Credit Card Surcharges: How AI Can Cut Fees Fairly

A credit card surcharge is basically a pricing confession: “We’re paying a lot to accept this payment method, and we’re done hiding it.” And in December—when checkout lines are long, carts are full, and patience is short—surcharges land like a penalty flag at the register.

That’s why American Express CEO Steve Squeri’s recent pushback on surcharging resonated. His argument is simple: surcharging creates a bad customer experience, and customers shouldn’t feel punished for using a card they like. He even suggested the “cleaner” approach is to embed costs in prices rather than surprise people at checkout.

Here’s the part that matters for payment leaders: surcharges aren’t just a PR issue. They’re a signal that payment economics and payment infrastructure are out of sync. And that’s exactly where AI in payments and fintech infrastructure can do real work—reducing unnecessary fees, making routing smarter, and bringing transparency to cost without turning checkout into a negotiation.

Why surcharges are getting louder in 2026

Surcharges rise when merchants feel trapped: they can’t easily steer customers to lower-cost payment methods, but they also can’t keep absorbing higher acceptance costs.

Two dynamics are colliding:

1. Premium cards are expensive to accept. American Express swipe fees reportedly average roughly 1.43% to 3.3% of a transaction, while Visa and Mastercard averages are often cited around 1.15% to 2.6%. Those ranges vary by merchant category, transaction type (card-present vs. card-not-present), and pricing model—but the direction is consistent: premium products often cost more.
2. Rules are shifting on what merchants can do about it. The proposed Visa/Mastercard merchant settlement in late 2025 (Amex wasn’t a party) raised the possibility of merchants gaining more practical flexibility, including differentiating treatment of premium cards. Analysts expect fewer outright refusals and more scaled surcharging—because most merchants won’t turn away high-spending customers.

Squeri’s stance—“embed it in your prices… but adding a surcharge is a really bad customer experience”—is a classic network position. Networks want acceptance to feel universal. Merchants want acceptance to feel affordable. Consumers want it to feel simple.

The truth: everybody is optimizing for their own outcome, and the checkout experience becomes the battleground.

The hidden math behind “bad customer experience”

Surcharging feels bad because it violates two expectations customers have at the register:

• Price certainty: the shelf price should roughly match the final price.
• Fairness: two people buying the same thing shouldn’t pay different totals for reasons that feel arbitrary.

Merchants counter with their own fairness argument: why should cash buyers or debit buyers subsidize premium rewards for someone paying with a high-interchange card?

That tension isn’t going away, especially in categories where margins are tight (restaurants, convenience, ticketing, local services) and where card-not-present volume keeps growing (ecommerce, invoicing links, subscriptions).

The practical issue I see most often isn’t ideology—it’s operations:

• Teams don’t actually know their effective cost per transaction in time to act on it.
• Steering tools are crude (blanket surcharge, blunt “cash discount,” or nothing).
• Routing decisions are siloed (fraud team, payments team, product team), so cost reduction increases declines, or approval optimization increases fraud, or both.

This is where AI-driven payment systems earn their keep: optimize the system so the merchant doesn’t need to “fix it” at the register.

Where AI actually reduces payment fees (without annoying customers)

AI can’t negotiate interchange away. It can reduce the avoidable costs around acceptance—especially the ones that push merchants toward surcharging in the first place.

1) Intelligent routing that optimizes for total cost, not just auth rate

Basic routing logic is usually a rules engine: “If network A fails, try network B,” or “Send everything to PSP X.” That’s better than nothing, but it’s not cost-aware.

AI routing can optimize across multiple objectives at once:

• Net approval probability (including issuer behavior patterns)
• Processing and network fees (including blended vs. interchange-plus implications)
• Fraud risk and chargeback probability
• Latency and customer drop-off risk
• Retry strategy (whether a second attempt helps or hurts)

When done right, this becomes a continuous decision system that learns. It doesn’t just chase approvals; it reduces expensive failures (soft declines, retries that trigger issuer suspicion, misclassified MCC/transaction types) that inflate effective costs.

Snippet-worthy take: The cheapest payment is the one that clears on the first attempt with low fraud risk.

2) AI-powered cost observability: know your effective rate in near real time

Most merchants get a monthly statement and a headache. By then, it’s too late.

With modern data pipelines, you can estimate effective processing cost per payment method, per channel, per basket size, per region, and spot what’s driving spikes:

• premium card mix shifts
• card-not-present growth
• increased manual review
• issuer-specific decline patterns
• higher chargeback ratios pushing up risk tooling costs

AI helps by classifying patterns and predicting which segments will be costly before the month closes.

