Ecommerce Website Development in Pune: What Agentic Commerce (AP2 and Instant Checkout) Means for Your Product Pages in 2026
Two of the largest platform companies in the world spent the last year building competing standards for letting an AI agent complete a purchase on a customer's behalf, and both launched within weeks of each other in September 2025: Google's Agent Payments Protocol (AP2) and OpenAI's Agentic Commerce Protocol, which powers Instant Checkout in ChatGPT. A year on, the picture is genuinely mixed — real infrastructure exists, real merchants are connected, and real purchases are happening through it, but the rollout has also stumbled in ways worth understanding before ecommerce website development in Pune treats this as a settled, must-build-for-immediately feature.
What actually launched, and what it does
Google's AP2 is a vendor-neutral, open protocol that lets an AI agent prove to a merchant or payment network that a real user genuinely authorised a specific purchase, at a specific price. It launched in September 2025 with more than 60 partners at launch, including Mastercard, PayPal, Coinbase, American Express and Salesforce. The technical core is what Google calls "Mandates" — cryptographically signed, tamper-proof digital records. An Intent Mandate captures the user's original instruction to their agent ("find me a pair of running shoes under ₹4,000 in my size"); a Cart Mandate then creates a verifiable record of exactly what's being purchased and at what price before the transaction completes. AP2 is designed to be payment-method agnostic, supporting cards and bank transfers as well as newer payment rails, and it's built to interoperate with OpenAI and Stripe's separate protocol rather than compete with it outright.
OpenAI's Instant Checkout, built on the Agentic Commerce Protocol developed jointly with Stripe, launched the same month, initially connecting US-based Etsy sellers and rolling out to over a million Shopify merchants including recognisable brands. The user experience is straightforward: a ChatGPT Plus, Pro or Free user in the US can complete a purchase from a connected Etsy seller directly inside a chat conversation, with the merchant remaining the merchant of record, fulfilling the order normally, and paying a small fee only on completed purchases — importantly, ranking in ChatGPT's recommendations is organic rather than pay-to-play, which is a meaningfully different model from how sponsored placement works in traditional search advertising.
The honest part of this story: it hasn't gone smoothly
This is where a lot of vendor marketing around "agentic commerce" glosses over an important, publicly reported detail. By March 2026, reporting on OpenAI's rollout described its "first try at agentic shopping" as having stumbled, with OpenAI and its retail partners moving away from the original Instant Checkout experience and toward building dedicated retailer apps within ChatGPT instead. In other words, the specific in-chat, single-click checkout flow that generated so much launch coverage in September 2025 was already being reworked within six months, even as the underlying protocol and partner relationships continued.
For an ecommerce website development in Pune decision, this matters a great deal. It's a genuinely useful signal that this category is real and well-funded, backed by serious infrastructure players, and worth building toward — but it is not yet a stable, finished feature with a fixed integration spec that will look the same twelve months from now. Building an entire checkout redesign around the September 2025 version of Instant Checkout in isolation would have meant reworking it again within two quarters. The more defensible approach is building the underlying data hygiene these protocols all depend on, rather than integrating against one specific, still-evolving implementation.
What both protocols actually need from your website, regardless of which one wins
Strip away the branding differences between AP2, ACP and whatever comes next, and both approaches depend on the same foundational things being true about your product pages and checkout — things that are also, not coincidentally, good ecommerce practice independent of AI agents entirely.
1. Accurate, structured product data. An agent completing a purchase on a user's behalf needs to know the exact price, current stock status, size or variant availability, and shipping cost before it can construct anything like Google's Cart Mandate. A product page where the "real" price only appears after JavaScript renders, where stock status is inaccurate, or where product schema markup is missing entirely, is invisible or unreliable to an agent in exactly the way it's already a weak signal for Google Shopping and rich results today. If your ecommerce website doesn't have clean Product and Offer structured data (schema.org markup) on every listing, that's the actual starting point — not a separate AP2-specific integration.
2. A checkout that doesn't assume a human is reading every screen. Agentic checkout flows need clear, unambiguous confirmation steps — what's being bought, at what price, with what shipping option — expressed in a way a machine can parse reliably rather than in marketing copy wrapped around the actual transaction details. This overlaps substantially with what a well-built checkout system already needs for accessibility and for reducing human cart abandonment: clear line-item pricing, no hidden fees that appear only at the final step, and an unambiguous order summary before payment.
