Buy now, pay later is likely to stop being a checkout button and start being a search filter. Signals from the last four weeks point to installment terms surfacing inside AI shopping results before a shopper ever reaches a cart, with the standard for doing so published by the end of Q1 2027. The base case here is a dated Universal Commerce Protocol (UCP) release carrying a pay-later capability, authored with Klarna and Affirm, and Google showing pay-later terms on US product results in AI Mode or Gemini by 31 March 2027, with NRF 2027 in mid-January as the first hard checkpoint.
This is not a story about whether shoppers can pay in four inside an AI assistant. That box was ticked in May. The shift that matters is one layer up: the moment an agent decides which products to show, the availability and cost of financing become an input to that decision, the way free shipping became a ranking input a decade ago.
In short
- The prediction: a dated UCP release ships a pay-later capability (discovery-time terms plus a BNPL payment handler) with Klarna and Affirm as named contributors, and Google surfaces installment terms on US product results in AI Mode or Gemini before checkout, both by 31 March 2027. First checkpoint: NRF 2027, mid-January.
- Signal 1 (2 September 2026): UCP seated its Payments Technical Council and, in the same notice, invited a BNPL working group with Klarna and Affirm, with a written mandate to “find BNPL terms during product discovery”.
- Signal 2 (18 August 2026): Klarna’s Q2 release reports 1.2 million merchants, up 54%, five PSPs enabling Klarna as a default-on option across more than $9.5 trillion of volume, and pay-later “coming to Google Search and the Gemini app”, still future tense three months after it was announced.
- Signal 3 (27 August 2026): Affirm’s fiscal Q4 call leaned entirely on the Affirm Card and merchant coverage (10% of US e-commerce merchants, 80 of the top 250 sites) and said nothing about agent surfaces, even though Affirm has been piloting UCP extensions since May. The pilot is running; the public narrative has not caught up.
- Main caveat: US credit-advertising rules make “from $25 a month” language expensive to display at scale, which could push the first version toward a generic “pay later available” badge rather than priced terms.
Why this matters now
Agentic commerce spent the first half of 2026 learning an uncomfortable lesson: native checkout inside an assistant did not convert. OpenAI wound down Instant Checkout in March and moved purchasing into merchant apps, with reporting at the time putting in-assistant conversion at roughly a third of the on-site rate for at least one large retailer. The value moved upstream, to discovery, and discovery is where financing has never lived.
Google’s protocol has been moving the same way. The v2026-08-25 release added store locations, operating hours and by-weight pricing, which are inputs to what an agent shows, not to how it pays. Buried in the same release notes were “deferred payments, deposits, and installment schedules” and multi-instrument split payments.
The schema slots for financing terms exist as of 25 August. What was missing was a party with the data to fill them, and that is what the 2 September notice supplied.
The commercial logic is simple. A shopper who asks an assistant for “a standing desk under $40 a month” is expressing a financing constraint as a product query. Today, no agent can answer that question well, because installment terms are computed at checkout by a BNPL provider that the agent cannot see. Whoever exposes those terms at discovery time gets to shape the shortlist, and the shortlist is where the money is.
Signal 1: UCP invites a BNPL working group with Klarna and Affirm
On 2 September 2026, the UCP maintainers announced the inaugural members of the protocol’s Payments Technical Council: Adyen, Ant International, Coinbase, Global Payments, Google, PayPal, Shopify and Stripe. The notice sits in the project’s public announcements feed and in GitHub Discussion #798. The council itself is notable (Stripe and PayPal both maintain or sponsor the rival Agentic Commerce Protocol, and they now sit on UCP’s payments roadmap), but the more specific tell is further down the page.
The same notice invited statements of interest for two Domain Working Groups. One is a Tokenization group with American Express, Mastercard and Visa as proposed initial members. The other is a Buy Now Pay Later group whose stated purpose is to “add UCP support to find BNPL terms during product discovery and BNPL payment requests processing via payment handlers”, with Klarna and Affirm as the proposed initial members alongside at least one council member.
