AI Shopping Assistants and Marketplace Selling in India

Executive summary

Amazon.in, Flipkart and Google now offer AI shopping assistants to Indian shoppers. These assistants answer questions and shortlist products using listing details, customer reviews, Q&A and web information. For marketplace sellers, the quality of that information now decides whether a product is recommended, not only whether it ranks for a keyword.

Key findings

• Every major channel has an assistant. Flipkart launched Flippi in October 2023 and announced its replacement, SLAP, in January 2026. Rufus reached Amazon.in in beta in August 2024. Google upgraded shopping in AI Mode and Gemini in April 2026, and in September 2026 began testing Flipkart purchases inside both.
• The best public data is from the US, not India. Adobe's analysis of more than 1 trillion US retail site visits shows traffic from generative AI sources up 4,700% year on year in July 2025. Those visitors stayed 32% longer, but were still 23% less likely to convert than other visitors. We found no comparable published data for Indian marketplaces.
• Widely shared statistics need care. Vendor conversion claims and market-size forecasts vary widely; this paper relies on neither.
• Indian rules limit the shortcuts. The review standard IS 19000:2022, the CCPA's dark-patterns guidelines and the DPDP Act and Rules all apply to how sellers handle reviews, offers and customer data.

Five actions for sellers

1. Complete every attribute and rewrite titles and bullets on the top 20 products so they answer real buyer questions.
2. Answer all open customer questions, and fold repeated ones into the listing.
3. Fix the product, packaging or listing issue behind the top 3 review and return complaints.
4. Each month, ask the assistants 10 real buyer questions and record whether your product is shortlisted.
5. Keep product facts the same on every channel and on your own website, and keep top sellers in stock.

1. What is changing

Indian shoppers can now ask an AI assistant what to buy inside the apps where Pexeta's clients sell. Every major channel has launched one since 2023:

The shift that matters for sellers: the shopper increasingly describes a need ("a non-sticky kadai under ₹1,500 for gas and induction") instead of typing keywords and scrolling a results page. The assistant reads listings, reviews and Q&A, and shortlists a few products. A product that the assistant cannot understand, or does not trust, is less likely to make that shortlist.

Amazon has said publicly how much it expects from this. Amazon CEO Andy Jassy has said Rufus is expected to add $10 billion in annual sales, as NewsBytes reported; that is a company expectation for its global business, not an Indian figure.

2. What the data says, and what it doesn't

The most reliable public numbers come from Adobe's analysis of more than 1 trillion visits to US retail sites. They show fast growth from a small base, and shoppers who research more but did not yet convert better. We found no comparable published data for Indian marketplaces.

Sources: Adobe, Digital Commerce 360, Retail Brew.

Three cautions before using numbers you see elsewhere:

• Vendor statistics are not benchmarks. The widely shared "12.3% vs 3.1% conversion with AI chat" comes from a chatbot vendor reporting on its own merchants (Rep AI).
• Market-size forecasts disagree. One widely copied estimate puts the AI-in-ecommerce market at $22.6 billion by 2032 (SellersCommerce, via WebProNews); another puts it at $11.1 billion by 2032 (Meticulous Research). This paper uses neither.
• US data is not Indian data. None of the figures above measures Amazon.in, Flipkart or quick commerce apps. Sellers should measure their own traffic and conversion instead (Section 4).

3. Why it matters for marketplace sellers

An AI assistant can only recommend what it can read. Amazon says Rufus is trained on Amazon's product catalog, customer reviews, community Q&As and information from across the web, and answers questions such as "is this jacket machine washable?" from listing details, reviews and Q&As. For a seller, that turns four things already on the listing into what the assistant uses to decide:

Two things follow for Indian marketplace sellers:

• Keyword search still matters, but it is no longer the only route to the product. A shopper asking for "a lunch box that keeps food warm for school, leak-proof" gets a shortlist, not a results page. Products whose listings state those facts clearly are easier for an assistant to match.
• The same listing data feeds several assistants. Flipkart's assistant, Rufus and Google's AI Mode all draw on product information and reviews. Cleaning up a catalogue once helps on every channel, including quick commerce apps, where listings are short and attributes carry more weight.

4. Seller playbook: the next 90 days

Most of this work improves today's keyword search and conversion as well, so it pays off whether or not AI assistants grow as fast as forecast.

What not to do: don't buy reviews or offer rewards for positive ones (Section 5), and don't stuff listings with keywords an assistant will read as noise.

5. Risks and rules in India

Three Indian rules apply to how sellers collect reviews, design offers and handle customer data. None of them is specific to AI, but each one limits shortcuts sellers might be tempted to take.

Two practical risks sit alongside the rules:

• Assistants make mistakes. Amazon itself says the technology won't always get it exactly right. Check what assistants say about your products, and correct the listing when they get a fact wrong.
• Platforms control the assistant. Sellers don't get a dashboard showing how often an assistant recommends their product. Measure it yourself with the monthly test in Section 4.

This section is a summary, not legal advice. Check specific obligations with a lawyer.

6. How Pexeta helps

Pexeta has worked with marketplace sellers since 2012 and runs each step of the playbook above as a managed service for brands on Amazon, Flipkart, FirstCry, Myntra and quick commerce apps.