Which merchant of record should you trust? I compared 9

For a first paid software product, my default merchant of record is Paddle. It becomes the legal seller, taking tax, refunds, chargebacks, and fraud off your desk. I compared nine providers because the fee is only part of the decision. Seller reviews, account holds, and who can sign up matter just as much.

What does a merchant of record actually take over?

A merchant of record, usually shortened to MoR, is a company that legally sells your product on your behalf. The customer buys from the MoR, so the MoR owes the VAT and sales tax, answers the chargebacks, screens the fraud, and fields the “where is my refund?” emails. You deliver the product and collect a payout.

Stripe’s own explainer puts the definition this way: the merchant of record is the entity legally responsible for processing customer payments, and with that role come tax collection and remittance, regulatory compliance, fraud prevention, and dispute handling [1]. With a plain payment processor (classic Stripe Payments, Mollie, and the like) the money moves just as smoothly, but every one of those obligations stays with you, including registering for tax in each country where your buyers live.

Who handles it	Plain payment processor	Merchant of record
Legal seller of the purchase	You	The provider
VAT, sales tax, GST worldwide	You register and file	Included
Refunds and chargebacks	You fight each case	Included
Fraud screening	You tune the rules	Included
Payment support requests	You	The provider
Fee per sale	Lower	Higher
Control over your balance	Yours between payouts	Reserves and holds possible

Choosing an MoR means choosing whose risk department you trust with your revenue. That is the practical meaning of the final row in figure 1, and it matters more than the sales pages suggest.

Because the MoR is the legal seller, its name is on the card statement and the chargebacks land on its records. It can demand more documents, hold your money in reserve, refund your customers, or suspend your account during a risk review. Stripe’s Managed Payments terms state that Stripe can issue refunds within 60 days of purchase on its own initiative [2], while Whop’s seller terms allow funds to be held in reserve for up to 180 days [3].

I treat seller reviews as the best public evidence of how those powers are used in practice. But the review platforms come with a trap of their own.

Can you trust merchant-of-record scores on Trustpilot?

Not on their own, because review sites measure different people. Trustpilot mostly reflects end customers who found the merchant of record’s name on a card statement, while sellers who integrate the platform write on G2 and Capterra. At least one provider’s Trustpilot page carried reviews that Trustpilot later removed as fake.

Paddle shows the split clearly. Its Trustpilot profile has collected roughly 11,000 reviews [4], and the mechanics of the MoR model mean most of them come from buyers: whoever saw Paddle on a receipt or a bank statement reviews paddle.com, whether or not they ever heard of the software company that made the product. The sellers who integrate Paddle rate it 4.5 out of 5 on G2 across about 237 reviews [5]. Capterra lands in between, at 3.0 from just 21 reviews with customer service at 2.6 [6], and that tiny sample is worth remembering before treating any single score as precise.

PayPro Global flips the same split the other way. Sellers give it 4.9 out of 5 on G2, from just over 110 reviews, which is the best seller score among the nine [7], while its Trustpilot score sits below 3 [8]. Same company, opposite verdicts, because the two sites poll opposite sides of the checkout.

Whop taught me the third trap personally. I was testing checkout on my own product with small live payments when Whop suspended the account without warning. The balance never came back, and support answered in what read like generated boilerplate. Its Trustpilot page looked spotless at the time, with page after page of five-star reviews from many freshly created profiles using oddly similar wording. I could not prove the reviews were fake, so I simply stopped trusting the score.

Trustpilot later established what I could not. As of 3 August 2026, Whop’s profile shows a warning instead of a rating: the score is unavailable due to a breach of Trustpilot’s guidelines, and Trustpilot states that it removed a number of fake reviews for the company [9]. Roughly 2,900 reviews sit behind that notice.

So this is how I read payment providers now: G2 and Capterra for the seller side, Trustpilot only sorted to newest and read from the one-star end, and always with an eye on whether the company answers its critics with specifics. Those habits produced most of what follows.

What does a merchant of record cost, and can individuals sign up?