Read that mandate carefully. It has two halves, and the order is deliberate. Payment-handler processing is the checkout half, and it is the part Google already demonstrated at Marketing Live in May. Finding terms “during product discovery” is the new half, and it is the half that turns a BNPL provider from a checkout option into a data source that feeds the agent’s ranking.
The council cadence adds context. UCP formed a Food council on 16 July (Block, DoorDash, Google, Toast, Uber Eats), a Lodging council on 11 August (Amadeus, Booking.com, Expedia, Google, Hilton, Marriott, Trip.com) and the Payments council on 2 September. Each council has been followed within weeks by vertical-specific schema work landing in the repository. The pattern suggests the BNPL group’s output is scheduled for the next dated release rather than parked as an intent.
Signal 2: Klarna’s Q2 shows the merchant side is done and the surface side is pending
Klarna’s second-quarter release on 18 August 2026 reported gross merchandise volume of $36.6 billion, up 18%, revenue of $1.04 billion, up 27%, and 1.208 million live merchants, up 54% year over year. The merchant growth is the number to watch, and management was explicit about its source: PSP partnerships. J.P. Morgan Payments went live on 6 August, Stripe and Nexi are ramping, and the release counts five PSPs enabling Klarna as a “default-on option across payment networks totaling over $9.5 trillion of volumes”.
That matters for the prediction because it removes the usual objection to discovery-time financing: coverage. If an agent shows an installment price on a product and the merchant cannot honor it at checkout, the display is worse than useless. Default-on PSP distribution means that for a growing share of US merchants, the terms are honorable without any merchant integration at all. The data can safely move upstream because the fulfilment of that data is now near-universal.
The second tell is a tense. Klarna’s release says its flexible payments “are coming to Google Search and the Gemini app within Google Pay”, and that its Shopping Search app “is live in ChatGPT”. The Google integration was announced on 12 May with a “coming weeks” timeline; on 18 August it was still future tense. Three months of slippage on a wallet integration between two willing partners is consistent with a scope that grew, and the working-group mandate suggests what it grew into.
Klarna’s own framing closes the loop. “PSPs bring merchants and merchants bring consumer surfaces,” the release says, before describing AI platforms as the next surface and noting that AI referral traffic converted at higher rates last holiday. A company that has just finished the merchant half of that sentence has every incentive to push the surface half into a standard it does not have to negotiate one platform at a time.
Signal 3: Affirm’s silence on agents while its UCP pilot runs
Affirm reported fiscal fourth-quarter results on 27 August 2026, its most profitable quarter on record excluding a tax allowance release. Management’s growth story on the call, per the published transcripts, was the Affirm Card (19% attach rate among active users, roughly twice the spend of a typical relationship, about 30% of card transactions offline), Pay-in-X growth of 41%, and merchant coverage of roughly 10% of US e-commerce merchants including 80 of the top 250 sites. Transcript summaries record no discussion of AI agents, ChatGPT, Gemini or agentic checkout.
That silence is the signal, because Affirm’s own 12 May press release said the company was “piloting extensions to the Universal Commerce Protocol” alongside its Google Pay, Search and Gemini rollout. Pilots that appear in a May release and disappear from an August call are usually one of two things: quietly abandoned, or quietly promoted into something the company would rather announce at a bigger venue. The 2 September working-group invitation, naming Affirm as an initial member, points to the second reading.
There is a structural reason Affirm needs discovery-time terms more than Klarna does. Affirm’s core product is a monthly installment loan on larger baskets, and the offer is only persuasive when the shopper sees a per-month figure early. Klarna’s pay-in-four is close to price-neutral and works fine as a checkout afterthought. If installment terms move upstream into the agent’s shortlist, Affirm’s product gets a merchandising surface it has never had; if they do not, it stays a checkout widget competing on a crowded button row.
