Agentic Commerce in India: What AI Shopping Agents Mean for Your Brand in 2026

Agentic commerce is the shift from people browsing shops to AI agents researching, comparing and, in some cases, buying on their behalf. For Indian ecommerce and D2C brands the useful question is not whether to allow it. It is whether your product data is clean enough for an agent to understand you and recommend you in the first place.

This guide covers what agentic commerce actually is, what has and has not happened so far, what is being built for the Indian payments stack, and the practical work that makes your store agent ready.

What Agentic Commerce Actually Means

In a normal online purchase, a person does the work. They search, open five tabs, compare prices, read reviews, then check out.

In agentic commerce, an AI agent does most of that. You describe what you want, the agent goes and finds options, narrows them down, and hands you a short list or completes the purchase inside the assistant itself. The shopper never sees a category page, a filter sidebar, or in many cases your homepage.

That matters because every layer of your store that was built to persuade a human is skipped. What the agent sees instead is your product data, your reviews, your prices, and what third-party sites say about you. This is the same shift covered in what is AEO and why Indian businesses need it, applied to a shopping cart rather than a blog post.

Where Agentic Commerce Actually Stands in 2026

There is a lot of noise here, so it helps to separate what has shipped from what has been announced.

Discovery is working. Adobe Analytics reported that traffic from AI sources to United States retail sites grew 393 percent year over year in the first quarter of 2026. More interesting than the growth is the quality. In March 2026, Adobe found AI-referred traffic converted 42 percent better than non-AI channels. A year earlier, in March 2025, the same traffic converted 38 percent worse. Adobe also measured 13 percent more pages per visit and 48 percent longer time on site from AI referrals.

That is a channel that flipped from curiosity to intent in twelve months. Note that this is United States retail data. India-specific figures at this scale are not published yet, but the underlying behaviour, people asking an assistant instead of a search box, is not country specific.

Checkout inside the assistant has been slower. OpenAI and Stripe launched Instant Checkout in ChatGPT on 29 September 2025, starting with United States Etsy sellers and promising over a million Shopify merchants. Forrester analyst Emily Pfeiffer noted that by February 2026 roughly 30 Shopify merchants had actually gone live on it. Consumer appetite is part of the reason. In a Forrester survey of 700 consumers, only about a third said they would be willing to complete a payment inside an answer engine at all, mostly over data privacy.

The volume is real even if the checkout is not. Research from OpenAI's economic research team with Harvard found that around 2 percent of ChatGPT queries are shopping related, which works out to roughly 50 million shopping queries a day.

So the honest picture for 2026 is this. People are using AI to decide what to buy far more than they are using it to pay. Discovery is the battleground right now, and that is good news, because discovery is the part you can influence with work you already know how to do.

The Plumbing Being Built Underneath

Three protocols are worth knowing by name, because your payment provider and your platform will start mentioning them.

• Agentic Commerce Protocol (ACP). Co-developed by Stripe and OpenAI and released as an open standard. It defines how an agent talks to a merchant and uses Shared Payment Tokens so the buyer's card details are never exposed to the agent. Merchants who do not use Stripe can still adopt it with their existing provider.
• Agent Payments Protocol (AP2). Announced by Google on 16 September 2025 with more than 60 partners including Mastercard, American Express, PayPal, Adyen and Worldpay. It solves the trust problem with Mandates, which are cryptographically signed records of what the shopper actually authorised. An Intent Mandate captures the request, a Cart Mandate locks the exact items and price, and both are signed as Verifiable Credentials so there is an audit trail if something goes wrong.
• Unified Agent Protocol (UAP). NPCI's proposed standard for agent-led transactions on UPI, being developed in consultation with industry and requiring RBI approval before it can launch. Reporting on it describes agents from merchant apps, payment apps or assistants handling low-consideration repeat purchases such as groceries and dairy, with registration and verification of trusted agents, limits on agent authority, dispute and chargeback handling, and audit trails.

The common thread is that all three are trying to answer the same question: how does a merchant know an agent is genuinely acting for a real customer, and who is liable when it is not.

The Amazon Signal

Amazon folded its Rufus shopping assistant into a unified Alexa for Shopping experience. The adoption numbers it reported are the part worth noting: over 300 million customers used Rufus during 2025, with monthly active users up more than 115 percent and engagement up nearly 400 percent year over year. It also added autonomous actions such as buying when a price hits a threshold and scheduled restocking.

