• AI visibility

Guide · AI search

How does a payment processor get onto the shortlist when merchants research with AI?

By making its pricing model, market coverage, performance and integration facts easy for AI assistants to find, verify and repeat. High-volume merchants and platforms choose processors through long evaluations, but the long list now forms partly in AI answers, and the engineers who integrate payments increasingly work through AI tools. A processor described vaguely or wrongly at that stage may never receive the request for proposal.

By Underneath Team Guide Updated

The short version

  1. A few merchants drive the business: Adyen (opens in a new tab) processed €803.8 billion in the first half of 2026, up 24%, and said 300 merchants accounted for roughly 60% of its growth.
  2. Enterprise merchants spread volume: in the 2025 Global eCommerce Payments & Fraud Report from Visa Acceptance Solutions and the Merchant Risk Council, merchants used 3 to 4 payment gateways and acquiring banks.
  3. Pricing models need explaining: the Federal Reserve (opens in a new tab) caps covered debit interchange at $0.21 plus 0.05% of the transaction, plus a $0.01 fraud-prevention adjustment, while credit interchange varies by card and network.
  4. Answers change by market: in our country study, the country’s share of the variation in brand lists was 0.434 for financial services and insurance, the highest of any industry measured.
  5. Integration work runs through AI tools: Stack Overflow’s 2025 survey (opens in a new tab) found 84% of developers use or plan to use AI tools in development.

Who buys payment processing at scale, and what is one merchant worth?

A group of payments, finance, engineering and procurement leaders; one large merchant can move a processor’s growth.

For an enterprise retailer, a subscription business, a travel company or a software platform that embeds payments, choosing a processor is a multi-year infrastructure decision. The head of payments owns authorization rates and cost; the chief financial officer owns the fee line and treasury; engineering owns the integration; procurement runs the process. Across industries, Forrester (opens in a new tab) found procurement professionals serve as decision-makers 53% of the time in the average business buying cycle.

The value of each win is unusually concentrated. Adyen reported net revenue of €1,302.9 million in the first half of 2026, up 19%, and said about 70% of its growth came from existing customers. That last figure is the shape of the business: processors often win a share of a merchant’s volume first and grow it later. Platforms are the fastest-growing segment: Adyen’s Platforms unit processed €135.0 billion, up 42%.

Stripe shows the same reach at the top of the market. Its 2025 annual letter (opens in a new tab) says it serves 90% of the Dow Jones Industrial Average, and that businesses on Stripe generated $1.9 trillion in volume. For processors, the commercial question is less “how many sign-ups?” and more “are we invited into the next large evaluation?”

Where does AI already sit in a processor evaluation?

In the early research, the questions engineers ask while scoping integration, and the drafting of evaluation criteria.

We found no public survey on how many payments leaders use AI assistants to build a processor long list. Across business purchases, Forrester reports that 94% of buyers use AI during their buying process and that they then validate the output with peers, analysts and experts. That is a cross-industry figure, but it describes the role AI plays before a request for proposal: breadth first, validation later.

Engineers are a second route in. Stack Overflow’s 2025 survey found that 84% of developers use or plan to use AI tools in their development process, while 46% said they do not trust the accuracy of the output. Processors have noticed. Stripe’s documentation (opens in a new tab) offers each page as Markdown, a button to copy it for AI tools, and a server that gives coding agents “up-to-date technical guidance.” When an engineer asks a coding assistant how hard it would be to add a second processor, the answer draws on whatever documentation it can read.

Agentic commerce is a third. Financial IT (opens in a new tab) reports that Adyen launched Adyen Agentic, “enabling enterprise merchants to securely process payments across AI agent protocols,” and Stripe describes its work with OpenAI on the Agentic Commerce Protocol. A reasonable expectation is that support for these protocols will start to appear as a line in evaluation criteria.

What do payments leaders and engineers ask AI about processors?

Pricing models, market coverage, performance, redundancy and integration effort. The sample prompts that follow are our own wording of typical buyer and engineer questions, not logged queries.

StageIllustrative prompt
Pricing model“Interchange-plus vs blended pricing for $80 million a year in card sales”
Coverage“Which processors offer local acquiring in Brazil, Japan and the EU?”
Comparison“Adyen vs Stripe vs Checkout.com for a marketplace paying out sellers”
Redundancy“How do we add a second acquirer without rebuilding checkout?”
Performance“How can we raise card authorization rates on recurring payments?”
Process“What should a payment processing RFP include?”
Integration“How long does a migration from our current gateway usually take?”

These map onto what merchants say they measure. The Visa and Merchant Risk Council report, based on 1,082 merchant professionals in 38 countries, found six metrics rated extremely important by more than 4 in 10 merchants: revenue, success rate, loss rates, authentication rate, authorization rate and cost of payments. A processor that publishes clear, checkable material on each gives an AI answer something to cite.

The pricing question deserves special care. Interchange is set by card networks and, for regulated US debit, capped by the Federal Reserve’s Regulation II; the processor’s markup sits on top. Public price lists show both models: Adyen (opens in a new tab) charges a fixed processing fee of $0.13 per transaction plus a payment-method fee, with one card line priced as “$0.13+Interchange+ + 0.60%,” while Stripe (opens in a new tab) offers IC+ pricing in custom packages for businesses with large volume. An answer that collapses these into one headline rate misleads the buyer.

How does an AI answer become processed volume?

By shaping the long list, which decides who receives the request for proposal, then a share of volume that grows.

The path, as we infer it from how processors are bought:

  1. Framing. A payments lead asks an assistant to explain pricing models or list processors for a set of markets.
  2. Long list. Analysts, peers and AI answers produce the names worth inviting.
  3. Request for proposal. Procurement invites a short list and scores responses.
  4. Technical proof. Engineers test documentation, sandboxes and migration effort.
  5. Pilot share. The merchant routes part of its volume to the new processor; multi-acquiring makes this common.
  6. Growth. Share rises as performance proves out.