What the pattern suggests
Put the three together and the sequence reads as: protocol slots opened (25 August), merchant coverage secured through PSPs (6 to 18 August), the two US pay-later leaders named to a working group with a discovery mandate (2 September), and the largest consumer surface (Google) still holding the integration in “coming” status. The missing step is the one the working group exists to produce: a standard way to publish and read pay-later terms per product.
| Signal | Date | Source type | What it establishes | Weight |
|---|---|---|---|---|
| UCP BNPL working group invited, Klarna and Affirm named, discovery-time mandate | 2 Sep 2026 | Public governance notice (GitHub) | Intent to standardize terms at discovery, not only at checkout | High |
| Klarna Q2: 1.2m merchants, five default-on PSPs, Google surface still “coming” | 18 Aug 2026 | Earnings release | Coverage problem solved; surface integration scope grew | Medium-high |
| Affirm FQ4: no agent commentary despite May UCP pilot | 27 Aug 2026 | Earnings call transcripts | Pilot moved from marketing into standards work | Medium |
| UCP v2026-08-25: installment schedules, deposits, split payments, Actions primitive | 25 Aug 2026 | Protocol release notes | Schema substrate exists | Supporting |
| Google Marketing Live: Affirm and Klarna in Google Pay, Universal Cart | 20 May 2026 | Product announcement | Checkout half already demonstrated | Context |
UCP’s release rhythm gives a timing anchor. Dated releases landed on 11 January, 23 January, 8 April and 25 August 2026, and the January date coincided with NRF. Google launched the protocol at NRF 2026 and used Marketing Live for the May expansion, so the two obvious venues for the next major cut are the next dated release (a four-month cadence points to December or January) and NRF 2027 in mid-January. That is why the prediction is framed as end of Q1 2027 with a mid-January checkpoint: it allows one release slip.
The precedent for financing terms becoming a ranking input is not hypothetical. Google’s Merchant Center has carried an installment attribute for years, but only for Argentina, Brazil, Chile, Colombia, Mexico and Peru across all categories, and for wireless products and vehicle ads elsewhere. In Latin America, where installment pricing is the default retail language, the attribute is a standard part of the product listing. The prediction is, in effect, that the Latin American product model reaches US general retail via the agent layer rather than via the classic Shopping tab.
| Precedent | Where | What moved upstream | What happened to ranking |
|---|---|---|---|
| Free-shipping and delivery-date feeds | Google Shopping, Amazon | Fulfilment promise shown on the listing | Became a filter and a de facto ranking factor; merchants without it lost visibility |
| Merchant Center installment attribute | Latin America | Monthly price shown on product listings | Listings without installment data underperform in installment-first markets |
| Prime badge | Amazon | Membership benefit surfaced at discovery | Buy Box and search weight shifted toward eligible offers |
| Klarna Shopping Search in ChatGPT | OpenAI Apps | Live prices and offers from a BNPL provider’s catalog | Provider’s data, not the merchant’s, defines the shortlist |
Wider context: financing is becoming the BNPL providers’ distribution, not the merchants’
The larger arc is that pay-later providers are moving their distribution away from merchant checkout integrations and toward consumer-held credentials and platform surfaces. Klarna’s card reached 6.5 million active users, up from 1.3 million a year earlier, and its paid memberships hit 2 million. Affirm’s card carries 19% attach and a growing offline share. Both companies are, in different words, telling investors that the consumer relationship, not the merchant button, is the asset.
Discovery-time terms inside an agent are the natural extension of that strategy. If the assistant already knows the shopper’s Klarna or Affirm standing, it can quote a personalized installment on every product it shows, regardless of whether the merchant has ever heard of either provider. The merchant becomes a fulfilment endpoint for financing decided elsewhere, which is precisely the position merchants spent 2025 trying to avoid in checkout and are now likely to accept in discovery because the alternative is exclusion from the shortlist.
The funding side reinforces the timing. Both leaders have spent 2026 re-plumbing their balance sheets, with the move toward deposit-funded lending lowering their cost of capital and making longer-dated installment offers cheaper to extend. Cheaper capital makes it rational to quote terms more aggressively and earlier in the journey, which is exactly what discovery-time display enables.
