Merchant of Record for AI Companies: What Founders Need to Know

AI products broke the billing playbook. The old SaaS model assumed a tidy mapping — one plan, one monthly charge, one predictable invoice. AI workloads don't behave that way. A single customer might burn through millions of tokens one week and almost nothing the next, run agent loops that spawn thousands of API calls, and expect to pay only for what they actually used. On top of that, you're selling globally from day one, which means tax and compliance in dozens of jurisdictions before you've even nailed your pricing.

That combination — usage-based billing plus global compliance — is exactly where a merchant of record (MoR) earns its place in an AI company's stack. This guide explains what AI founders need to know: why AI billing is different, how usage-based billing actually works, where an MoR fits, and how to tell a true MoR from a billing tool that just looks like one.

New to the model itself? Start with what a merchant of record is.

Why billing for AI companies is different

Traditional billing tools were built for flat subscriptions. AI introduces problems they were never designed for:

Consumption is variable and unpredictable. You're billing tokens, API calls, compute seconds, generations, or agent actions — not a fixed seat count.
One request isn't one billable event. An AI workload can fan out into many metered events, with deduplication, late events, and corrections all in play.
Credits are a liability, not just a payment method. Prepaid credits sit on your books until consumed, can roll over or expire, and must be reconciled against real usage. "Credit management" bolted onto a subscription tool is often four spreadsheets pretending to be a ledger.
Costs can run away. Without cost containment — credit reservations, circuit breakers for runaway agent loops, anomaly detection — a single customer's usage can spike before anyone notices.
Global tax lands immediately. AI products sell worldwide instantly, so VAT, GST, and sales tax obligations show up far earlier than they would for a local business.

Get the metering wrong and you either under-bill (lost revenue) or over-bill (churn and disputes). Get the tax wrong and you inherit a compliance mess across every market you've touched.

Usage-based billing, briefly

Usage-based (or metered) billing charges customers in proportion to what they consume rather than a flat fee. Under the hood it's a three-stage pipeline, and each stage has to be independently correct:

Ingest — usage events flow through a pipeline that validates, buffers, and durably stores them. It has to absorb traffic spikes without dropping records.
Meter — raw events are aggregated into billable quantities for the billing period, applying your unit definitions and pricing logic consistently.
Invoice — metered totals become invoice line items, with credits and discounts applied, and the customer is charged.

Determinism and idempotency aren't optional here. If ingestion double-counts or a late event isn't handled, you don't find out until a revenue discrepancy surfaces weeks later. Common billing units in AI: tokens (often priced differently for input vs. output), API calls, compute/GPU seconds, generations (per image/video/audio), credits, seats, and increasingly outcome- or agent-action-based pricing. Most AI companies land on a hybrid model — a base subscription plus usage overage.

Where a merchant of record fits

A billing engine can meter and invoice. It can't, on its own, take the legal and tax liability off your books. That's the MoR's job. When you sell through a merchant of record, the MoR becomes the legal seller: it processes the payment, collects and remits VAT/GST/sales tax in each jurisdiction, absorbs chargeback and fraud liability, and handles PCI compliance. You keep building the product.

For AI founders specifically, an MoR delivers a few things that matter more than they first appear:

Global tax handled from day one — no registering for VAT across the EU while you're trying to ship features. Your job is the model, not EU tax returns.
Payment credibility and risk models — an established MoR brings credibility with banks and acquirers, plus fraud/chargeback models tuned for the unusual risk profile of AI usage.
One relationship instead of a stack — checkout, metering, tax, and payments in one place rather than stitched-together tools.

The critical distinction: MoR vs. "AI billing tool"

This is the single most important thing for a founder to understand, because the market blurs it. Many tools marketed for "AI billing" are metering and billing engines, not merchants of record — they aggregate usage and generate invoices, but require you to plug in a separate payment processor and still leave tax and compliance liability with you.

A rough map of the landscape:

Billing/metering engines (not MoRs): great at usage metering and pricing logic, but you bring your own processor and own the tax/compliance. Some are tied to a single payment ecosystem, which can mean lock-in.
Merchants of record: handle payments, tax, and liability as the legal seller — and the better ones also do the usage metering, entitlements, and checkout so you don't need a separate billing layer.

When you evaluate a platform, the questions that cut through the marketing:

Is it a true MoR (legal seller, remits tax, absorbs chargebacks), or a billing tool that needs a processor bolted on?
Does it handle usage-based metering natively, or only flat/seat plans?
Can it manage entitlements and usage caps (gate features, enforce credit limits) reliably when billing changes?
Are you locked into one payment ecosystem, or is it flexible?
Does it actively recover revenue — dunning, smart retries, checkout optimization — or just process what comes through?

(For the underlying difference between a true MoR and a processor, see Merchant of Record vs. Payment Service Provider — Note that standard Stripe is a PSP, not an MoR.)

What to prioritize when choosing

For an AI company, weigh these in roughly this order:

Accurate, real-time usage metering for your billing units (tokens, calls, compute, credits).
True MoR coverage so global tax and liability are off your books.
Pricing flexibility — hybrid, tiered, volume, credits, overages — that you can change without an engineering project every time.
Entitlements and cost containment — feature gating, usage caps, and protection against runaway usage.
Revenue recovery — failed payments and churn quietly cost more than transaction fees, so dunning and checkout optimization matter.
Speed and support — how fast you launch, and whether the provider's team handles billing issues so yours doesn't.

This is the gap Comecero is built for: merchant-of-record coverage and billing infrastructure aimed at SaaS and AI sellers, with a focus on recovering the revenue most platforms leak — checkout optimization, dunning, and pricing-workflow automation — without a heavy setup.

The bottom line

AI billing is a metering problem and a compliance problem at the same time. You need usage-based billing accurate enough to trust your invoices, and a merchant of record that takes global tax and liability off your plate — ideally in one system, without locking you into a single ecosystem, and with real revenue recovery built in. Sort out which tools are true MoRs versus billing engines early, and you'll avoid the most expensive mistakes.

If you're building an AI product and want MoR coverage plus billing that's built to recover leaked revenue, talk to the team at Comecero. We build merchant-of-record and billing infrastructure for SaaS and AI companies — without the complex setup.