AI in ecommerce is used in five places today. They are product discovery through AI assistants and Google's AI Overviews, on-site search and recommendations, support chat, content and catalogue work, and pricing and operations. Each one changes a different step between a shopper's question and a delivered order.
This guide is for founders and marketers at consumer brands and online stores in India. Each use comes with what the AI does, one real example with its source, and a first step for a small team. An AI assistant here means a chat tool, such as ChatGPT or Gemini, that writes an answer in place of a list of links. An AI Overview is the AI-written summary Google shows above some search results.
Most guides to AI for ecommerce stop at chatbots and product descriptions. Discovery is newer, and harder to see from inside your own store. Shoppers ask an AI what to buy, and the answer names a few brands.
Where is AI used in ecommerce today?
AI is used at five points on the shopper's path: before the store, inside it, after a question, behind the product page and behind the order. The table gives each use, what it changes and a first step that costs little or nothing.
The role of AI in ecommerce differs at each point. In discovery, another company's AI decides whether your brand is named. In the other four, you pick a tool and switch it on.
| AI use | What it changes | First step for a small brand |
|---|---|---|
| Discovery in AI answers | A shopper gets a short list of brands and products from an assistant before visiting any store | Ask your 5 main buying questions in ChatGPT, Gemini, Perplexity and Google, and note which brands are named |
| On-site search and recommendations | Search understands everyday phrases, and product pages suggest related items | Open your search report, find the searches that returned nothing, and add those words to your products |
| Support chat | Common questions get an instant answer or a drafted reply | Write clear shipping, returns and cash-on-delivery policies, because the AI answers from them |
| Content and catalogue work | A first draft of a description or title takes minutes | Draft with AI, then check every fact against the product before you publish |
| Pricing and operations | Orders are scored for fraud or return risk before they ship | Review the orders flagged as high risk before dispatch |
How do shoppers use AI assistants to decide what to buy?
Shoppers type a full buying question, such as "best sunscreen for oily skin under ₹500". The assistant replies with a few named brands and a reason for each. Many replies link to the pages they drew from, and ChatGPT can also show products as cards inside the answer. OpenAI says shopping in ChatGPT is live for users in the US today and will expand to more regions (OpenAI's merchants page, checked 7 Oct 2026).
The shortlist forms before the shopper sees your store, your ad or your marketplace listing.
What OpenAI says about ChatGPT's product results
OpenAI's help page on shopping says a product appears when ChatGPT judges it relevant to what the shopper wants. In its example, ChatGPT may consider the options available, price, reviews and ease of use. OpenAI says product results are chosen independently by ChatGPT and are not ads (OpenAI help centre, checked 7 Oct 2026).
OpenAI says ChatGPT considers structured metadata from first-party and third-party providers, such as price and product description, next to other third-party content. Structured metadata means product facts in a format a machine can read. Merchants are ranked on factors like availability, price, quality and whether they are the maker or primary seller (OpenAI help centre, checked 7 Oct 2026). Our guide to ChatGPT shopping covers each part.
Ads are separate from product results. OpenAI says ChatGPT Ads have launched in India (OpenAI announcement, published 23 September 2026, checked 7 Oct 2026). Its help page says ads may appear below an answer for users on the Free and Go plans, labelled as sponsored (OpenAI help centre, checked 8 Oct 2026).
What Google says about AI Overviews
Google says AI Overviews and AI Mode, its conversational search option, may run several related searches for one question. It calls this query fan-out. To appear as a supporting link, a page must be indexed, meaning stored in Google's database of pages. It must also be eligible to show with a snippet, the preview text under a result.
Google adds that there are no extra technical requirements and no special optimisation (Google Search Central, checked 7 Oct 2026). Every site also has a Search generative AI control in Search Console, Google's free tool for site owners, under Settings. Sites are included by default, and a site its owner excludes does not appear in these features (Google Search Console Help, checked 8 Oct 2026). A store Google cannot crawl and index cannot be linked from an AI Overview.
Which websites the answers cite in India
AI shopping answers lean on other people's pages as much as on brand sites. Depra AI's study collected 480 answers from ChatGPT, Gemini and Perplexity in India on 14 August 2026 and counted 4,640 citations. The most cited site was youtube.com with 534 citations, then reddit.com with 191, nykaa.com with 176 and magicpin.in with 133.
So videos, forum threads and marketplace pages shape what an answer says about your category. The full list is in which websites AI shopping answers cite in India.
Where the purchase happens
In ChatGPT today, OpenAI says shoppers discover and compare products in the chat and complete the purchase on the merchant's own website or app. Its merchants page says OpenAI is moving away from a standalone Instant Checkout (OpenAI's merchants page, checked 7 Oct 2026), yet its help article says the option may show for some eligible products (OpenAI help centre, checked 7 Oct 2026). Agentic commerce explains what this means for a store.
How does a brand find out whether AI recommends it?