3. Reliable inventory and pricing sync. Both AP2's Cart Mandate and ACP's checkout flow depend on the price and availability an agent sees being the actual, current price and availability — a stale product feed or a delayed inventory sync that shows an item as available when it's actually sold out becomes a much more visible failure when an autonomous agent completes (or fails to complete) a transaction based on it, compared to a human who might notice and adjust.
4. Being present in the product feeds these systems actually read. Whether a customer's AI assistant recommends your product at all depends on the same product feed infrastructure already powering Google Shopping and Meta's catalogue ads — accurate GTINs, category mapping, and complete attribute data. A product not properly represented in your existing product feed isn't going to appear correctly in an agentic shopping flow either, regardless of which specific protocol becomes dominant.
How AP2 and ACP actually relate to each other
It's a common misconception that AP2 and OpenAI/Stripe's Agentic Commerce Protocol are rival standards locked in a winner-takes-all fight, the way VHS and Betamax were. They're solving overlapping but distinct layers of the same problem, and Google has explicitly designed AP2 to interoperate with ACP rather than replace it. Google also introduced a separate, complementary protocol called UCP (Universal Commerce Protocol) around the same period, which governs how an agent and a merchant communicate about the shopping journey itself — product discovery, cart building — while AP2 specifically governs the payments layer: proving a real user authorised a specific charge. ACP, in Stripe and OpenAI's implementation, bundles a version of both concerns into one flow for the specific case of ChatGPT-driven purchases. The practical takeaway for a merchant is that these protocols are converging toward a shared set of underlying requirements — verifiable user intent, an auditable record of what was agreed to, and standard payment authorisation — rather than forcing a business to pick a side the way an early web developer once had to bet on Flash versus HTML5.
Liability and dispute questions worth asking before you connect anything
One question every ecommerce business should have answered before connecting to any agentic checkout system: if an AI agent misinterprets an instruction and buys the wrong size, colour or quantity, who is responsible for the return, and does it follow the same return policy as a normal order? AP2's Mandate system is specifically designed to create a cryptographically verifiable record of exactly what was authorised, which exists precisely to answer this kind of dispute cleanly — a Cart Mandate showing the exact item, price and quantity a user's agent agreed to removes a lot of the ambiguity a phone-order dispute might otherwise involve. Even so, a Pune ecommerce business connecting to any agentic checkout flow should confirm in writing, before going live, how return and dispute handling works for agent-initiated orders, and whether existing consumer protection obligations under India's ecommerce rules apply identically to an agent-initiated purchase as to a manually placed one. This is a question for your payment gateway and legal counsel specifically, not something to assume by analogy.
Should a Pune ecommerce business actually integrate today?
For most small and mid-sized Indian ecommerce businesses, the honest answer is not yet, and not because the technology isn't real — it's because both protocols currently centre on US-based merchants and US ChatGPT users, and neither has confirmed a general availability timeline for Indian merchants or Indian payment rails like UPI as of this writing. Building direct AP2 or ACP integration today would mean investing engineering time against a partner list and API surface that may look different by the time it reaches India. TODO: watch for an official Google or OpenAI announcement naming India-specific merchant onboarding or UPI support before treating direct integration as a near-term priority.
What's genuinely worth doing now, and worth doing regardless of the timeline for India-specific rollout, is the foundational data work described above — clean product schema, accurate real-time inventory sync, and an unambiguous checkout summary. That work pays off immediately in traditional search and shopping ads performance, and it happens to be exactly what any future agentic commerce integration will require as a prerequisite, so none of it is wasted effort if the India timeline moves faster or slower than expected.