Note also who is not in the room. OpenAI’s ChatGPT does not run on UCP; its commerce path runs through the Agentic Commerce Protocol and merchant apps, and Klarna reached ChatGPT through its own MCP server rather than any standard. A UCP pay-later capability therefore covers the Google surfaces and any UCP-adopting platform (Microsoft has wired UCP into Merchant Center, and Shopify’s tooling ships against it), but not the assistant with the largest consumer base. The prediction is scoped to Google for that reason.
Implications for retailers, BNPL providers and platforms
Retailers. The practical consequence is that financing eligibility becomes a feed attribute, alongside price, availability and shipping. Merchants that already publish clean structured data will find this a small addition; merchants whose product feeds are thin will have one more field they cannot supply. The uncomfortable part is that the terms shown may be the provider’s, computed against the shopper’s profile, and the merchant will be asked to honor them, which is the reverse of today’s flow where the merchant chooses the provider and the plan.
For big-ticket categories (furniture, fitness equipment, electronics, travel) the effect on the shortlist is likely to be large, because a monthly figure changes the perceived price band. For low-ticket categories it is close to irrelevant. Retailers should model the incremental provider cost against the visibility gain rather than against checkout conversion, since BNPL merchant fees run well above card costs and the lift now has to be earned earlier in the funnel.
BNPL providers. The working group is a two-seat table, and the seats have names. Klarna and Affirm are positioned to write the terms schema everyone else implements. PayPal Pay Later sits on the council rather than in the working group, Splitit endorsed UCP in March with card-linked installments, and Apple Pay Later is gone.
The likely outcome is a schema that fits the two authors’ products best, which is a standard-setting advantage the rest of the sector will spend 2027 catching up to.
Platforms. Shopify’s Shop Pay Installments runs on Affirm, and Shopify holds a permanent seat on the UCP governing council and one on the payments council. That is the shortest path from working-group output to a million storefronts. For Google, discovery-time financing is a differentiator against ChatGPT that does not depend on winning native checkout, which is a fight the evidence says nobody has won.
Investors. The scoring event for Affirm and Klarna is not a partnership press release; both have those already. It is a dated UCP release with a pay-later capability and an observable change in what Google shows on a product result. The first would likely be treated as a non-event by the market and the second as a re-rating catalyst, which is the usual order for infrastructure stories.
Caveats: what could go wrong
US credit-advertising rules. This is the strongest counter-signal. Under Regulation Z, advertising a specific periodic payment amount for closed-end credit triggers mandatory disclosures (payment terms and annual percentage rate). Pay-in-four with no finance charge generally sits outside that regime; Affirm-style monthly loans do not.
Showing “$27 a month” next to a product in an agent result is legally doable (Affirm’s on-site widgets do it with “as low as” language and APR ranges) but at agent scale, with personalized terms, the compliance surface is large. The plausible fallback is a generic “pay later available” flag at discovery with priced terms deferred to the cart, which would count as a partial miss.
Working-group process risk. The 2 September notice invited statements of interest; it did not charter the group. UCP’s own governance requires a council member in each working group, and the payments council is three weeks old. A group that forms in October and targets a first schema by January is a tight schedule for an industry standard, even one with two motivated authors.
Google may keep financing in the wallet. The May announcement embedded Affirm and Klarna in Google Pay, which is a checkout-side design. Google could reasonably decide that terms belong with the payment instrument and not with the listing, especially given the Merchant Center installment attribute’s limited geographic footprint. That would leave the checkout half standardized and the discovery half stuck in “coming” status.
The stall might be the story. A tense that stays future across two quarters can also mean the integration is stuck rather than growing. If Klarna’s Q3 release in November still says “coming to Google Search and Gemini”, the scope-growth reading weakens and the stuck reading strengthens.
