The discipline of one extra question changes what gets printed.
For operators benchmarking themselves against headlines: assume the public number is 30 to 70% above reality. Plan accordingly.
The Bigger Picture
This isn’t only an AI story. It’s what happens when capital floods a category faster than accounting norms can keep up. Garry Tan at Y Combinator has been publishing explainers on proper revenue metrics, and Clio’s Jack Newton told TechCrunch AI the attention is overdue. Expect more LPs to start pressing GPs on diligence standards, and expect a few embarrassing markdowns when contracted revenue fails to convert.
The AI boom is real. The revenue figures aren’t always. The gap between those two facts is where the next round of disappointments will land. Full reporting at the original source.