Inflated ARR at AI startups: how to really evaluate them

The context: why AI startups' ARR metrics are under scrutiny

In May 2026, TechCrunch published an in-depth analysis of a practice that is now widespread in the AI ecosystem. Some startups inflate their Annual Recurring Revenue metrics to build compelling growth stories. What's more, the venture capitalists who fund them are often aware of the distortion. Yet they keep sharing those numbers publicly.

The phenomenon isn't new in tech. In the age of artificial intelligence, however, it takes on a particular shape. AI startups' business models are often hybrid: they combine software licenses, professional services, API credits and pilot contracts. As a result, the very definition of "recurring revenue" becomes elastic and easy to manipulate.

For Italian SMEs operating in B2B or retail, this scenario has direct implications. It concerns anyone considering adopting AI tools, choosing technology partners or integrating third-party platforms into their workflows.

The numbers that matter: how ARR gets distorted in the AI sector

ARR, or Annual Recurring Revenue, measures annualized revenue from recurring contracts. It's a standard metric in traditional SaaS. In the AI context, however, it gets manipulated in several ways.

The first technique is aggressive annualization. A startup with a monthly contract worth €10,000 reports an ARR of €120,000, even if the contract lasts only three months. The number looks solid, but it reflects no multi-year commitment from the client.

The second technique is including professional services. Consulting, implementation and training get counted as recurring revenue. By definition, they aren't: they end when the project does.

The third distortion involves pilot contracts. Some startups count proof-of-concept projects as ARR, even though they often don't convert into final contracts. The real conversion rate stays hidden from outside observers.

According to a Gartner analysis of AI startup metrics, over 40% of companies in the sector use non-standard ARR definitions in their public communications. Moreover, fewer than 20% clearly specify which revenue components they include in the calculation.

Strategic takeaways: what it means if you're choosing an AI vendor

B2B SMEs often have to evaluate AI solution providers without the analytical tools of venture capital funds. Still, there are signals you can read even without access to audited financial statements.

The first signal is the makeup of the customer base. A startup with 500 paying low-ticket customers is structurally stronger than one with 5 enterprise clients on pilot contracts, even if the reported ARR is identical. So ask how many active customers pay every month, not just the total number.

The second thing to check is Net Revenue Retention (NRR). This metric measures how much existing customers spend over time. An NRR above 100% indicates organic expansion. Conversely, an NRR below 90% signals high churn, regardless of the reported ARR.

The third criterion is the contract structure. Prepaid annual contracts show real commitment from customers. Monthly agreements that can be terminated at any time, on the other hand, significantly weaken the strength of the recurring revenue.

Along these lines, Harvard Business Review has published an analysis on assessing the stability of AI vendors. The article stresses that operational due diligence now matters more than a vendor's media reputation.

The hidden risk for SMEs: dependence on fragile platforms

Adopting an AI tool from a startup with inflated metrics exposes SMEs to real risks. The main one is service discontinuity. If the startup runs out of funds or gets acquired, existing contracts can be terminated with minimal notice.

At SHM Studio, we've seen this scenario in at least three cases over the past eighteen months. Client companies had integrated third-party AI tools into their digital marketing and SEO processes. Later, the vendors shut down or pivoted the product. An urgent migration became necessary, with unplanned costs and disruption.

There is also the risk of technology lock-in. Some AI platforms build deep dependencies into company workflows. Switching vendors then becomes costly, even when the service gets worse. That's why your initial vendor evaluation should also cover how easy it is to exit the contract.

Finally, there's reputational risk. Publicly associating with a startup that later turns out to have inflated its metrics can damage the client company's credibility, especially in B2B contexts where trust is a critical asset.

Operational implications: a checklist for evaluating AI vendors

Based on these points, you can build a structured evaluation process. Below are the main criteria we suggest applying before signing any contract with an AI vendor.

• Request the ARR breakdown: how much comes from prepaid annual contracts, how much from monthly ones, and how much from non-recurring professional services.
• Verify the number of active customers: not just the historical total, but how many are paying today and how often.
• Ask for Net Revenue Retention: a figure above 100% is a positive sign of product strength.
• Analyze the financial runway: how many months of operations the current cash guarantees, regardless of reported revenue.
• Review the exit clauses: notice periods, data portability, API availability if the service shuts down.
• Check customer base concentration: if 50% of ARR comes from one or two customers, the risk of volatility is high.

These criteria apply both to choosing tools for managing AI in your company and to selecting platforms for Google Ads campaigns or LinkedIn campaigns with intelligent automation features.

What nobody says openly: the role of VCs in building the myth

The TechCrunch investigation raises an issue that goes beyond creative accounting. Venture capitalists aren't unwitting victims of these practices. On the contrary, they often actively encourage them.

The mechanism is simple. A fund that has invested in a startup has every incentive to maximize its valuation at exit. So spreading growth narratives built on generous ARR figures helps set the stage for later rounds or acquisitions at high multiples. As a result, the secondary market absorbs valuations that wouldn't survive rigorous scrutiny.

According to McKinsey's State of AI 2025, the pressure to show rapid growth has led many organizations to favor vanity metrics over indicators of real value. On top of that, competition for talent and media visibility amplifies this effect in AI more than in any other tech vertical.

For SMEs, this means operating in a market where public information about vendors is structurally skewed. Independent due diligence is therefore not an optional extra: it's an operational necessity.

The SHM Studio perspective: how we guide our clients' technology choices

SHM Studio supports Italian SMEs in selecting and integrating digital technologies, including AI solutions applied to web development, SEO copywriting and marketing automation. In this context, critical vendor evaluation is an integral part of our consulting process.

In particular, when we assess tools to integrate into our clients' workflows, we systematically apply the criteria described in this article. We also monitor our suppliers' financial stability over time, not just at onboarding. That way, we reduce the risk of having to manage urgent migrations or unplanned service interruptions.

If your company is considering adopting AI tools or wants to review its tech stack, the SHM Studio blog offers regular updates on these dynamics. Our team is also available for a personalized assessment through the contact page.

Finally, if you run digital marketing activities with AI components, we suggest reviewing your contracts with automated platform providers on a regular basis. It also helps to set up internal alerts for when a vendor changes its pricing, modifies its APIs or announces funding rounds with valuations that look out of line with reported revenue. Read together, these signals give you a more reliable picture of an AI startup's real health than any press release.