At the same time, Amazon blocks external shopping agents from its site and has taken legal action against Perplexity over access. That tension is the strategic story of agentic commerce. Marketplaces want the agent to be theirs. Your own store is the one place where you decide what an agent is allowed to see.

If you sell on marketplaces as well, the listing hygiene that helps Rufus is the same work covered in Amazon listing optimisation for India and our Amazon and Flipkart listing service.

What Indian Brands Should Actually Do

The work splits into three layers. Most brands have done none of it.

1. Make your product data machine readable

This is the unglamorous foundation and it is where most stores fail. Adobe scored retail pages with its AI content visibility checker and found product pages averaged 66 percent machine readable, below homepages at 75 percent. Product pages are exactly the pages an agent needs to read.

Concretely, that means:

• Product schema on every product page with price, currency, availability, GTIN or SKU, brand, and aggregate rating. Our schema markup guide covers the setup, and the schema generator will produce the markup.
• Prices and stock status that are correct in the HTML, not only after JavaScript runs. An agent that reads a stale or empty price will either skip you or quote a wrong number.
• Real specifications in structured fields rather than buried in a paragraph of marketing copy. Size, material, weight, colour, warranty, compatibility.
• Shipping and returns stated plainly on the product page. Agents compare on total cost and risk, not just sticker price.

2. Answer the questions people ask agents

Shoppers do not ask agents for a keyword. They ask for a situation. Best running shoes for flat feet under a certain budget. A gift for a colleague who likes coffee. A CRM that works for a small team in India.

Your product and category pages should answer those framings directly, in short extractable blocks near the top. This is the same discipline as getting cited by ChatGPT, and the mechanics of what engines look for are covered in how AI engines decide which sources to cite.

3. Build the outside signals

An agent recommending a product is making a small bet on your behalf. It leans on things it did not have to take your word for.

• Genuine reviews with volume and recency, on your site and off it.
• Consistent brand details everywhere, so the agent knows the Growzai on one site is the Growzai on another. This is entity SEO.
• Third-party coverage. Directories, comparison pages, and industry write-ups create the consensus an agent looks for before it names you.
• An llms.txt file that gives AI systems a clean summary of what you sell.

The Decision Nobody Has Made Yet

At some point you will need a position on whether agents are welcome on your site. Blocking them protects your merchandising and your ad-driven upsells. Allowing them puts you in consideration sets you would otherwise never appear in.

For most Indian D2C brands the answer right now is straightforward: allow the crawlers, because you are not big enough for a shopper to seek you out by name. You need to be found. Check your robots file and confirm you are not accidentally blocking AI crawlers, which is one of the most common and most expensive mistakes we find in audits.

Mistakes to Avoid

• Treating this as a payments project. Payments are the last step, and the protocols are not settled. Product data is the step that pays off today.
• Optimising the homepage and ignoring product pages, which is exactly the gap Adobe measured.
• Letting prices, stock and delivery timelines go stale. An agent that quotes you wrong once will deprioritise you.
• Assuming marketplace listings cover you. Amazon blocks outside agents, so your marketplace presence does not make you visible to ChatGPT or Perplexity.
• Waiting for UAP or ACP to be finalised before doing anything. None of the preparation work depends on which protocol wins.

Frequently Asked Questions

Can AI agents already buy things in India?

Not through a settled, regulated framework yet. NPCI is developing the Unified Agent Protocol for agent-led UPI transactions in consultation with industry, and it needs RBI approval before launch. What is happening in India today is agent-assisted discovery, where people use AI to research and then buy through normal channels.

Should I block AI agents from my store?

Most Indian D2C and ecommerce brands should not. Blocking removes you from consideration in the assistants where buying decisions are increasingly made. Large marketplaces block because they already own the demand. You probably do not.

Is this different from SEO?

It builds on SEO rather than replacing it. An agent can only recommend a page it can crawl, index and parse. The difference is that agents read structured data and specifications where humans read persuasion, so the work shifts toward clean, complete, machine readable product information. See our SEO vs AEO comparison.

How do I know if AI is already sending me traffic?

Segment your analytics by referrer to isolate visits from assistant domains, and watch for branded queries that mention your product category in a conversational way. Our guide on measuring AI search visibility walks through the setup.

What should I fix first?

Product schema and accurate server-rendered prices and stock. It is the cheapest fix with the largest effect, because everything else an agent does depends on being able to read those fields correctly.

The fastest way to see where you stand is to run your site through our free AEO Checker. If you want your product pages, schema and entity signals audited properly for agent readiness, book a free strategy call with Growzai and we will show you exactly what an AI agent can and cannot see on your store today.