Got blocked by Stripe? Here's the 72-hour playbook.

On a random Tuesday in February 2025, a customer of ours — a 12-person AI inference company doing $480k/month — got an email at 4:53am Pacific that started with "After a recent review of your account, we've identified activity that's outside of our Stripe Services Agreement." By 7am the dashboard showed a 25% reserve on incoming volume. By 11am payouts were paused. By 2pm a customer Slack escalation hit their support inbox because their billing portal was returning 402s.

This was the eighth time we'd watched this exact sequence in twelve months. By the eighth time we'd learned the only thing that matters: don't spend day one negotiating with Stripe Risk. Spend day one getting processing back up.

Here's the full playbook.

What "blocked" actually means at Stripe

Stripe has a few different escalation states. They look similar from the outside but produce very different outcomes:

The escalation almost always goes top-to-bottom. If a reserve appears, payouts will usually pause within 1–2 weeks unless you change the underlying signal. If payouts pause, the "new charges disabled" step is usually 5–14 days behind. Once new charges are disabled, full closure within 30 days is the modal outcome.

Why AI companies in particular get flagged

Stripe's underwriting model was built for SaaS in the 2015–2020 era — fixed monthly subscriptions, low chargeback rates, predictable MRR. AI inference businesses trip almost every signal that model uses to flag risk:

1. Volatile MRR. Usage-based revenue spikes and dips. Stripe's models read a 4× weekly spike as "sudden volume increase" — a fraud signal. For you, it's Tuesday after a Product Hunt launch.
2. High-ticket international cards. A $400 annual plan paid by a card issued in Brazil or India looks identical to fraud-pattern card testing. AI buyers often pay in international currency, on consumer credit cards, in non-standard amounts.
3. Chargebacks on "I didn't recognise this charge." Consumer customers forget what they subscribed to. AI brand names are short and unfamiliar ("PromptCo", "Lattix", "Atria"). When the customer's spouse sees the charge on the statement, the dispute reason code is "fraud" — which counts against your fraud rate, not your dispute rate.
4. Adult / political / financial content adjacency. If your model can generate it, Stripe's content review treats your business as if your business does. We've seen image-generation companies flagged because users typed NSFW prompts that never produced an output.
5. Refund spikes after a product change. One pricing change → a wave of refund requests → Stripe's refund-rate signal trips.
6. Crypto-adjacent verbiage. If your marketing site says "tokens" — and yours does, because LLM context windows are measured in them — risk systems built before 2023 still occasionally flag this.

None of these are fraud. All of them produce the same signal as fraud in a model that was tuned in 2018.

The 72-hour playbook

Hours 0–6: triage, don't respond

The temptation is to reply to Stripe Risk immediately, in detail, with documentation. Don't. Risk reviews are queued; the first 24 hours of your response window is spent in a queue regardless. Use those hours to:

• Export everything from Stripe. Customers, subscriptions, invoices, payment methods (yes, you can export the IDs; the cards themselves are migrated via PCI-compliant network). Use the data export tooling under Dashboard → Settings → Data exports.
• Pull the last 90 days of chargeback evidence from Dashboard → Disputes.
• Pull your tax remittance history if you're on Stripe Tax.
• Quietly notify your bank that payouts may stop arriving for ~30 days. They'll be more flexible on the credit line if they know in advance.
• Take inventory of subscription state. Which customers are mid-trial, which are on annual plans, which are in dunning.

Hours 6–24: stand up a backup processor

Sign up for a Merchant of Record (Macropay, Paddle, FastSpring, Lemon Squeezy) or a second processor (Adyen, Braintree, Checkout.com). The MoR path is much faster — you get a working checkout in under a day because the MoR is the seller; you don't need to be re-underwritten. The second-processor path is slower (3–10 days for underwriting).

At Macropay, we have a dedicated "Stripe migration in progress" intake that gets a working checkout in < 4 hours and runs the import job (subscriptions, customer records, card-on-file tokens via network token exchange) in parallel.

Hours 24–48: switch new charges, keep Stripe subscriptions live

Point new signups at the new processor. Don't cancel your existing Stripe subscriptions yet. If Stripe is still processing them (most accounts can still process existing recurring charges even after "new charges disabled"), that revenue keeps flowing. You'll migrate it in waves over the next 4–8 weeks.

