AI x Crypto: How to Tell Real Usage from a Narrative

AI x Crypto should be evaluated through verifiable usage, not through words like “agents,” “neural networks,” or “automation.” The useful questions are: who pays, which actions repeat, what data the system needs, and why a token is required. If a project shows only announcements, it is more narrative than product.

What counts as real usage?#

Real usage is a repeated action that remains useful without market hype. Examples include an agent paying for a service, a model accessing data, an app coordinating compute, users paying for output, and transactions that can be verified.

Visa’s work on agentic payments shows why machine payments need rules, limits, and controls. For crypto, this matters even more because many transactions are irreversible.

Which metrics separate product from presentation?#

Look at active users, recurring operations, revenue or fees, unique payers, retention, partner quality, and the share of activity that does not look like incentive farming.

On-chain metrics are useful, but they can be inflated. If a project claims millions of transactions, check who made them, how many addresses are unique, whether there is economic value, and whether activity is subsidized by the project itself.

How can token utility be checked?#

A token is useful when it solves a real coordination problem: paying for a resource, securing a network, distributing access, or aligning incentives. If the same product could work more cheaply without a token, users should understand why the token exists.

A simple question helps: what breaks if the token is removed? If the answer is mainly “market attention,” the reason is weak.

What risks come with AI payments and data?#

AI systems involve data, permissions, and automated actions. Risks include privacy leakage, incorrect permissions, agent error, malicious instructions, external API dependency, and unexpected spending.

The IMF’s Fintech Notes on AI and finance highlight the need for risk management and oversight in automated financial systems. That concern is especially relevant when actions can scale quickly through APIs and wallets.

What are the checklist’s limits?#

The checklist does not guarantee a project will succeed. Even a real product can lose to competitors, face regulation, lose data access, or fail to maintain token economics.

But it removes the weakest layer: projects with AI slogans, a token, and a roadmap, but no verifiable users. For lasting interest, that is not enough.