Headline rates run from 2.7% + $0.30 at Whop to 5% + $0.50 at Paddle, Polar, and Lemon Squeezy, with PayPro Global and FastSpring quoting per customer instead of publishing a price. Five of the nine state in writing that individuals without a registered company can sign up. Figure 2 holds the published numbers.

Provider	Headline fee	What stacks on top
Paddle	5% + $0.50	Custom pricing for products under $10
PayPro Global	Quote only	Priced per business
FastSpring	Quote only	Volume-based, by agreement
Stripe Managed Payments	3.5% + Stripe processing	Payment-method and currency fees vary
Creem	3.9% + $0.40	Bank payouts cost $7 or 1%
Polar	5% + $0.50	+1.5% international cards, cheaper paid tiers
Lemon Squeezy	5% + $0.50	+1.5% international, +0.5% subscriptions, +1.5% PayPal
Dodo Payments	4% + $0.40 (US cards)	+1.5% international, +0.5% subscriptions, +3% PayPal or BNPL
Whop	2.7% + $0.30	+1.5% international, per-payout fees

The cheap-looking headline is rarely the real price. Whop’s 2.7% is the lowest base rate but sits on top of per-payout fees and the reserve policy covered below. Lemon Squeezy’s 5% becomes roughly 7% for an international subscription paid through PayPal once the three surcharges stack.

Fixed fees also punish small charges. At Creem’s 3.9% + $0.40, a $2 sale loses about 24% to fees while a $20 sale loses about 6%. I calculated those percentages from the published rates, and they are why I would sell an AI product through subscriptions or credit packs instead of single one-dollar top-ups. The costs underneath such a product are metered too, which is why I looked at what OpenAI’s Agents API takes over and which costs stay yours.

Five providers state that individuals can sign up. Paddle skips business verification for individuals and sole traders, with identity checks remaining [21]. PayPro Global says you can accept payments without a registered company [22]. Creem’s terms define a merchant as any natural person or legal entity [23], Polar offers a personal payout account [24], and Dodo lets you verify as an individual creator or sole proprietor [25].

Stripe, by contrast, gates Managed Payments behind an eligibility review and does not state which legal forms pass [2]. FastSpring asks for registration or identification documents and settles the question case by case with a representative [26].

One boundary applies everywhere: the MoR takes over your customers’ taxes, not yours. Income tax and any business registration your country requires once revenue is real remain your job.

Account holds matter more than the headline fee

Fees decide a few percent. Reserve, suspension, and payout policies can decide whether you get paid at all, yet none of them appear in the pricing table.

FastSpring places a 45-day monitoring hold on a new account when live payments start, so your first payout arrives a month and a half after your first sale [27]. That is survivable if you know it in advance, and a nasty surprise if you do not.

Whop’s paperwork is blunter than its marketing. The seller terms allow reserves for up to 180 days and scope the merchant-of-record role to card network rules and payment settlement only [3]. The trust-and-safety docs add that a suspended account’s balance is placed on hold and is not automatically released after any set period [28]. The newest reviews on the flagged Trustpilot profile describe that exact pattern; one from 1 August 2026 reports a balance held for 207 days on an account with a zero dispute rate, a written release date that passed unhonored, and a suspension right after the payout request [9].

Dodo Payments sits at 3.0 on Trustpilot across 122 reviews, and the replies under the reviews say more than the score. Responding to angry sellers, Dodo’s team confirms 120-day holds as standard practice and describes terminating already-approved accounts after later compliance reviews [29]. Credit where it is due: the team answers with specifics and names the policy involved. But approve first, re-review later, hold the money meanwhile is the pattern I got burned by, so I weight it heavily.

Lemon Squeezy is the clearest no. Its Trustpilot score is 1.2 from 163 reviews, with sellers in July 2026 reporting payouts stuck for weeks and one describing more than 100 completed orders refunded back to customers after a ban [30]. The silence has an official explanation. Stripe bought the company in July 2024 [31], and in January 2026 the CEO wrote that the team had been heads-down on Stripe’s new merchant-of-record product, admitted slower support as a consequence, and said the goal is to give Lemon Squeezy users an easy way to migrate to Stripe Managed Payments [32]. However you read that, you should not start a new product on a platform whose own maker is pointing somewhere else.