Steps 1 and 2 are where AI search can matter, and they are the hardest to see. A processor missing from the long list never reaches step 3, and nothing in its pipeline reports shows the loss. Our guide to how AI assistants shape enterprise software shortlists covers the same pattern in a neighboring market, and linking AI answers to pipeline explains how to measure it.

What decides whether an assistant names a processor for a merchant’s needs?

Platforms document little about vendor choice; studies point to market-specific sources, independent coverage and accurate pricing facts.

Documented by platforms. Google says its AI features may use “query fan-out” (opens in a new tab), issuing several related searches across subtopics. A question about processing in Brazil, Japan and the EU may therefore pull in separate sources for each market. We found no platform documentation on how assistants choose payment processors.

Observed in studies.

  • Market matters. In our country study, the gap between same-country and different-country answers on ChatGPT was 0.254 for high-dependence questions such as insurance, lending and tax, against 0.053 for global products and software. About half (52.7%) of ChatGPT’s same-country advantage went with differences in the websites it cited.
  • Independent coverage. Analysts and payments press carry weight: in our brand study, a tenfold rise in independent sites mentioning a brand across the cited pages was linked to 4.7 times the odds that assistants recommended it.
  • Pricing conditions get lost. In our pricing study of software plans, 61.9% of 840 quoted plan prices were fully faithful, and another 3.8% had the right amount but dropped a condition that changes what a buyer pays.

Our inference for processors. Coverage claims should be specific and verifiable: local acquiring licenses, domestic clearing access and supported payment methods by country. Adyen’s own results name such facts, including direct access to France’s domestic interbank clearing system and a license from the Central Bank of the UAE. Performance claims need a stated method, because buyers and assistants both need something to check. Pricing pages should say plainly which model applies to whom.

What does a processor lose when it is missing from AI answers?

Invitations to evaluations it never hears about, in a market where a few merchants drive most of the growth.

We found no public data measuring lost requests for proposal due to AI answers. The exposure follows from concentration: when 300 merchants account for roughly 60% of a leading processor’s growth, a single missed evaluation can matter more than thousands of small sign-ups. And because merchants typically spread volume across several gateways and acquirers, being absent from the next evaluation means losing the chance at a share, not just one deal.

What does GEO look like for a payment processor?

Generative engine optimization (GEO) here means making pricing, coverage, performance and integration facts findable and confirmed.

  1. Pricing-model explainers. Plain pages on interchange-plus, blended and custom pricing, who each suits, and a worked example, consistent with sales materials.
  2. Coverage by market. A current table of countries, local acquiring, settlement currencies and payment methods, with the licenses and clearing access that back it.
  3. Readable documentation. Public, current integration guides in formats AI tools can read; our guide on API companies in AI search covers this in depth.
  4. Performance with method. Authorization, fraud and cost results stated with how they were measured, never as unqualified promises.
  5. Independent coverage. Analyst reports, payments trade press, merchant case studies with named customers, and conference talks by merchants.
  6. Evaluation-ready facts. Public answers to the questions procurement asks: security certifications, uptime reporting, data residency and migration support.
  7. Visibility by market. Track the evaluation questions in each region you sell into, because answers differ by country.

For adjacent buyers, see our guides on fintech software sold to banks and finance teams and ecommerce platforms. Small merchants choosing a provider at founding are covered in how payment companies win new businesses. No one can guarantee a processor appears in AI answers; GEO makes the evidence they draw on accurate, current and specific.

Which questions about AI and processor selection remain open?

How often payments leaders consult AI before a request for proposal, and whether being named changes who gets invited.

  • No payments-specific survey. The 94% figure covers business buyers in general, not payments teams.
  • Developer trust is mixed. Many developers use AI tools but distrust their accuracy, so the weight engineers give to AI answers is uncertain.
  • Studies outside payments. Our country, brand and pricing studies did not test processors specifically.
  • Agentic commerce is early. Protocol support is new, and we found no data on how often it decides an evaluation.

Where should a payment processor start?

Start by running the questions your next evaluation will raise, by market, and checking how assistants describe you.

That first check shows which processors are named for each region and use case, whether your pricing model and coverage are described correctly, and which analyst, press and documentation pages the answers rely on. It also shows what an engineer’s coding assistant says about integrating with you.

If more invitations to enterprise and platform evaluations are the goal, ask us to review your evaluation-stage visibility. We will test the pricing, coverage, comparison and integration questions buyers and engineers ask, find where answers miss or misstate you, and plan the documentation, coverage and content work that gives assistants accurate evidence. The way we run it, market by market for payments leads, engineers and procurement, is described on our generative engine optimization service page.

Frequently asked questions

Do payments teams really use AI assistants to compare processors?

No payments-specific survey exists that we found. Across industries, Forrester reports 94% of business buyers use AI during buying, then validate with peers and analysts.

Why do AI answers get processing fees wrong?

Processor pricing mixes interchange, network fees and markups, often on custom terms. In our software pricing study, some answers kept the amount but dropped conditions that change the price.

Does our developer documentation affect AI visibility?

Plausibly, for integration questions. Stripe offers its documentation as Markdown and through a server for coding agents; our llms.txt study found 11.5% of top websites publish an llms.txt file.

Should a processor publish its prices if most deals are custom?

Publishing the pricing model and how it works helps answers stay accurate even when final rates are negotiated. Adyen and Stripe both publish their models.

Sources

Free strategy call

Some questions are easier to answer about your own business.

Bring the one that matters most. On a free 30-minute call we’ll take a first look at it and send you a short written read afterward.

Ask about my businessHow we work