Governance reshuffle. This site has argued that one of the commerce protocols is likely to move under neutral governance by Q1 2027. If UCP is the one that moves, the transition could pause working-group output for a quarter or more, which would push the schema past the March deadline without changing the direction.
| Scenario | What it looks like by 31 March 2027 | Rough odds | Early tell |
|---|---|---|---|
| Base: priced terms at discovery | Dated UCP release with a pay-later capability naming Klarna and Affirm; Google shows per-month terms on US product results in AI Mode or Gemini | 50% | Working-group charter merged in October; installment fields in a UCP pre-release branch by December |
| Partial: badge, not terms | Capability ships; Google shows “pay later available” without a figure pending compliance work | 25% | Schema carries eligibility flags but no amount fields |
| Slip: standard lands after Q1 | Working group active, release cadence pushes the cut to Q2 2027 | 15% | No dated UCP release by NRF 2027 |
| Miss: wallet-only | Financing stays inside Google Pay at checkout; no discovery-time capability | 10% | Klarna’s Q3 release still says “coming”; group stays uncharted |
FAQ
What exactly is being predicted, and how would a reader check it?
Two observable events by 31 March 2027. First, a dated UCP release whose notes include a pay-later or installment capability with Klarna and Affirm listed as contributors, checkable on the protocol’s public GitHub releases page. Second, installment terms visible on US product results inside Google AI Mode or Gemini before the shopper enters a cart, checkable by running a big-ticket product query. Mid-January (NRF 2027 and the expected dated release) is the first checkpoint.
Didn’t Google already add Affirm and Klarna to agentic checkout in May?
Yes, at Marketing Live on 20 May, Google embedded both providers in Google Pay for agentic transactions. That is the checkout half of the working group’s mandate. The prediction is about the discovery half: terms shown while the agent is still deciding what to recommend, which requires a per-product data standard that did not exist in May and that the 2 September notice explicitly sets out to create.
Why would financing terms affect what an agent recommends at all?
Because a large share of considered purchases carry an implicit budget constraint expressed per month, not per item. An assistant that knows a shopper’s approved installment terms can answer “under $40 a month” queries accurately and can rank products by affordability. Once one surface does this, the others have a reason to follow, which is the same dynamic that made delivery dates a ranking input.
Isn’t this just Klarna and Affirm marketing dressed as a standard?
Partly, and that is a reason to expect it to ship. Standards with two motivated authors and an established host tend to move faster than neutral consortia. The counterweight is UCP’s governance, which requires council participation and public review, and the presence of PayPal, Adyen and Stripe on the payments council, none of whom benefit from a schema that favors two competitors. Expect the output to be more generic than either author would write alone.
What about ChatGPT? It has more users than Gemini.
ChatGPT is outside the scope of this prediction. OpenAI’s commerce path runs through the Agentic Commerce Protocol and merchant apps, and Klarna’s ChatGPT presence is its own app built on an MCP server, not a standard. If OpenAI adopted UCP or shipped an equivalent capability in ACP, the prediction would apply there too, but nothing in the last 30 days suggests that step.
Could regulation stop priced terms from appearing at discovery?
It could shape them. Regulation Z requires disclosures whenever a specific payment amount is advertised for covered credit, and Affirm’s longer loans are covered. That does not prohibit discovery-time display; it makes the display carry APR and term disclosures, which providers already manage on merchant sites. The risk is that Google opts for a generic badge to avoid the compliance surface, which is the partial-hit scenario in the table above.
What should a retailer do in the next 90 days?
Three things. Audit whether the product feed carries the fields agents already read (price, availability, shipping, identifiers) because a pay-later field will sit beside them. Decide which provider’s terms the business is willing to honor if quoted upstream, since that choice is becoming a distribution choice. And model the provider fee against visibility in big-ticket categories rather than against checkout conversion, because that is where the lift would come from.
Why frame this as a tech-adoption signal rather than a product launch?
Because the evidence is in the plumbing, not in a press release. A working-group mandate, a release note listing installment schedules, and PSP default-on coverage are adoption-curve markers: they show a capability being built into the substrate before any surface announces it. Product launches follow substrate; the substrate here is visibly being poured.
What would falsify the thesis early?
Three tells before year-end. No working-group charter merged in the UCP repository by the end of October. Klarna’s Q3 release in November still describing Google Search and Gemini as “coming”. And a dated UCP release in December or January whose payment section carries no pay-later capability.
Any two of the three would move this call from likely to unlikely.