Creem and Polar I cannot judge this way, because neither has enough public review history for the scores to mean anything in praise or in blame. Treat both as bets on a young company rather than on proven infrastructure, and size the bet accordingly.

My picks: Paddle by default, PayPro Global for support

Paddle runs my products and remains my default recommendation. The 5% + $0.50 fee is public [10], the sandbox has its own API keys, data, webhooks, and test cards [33], and individuals are accepted [21]. Nothing you do there touches real money, so you do not have to test with live charges the way I once did on Whop.

My own support experience has been uneventful, and a human answers when I write in. That is a first-hand impression, not a guarantee: Paddle’s 2.6 customer-service score on Capterra shows that other sellers have had a worse experience [6].

There is also one thing you should know before choosing Paddle. In June 2025 it paid $5 million to settle a US Federal Trade Commission complaint alleging it had processed payments for deceptive tech-support merchants and masked their dispute rates from the card networks [34]. Paddle neither admitted nor denied the allegations, accepted a narrow ban on serving that merchant category, and says the case traces back to two clients out of thousands [35]. I read the settlement as pressure toward stricter merchant screening rather than a reason to walk away, but read it yourself; it is the kind of fact a recommendation should carry openly.

If reachable humans are your first requirement, start with PayPro Global instead. It holds the best seller reviews of the nine [7] and advertises phone, chat, and email support around the clock, with account managers during onboarding [36]. The trade is opacity: you must ask for a price [11], and onboarding runs through sales rather than a self-serve signup. FastSpring occupies similar ground with solid seller reviews on G2 [37], quote-based pricing [12], and that 45-day first hold [27]: reasonable for an established product moving real volume, heavy for a first launch.

Stripe Managed Payments is the one to watch. Link, Stripe’s consumer brand, becomes the seller of record, and Stripe takes on tax in more than 80 countries, disputes, fraud, and payment support [38] for 3.5% on top of your normal Stripe processing fees [13]. It accepts only fully automated digital products and runs every business through an eligibility review [2]. What it lacks is a track record: I could not find any meaningful body of seller reviews for it yet, so you would be trusting Stripe’s general reputation rather than evidence about this product. Apply, test it, but let someone else be the case study.

For a cheap experiment, Creem’s 3.9% + $0.40 is the lowest fully published MoR rate here [14], and Polar is built around usage meters and credit billing aimed at AI products [39], with tiers that get cheaper as monthly fees rise [15]. I would happily run a side project on either. I would not park a main income on a provider this young, for the reasons in the previous section.

How should you integrate a merchant of record?

Get approved before you build, test only in the sandbox, and keep the provider behind a small interface so you can leave later. Then verify one real refund and payout before spending money on traffic.

Paddle’s own docs recommend starting live verification before you integrate against the sandbox, because domain and product review takes days [33]. The same order applies everywhere: put the real website online first with visible pricing, a refund policy, terms, and contact details, submit it for review, and only invest integration time once the live account exists. People do build the whole thing first and then get rejected.

Test only in the test environment. Every live charge you run against your own product during development is a fraud signal waiting to be misread, and that is the mistake I made on Whop, partly because Whop had no sandbox back then. Today it does, with test cards and separate hostnames, though payouts still cannot be simulated there [40], which is unfortunate given that payouts are the thing its reviews complain about.

Keep a thin wall between your code and the provider. One small interface for checkout, cancellation, refunds, and webhooks; your own database as the record of who is entitled to what; the provider’s customer and subscription IDs stored on your side; webhook events verified, deduplicated by ID, and treated as the only trigger for granting access, never the success-page redirect. This is a day of work, and it turns a future provider switch from a rewrite into a migration. My own database for that record is Convex, for the reasons I give in my comparison of eight Convex alternatives.

Then run one real purchase after live approval, refund it, and wait until the first payout actually lands in your bank account before spending anything on traffic. A merchant of record removes the tax filings, the chargeback letters, and the fraud rules. It does not remove the need to know exactly who is holding your money. I check that now the way I check the license on a dependency: before I build, not after. When you are ready for that traffic, how founders get their first 1,000 users without ads covers where to find it.