A practical readiness checklist
Rather than waiting for a formal India launch announcement and then scrambling, most of the following can be done as part of ordinary catalogue and storefront maintenance:
- Audit every product page for complete schema.org Product and Offer markup — price, availability, SKU/GTIN, and variant data
- Confirm inventory sync between your storefront and your actual warehouse or POS stock has no meaningful lag (same-day sync at minimum)
- Review your checkout summary screen for unambiguous line-item pricing, with no fee that only appears at the final confirmation step
- Check that your product feed (the one already feeding Google Shopping or Meta catalogue ads) has no missing or malformed attributes for your top-selling SKUs
- Confirm your returns policy language works for a scenario where the initiating "customer" is technically an AI agent acting on a real buyer's behalf
- Ask your payment gateway provider directly whether they have any stated roadmap for AP2 or ACP support relevant to Indian merchants
None of this requires committing engineering time to a specific protocol integration before it's clearly available and stable for Indian merchants — it's the same groundwork that improves your existing Google Shopping and marketplace performance today, with the added benefit of being ready quickly whenever agentic commerce does extend to India in a stable form.
What to ask a web development company in Pune before your next catalogue update
If you're planning a product catalogue migration, a new theme, or a checkout redesign in the next few months, it's worth explicitly asking whether structured product data (schema.org Product/Offer markup) is part of the scope, rather than treating it as an SEO afterthought bolted on later. It's meaningfully cheaper to build this in from the start of a redesign than to retrofit it onto an existing catalogue of hundreds of SKUs later — and building it now means you're ready the moment agentic commerce protocols do extend properly to Indian merchants, rather than starting the data cleanup from scratch at that point.
Frequently asked questions
Do I need to sign up for AP2 or ACP directly right now?
Not urgently, and for most Indian merchants there isn't yet a clear signup path tied to Indian payment rails or a confirmed India launch date. The useful preparation is the underlying data quality work, which benefits your existing search and shopping visibility regardless of when (or whether) direct integration becomes relevant for your specific business.
Is this the same thing as the AI crawlers and llms.txt conversation my web developer already raised?
Related but distinct. Crawler access (GPTBot, ClaudeBot, PerplexityBot reading your content) determines whether an AI assistant can describe or recommend your products at all. Agentic commerce protocols like AP2 and ACP are a separate, later layer — they govern how a completed transaction actually gets authorised and processed once an agent has decided to buy something. You can be crawler-friendly without being agentic-commerce-ready, and vice versa; they're solved by different technical work.
What happens to my margins if a customer buys through an AI agent instead of my own website?
Under OpenAI's current model, the merchant remains the merchant of record and pays a small fee only on completed purchases, similar in structure to a marketplace or payment gateway fee rather than a full commission model — though exact fee structures are set by each protocol's commercial terms and are worth confirming directly with the platform before assuming a specific rate applies to your business.
Will this replace my existing website and checkout?
No. Every version of agentic commerce launched so far routes through a merchant's existing product catalogue, checkout logic and order fulfilment — it's an additional purchase channel, similar in concept to how a marketplace listing is an additional channel alongside your own site, not a replacement for your own site's checkout infrastructure.
Given OpenAI already reworked Instant Checkout once, why invest in this at all?
Because the specific finding from the March 2026 reporting was that the narrow, single-item, in-chat checkout flow needed rework — not that agentic commerce as a category failed. The underlying infrastructure (Stripe's Agentic Commerce Protocol, Google's AP2, the partner network around both) continued to develop through the pivot. Treat the specific UI implementation as unsettled, but treat the underlying requirement — clean, machine-readable product and pricing data — as a safe, durable investment regardless of which specific interface wins out.
Where to start
If your product pages don't currently carry complete, accurate schema.org structured data, or your inventory sync between your storefront and your actual stock has any known lag, those are the two highest-value fixes regardless of how agentic commerce specifically plays out over the next year. As a team handling ecommerce website development in Pune, Govindani Infotech builds structured product data and real-time inventory sync into every storefront we deliver, and can audit an existing catalogue for gaps — contact us to have your product feed and checkout reviewed against this.
Sources:
- Announcing Agent Payments Protocol (AP2) — Google Cloud Blog
- AP2 - Agent Payments Protocol Documentation
- Stripe powers Instant Checkout in ChatGPT and releases Agentic Commerce Protocol — Stripe Newsroom
- Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol — OpenAI
- Etsy pops 16% as OpenAI announces ChatGPT Instant Checkout — CNBC
- OpenAI's first try at agentic shopping stumbled. It's trying again — CNBC