This matters because it enables a better alternative to surcharging: quiet optimization.

3) Smarter steering that doesn’t feel like punishment

If you want to reduce reliance on surcharges, steering has to feel like a benefit, not a threat.

AI can personalize and test incentives in a way humans rarely have time to manage:

• Offer ACH or bank transfer for high-ticket invoices where card fees are brutal
• Offer debit-first prompts in categories with low fraud and high debit adoption
• Promote wallet payments where they improve approval rates and reduce fraud
• Use fee-aware subscription billing strategies (timing, retries, token updates)

The win is subtle: instead of a visible “3% surcharge,” customers see “Pay by bank and save $12.” Same economics, different psychology.

4) Fraud prevention that reduces cost pressure (and the urge to surcharge)

Fraud tools aren’t free. Manual review isn’t free. Chargebacks are definitely not free.

When AI fraud models reduce false positives, you don’t just improve approval rates—you cut:

• operational costs
• chargeback fees
• lost inventory and shipping losses
• the “risk tax” that forces finance teams to squeeze acceptance elsewhere

Surcharging often appears after a business has already absorbed growing invisible costs. Fraud is one of the biggest invisible costs.

What scaled surcharging could look like—and why it’s risky

Industry voices have suggested that if network rules loosen, merchants may introduce scaled surcharges: a higher fee for more expensive cards.

Operationally, that’s harder than it sounds.

The two real risks

1. Customer backlash at the moment of highest intent. A surprise fee at checkout is a conversion killer—especially online where a customer can abandon in one click.
2. A pricing integrity problem. The more versions of “the price” you have, the more customer service tickets and disputes you invite.

There’s also a long-term strategic risk: if customers associate your brand with “gotcha fees,” you’ll pay for it in retention.

If you’re going to do any surcharge-like mechanism, my stance is: make it predictable and explainable. If it can’t be explained in one sentence by a frontline employee (or a checkout tooltip), it’s too complex.

A better playbook: reduce the need for surcharges with AI

If you run payments for a merchant, marketplace, SaaS platform, or PSP, here’s a practical sequence that tends to work.

Step 1: Measure the real problem (effective cost, not stated rate)

Start with a clean set of metrics:

• effective cost rate by tender type and channel
• approval rate by issuer/bin, country, and entry mode
• chargeback rate and reason code distribution
• retry rate and retry success rate
• top drivers of interchange qualification misses (where applicable)

If you can’t measure it, you’ll “solve” it with surcharges because that’s visible and immediate.

Step 2: Introduce AI routing with guardrails

Routing needs governance, not vibes. Put boundaries in place:

• never optimize cost at the expense of fraud risk
• cap retries to avoid issuer distrust
• protect latency for mobile checkout
• monitor fairness (avoid systematically disadvantaging certain customer segments)

Step 3: Replace blunt surcharges with targeted, tested incentives

Use experimentation:

• A/B test bank-pay discounts on large baskets
• Offer debit incentives for repeat customers
• Use post-purchase prompts for lower-cost payment on the next transaction

In December retail cycles, even a small shift in payment mix can be meaningful because volume is so high.

Step 4: Make fee transparency a product feature

If you’re a platform or fintech infrastructure provider, transparency isn’t just compliance—it’s sales.

Give merchants:

• a fee dashboard that maps cost to decisions (routing, fraud settings, payment methods)
• alerts when premium card mix or declines spike n- recommended actions with estimated impact (in dollars, not percentages)

Snippet-worthy take: Transparency is what turns “fees” from a complaint into an optimization problem.

What this means for fintech infrastructure teams

Squeri’s argument focuses on customer experience. Merchants focus on margin. Networks focus on acceptance and brand. The industry will keep debating who should pay for rewards.

Meanwhile, the teams building payments infrastructure can reduce the friction by doing something more practical: make costs measurable, controllable, and explainable—without forcing a checkout penalty.

That’s the north star for this AI in Payments & Fintech Infrastructure series: AI should make payment systems more reliable and more honest. Not more confusing.

If you’re considering surcharges (or already using them), treat it as a sign you need better tooling—especially around intelligent routing, fraud optimization, and cost observability. The companies that win in 2026 won’t be the ones with the harshest fee policies. They’ll be the ones whose payment stack is smart enough to avoid needing them.

What would happen to your conversion rate—and your support volume—if you could cut acceptance costs without changing the price your customer sees at checkout?