Cancelling existing subscriptions on day two is the single biggest unforced error we see. It accelerates the cash crunch by 60+ days.

Hours 48–72: respond to Stripe Risk — once, in writing

Now you reply. Three things only:

1. A factual description of your business (one paragraph, no marketing copy).
2. The specific signals that likely tripped: a chargeback spike, a usage spike, a refund wave — whatever you actually know happened. Be honest. Risk reviewers can see the data; lying loses you the appeal.
3. The remediation you've already done: tighter signup verification, address validation, an explicit refund policy on the checkout page, an updated descriptor on the credit card statement.

Don't threaten to leave. Don't mention your VC. Don't escalate to sales reps you talked to at Stripe Sessions. None of that helps. The risk team is siloed from sales and account management for exactly this reason.

What not to do

• Don't open a second Stripe account under a sister entity. Stripe's underwriting catches this in 99% of cases (same EIN/ABN, same beneficial owners, sometimes same descriptor). It moves you from "reserve added" to "account closed" in days.
• Don't initiate chargebacks against Stripe to retrieve funds. The funds are legally Stripe's during the hold period (per Services Agreement section 6). Bank chargebacks against PSPs almost universally fail and burn the relationship with your bank.
• Don't tweet about it. The temptation is real. Public escalation has, in our observation, never accelerated a Stripe Risk decision. It has occasionally accelerated closure, which is the opposite of what you want.
• Don't lawyer up before the 14-day mark. Stripe's appeal process is real and reasonable on the 7–14 day timeline. Engaging counsel before then is expensive theatre. After 14 days with no movement, then yes.

If your payouts are frozen (and they will be)

Three things help during a payout freeze:

1. Float from your new processor. If you moved fast on hours 6–24, you have revenue arriving from a different source 48 hours later. That covers payroll if you have under ~$300k/month payroll.
2. Venture debt or a credit line draw. Most facilities allow draws of up to ~70% of AR. The frozen Stripe balance counts as AR for most lenders, though they'll discount it.
3. Customer prepayments. Email your top 20 customers and offer 10–15% off annual prepayment. Most respond. We've seen this raise 30–60 days of runway in a week.

Using the suspension as a migration window

Here's the counterintuitive bit. The companies who came out of a Stripe pause stronger are the ones who treated it as a forcing function to actually rebuild their billing: move to a Merchant of Record, get the tax exposure off their balance sheet, switch to per-transaction pricing that's legible to customers, fix the dunning flow.

The companies who came out worse are the ones who got reinstated by Stripe in 30 days and went back to the same setup. Every single one of them got flagged again within 12 months.

If you've been on Stripe and you're reading this before a pause: the most valuable thing you can do is have a working second processor / MoR account spun up now. It costs nothing to have it on standby. Macropay's migration endpoint (POST /v1/migrations/import) is idempotent and SSE-streamed, which means you can dry-run a Stripe import today and have a tested cutover ready to fire when you need it.

We were live on Macropay 14 hours after Stripe paused us. The 60 days that followed were the most productive eight weeks the company had ever had — because we finally owned the billing stack instead of hoping a black box wouldn't flag us again.— Anonymous founder, AI inference company, 2025

Preventing it next time

1. Don't be the seller of record. An MoR carries the underwriting risk. You become a B2B vendor to the MoR, which is a much smaller surface.
2. Run two rails. One MoR for B2C, one PSP for B2B contracts you invoice directly. If one rail stalls, the other keeps running.
3. Stable descriptor. The descriptor on the credit card statement should match the product name your customer recognises. "NIMBUS AI", not "NIMBUS LABS INC".
4. Explicit refund policy in the checkout flow. Three sentences. Cuts chargeback rate noticeably.
5. Watch your refund and dispute rates weekly. Above 1% refund or 0.5% disputed is your warning zone. Above 1.5% / 0.9% you're in the "reserve added" risk window for most processors.

If you're mid-suspension right now, our calculator will show you what Macropay actually costs at your volume. The runbook above is what we walk every paused-Stripe team through; the migration endpoint is real and the four-hour timeline is what we measure ourselves against. The full path is documented under Quickstart.