Ask the assistants your buyers' questions yourself, several times, and count how often your brand is named. One answer proves little, because the same question returns different brands on different runs.
- 1. Write 5 buying questions. Use the words a shopper would use and leave your brand name out, for example "best running shoes under ₹3,000".
- 2. Ask each one in a private window. Use ChatGPT with web search on, Gemini, Perplexity and Google. A private window keeps your own history out of the answer.
- 3. Record four things. Whether you are named, your place in the list, the competitors named and the websites cited.
- 4. Repeat on another day. Two passes give you 40 answers, enough to spot a clear gap.
- 5. Count. Your visibility is the answers that named you divided by all answers. The AI visibility checker page has the full method.
What a small brand can do with the result
- Let the AI search crawlers in. OpenAI says sites that opt out of OAI-SearchBot, its search crawler, are not shown in ChatGPT search answers (OpenAI crawler docs, checked 7 Oct 2026). To check, open yourstore.com/robots.txt in a browser. It is the file that tells bots what they may read. If it has a User-agent: OAI-SearchBot line, the rules under it apply. If not, the rules under User-agent: * apply. A Disallow: / line there blocks the whole site. Our AI crawlers list names the other bots.
- Put product facts in text. Price, size, ingredients and who the product suits should be readable on the page. Ecommerce SEO covers the page and structured data work, and Shopify SEO covers it for Shopify stores.
- Get named where the answers look. If the cited pages name your rivals and skip you, that is the gap to close. Why ChatGPT doesn't mention your brand gives a test for each cause.
Tracking it with Depra AI
A manual check is a snapshot. Depra AI asks your buying questions to ChatGPT with web search, Gemini and Google AI Overviews every day, and to Perplexity every week. It asks from inside India or another market you pick, and stores each answer with the brands named and the websites cited. For ChatGPT it also records the products shown. Product visibility ranks brands by the share of shopping answers that show their products.
OpenAI says shopping in ChatGPT is live for users in the US today and will expand to more regions (OpenAI's merchants page, checked 7 Oct 2026). If you track questions for India, your Shopping report may show few or no products until ChatGPT shows products there. Depra AI records products whenever ChatGPT shows them in a tracked answer.
Its limits, stated plainly: only ChatGPT returns product cards, so the other three engines are tracked for brand mentions and sources. Depra AI does not track Google AI Mode or Amazon Rufus. It does not write content, and it never changes your store. For a Shopify store, setup needs no Shopify app and no access token, as Depra AI for Shopify explains. To weigh other options, see the best AI visibility tools for D2C brands. D2C means direct-to-consumer.
How is AI used in on-site search and recommendations?
AI search matches what a shopper means, not only the exact words typed. AI recommendations suggest products that are often bought or viewed together. Both work after the shopper has arrived.
Shopify's semantic search is one example. Its help centre describes it as AI-powered search that uses related words, concepts and categories to widen results. It is included for stores on the Grow, Advanced or Plus plan with fewer than 200,000 products, with nothing to switch on (Shopify Help Center, checked 8 Oct 2026). A Shopify changelog entry dated 24 June 2024 gave the example of a shopper who types "something to wear in the summer" instead of "shorts" (Shopify changelog, checked 7 Oct 2026). Shopify's free Search & Discovery app adds synonym groups, filters, related products and search analytics (Shopify App Store, checked 8 Oct 2026).
Marketplaces do the same at a larger scale. Amazon's assistant is Amazon Rufus (renamed Alexa for Shopping in May 2026), according to Amazon's own page (About Amazon, checked 7 Oct 2026). In an announcement first published on 20 November 2024, Amazon India said Rufus was available to all India customers in the Amazon Shopping app and on desktop. For a question about one product, it said Rufus answers from listing details, customer reviews and community Q&As (About Amazon India, checked 7 Oct 2026).
What a small brand can do: find the searches in your store's search report that returned nothing. On Shopify, the Search & Discovery app has search analytics. Add those words to product titles, descriptions or synonym groups. On marketplaces, fill in every listing field and answer buyer questions, since the assistant reads them.
How is AI used in ecommerce customer support?
AI answers common support questions from your policies and order data, or drafts a reply for staff to check. This is often called conversational AI for ecommerce. It suits repeat questions: where an order is, how to return an item, whether cash on delivery is offered.
Shopify Inbox, Shopify's customer chat app, is an example of the drafting kind. Shopify says its AI-generated suggested replies help staff compose responses faster and are built from your store information. Staff must read each reply and edit it where needed before sending. The feature works only in English. Its on and off switch is in Shopify Inbox: click the settings icon, then Message preferences (Shopify Help Center, checked 7 Oct 2026).
WhatsApp sets rules for support on the WhatsApp Business Platform, which businesses use through software instead of the phone app. Its business policy allows automated replies there inside the 24-hour customer service window. That window opens and resets with each customer message. The policy also requires prompt, clear and direct escalation paths, such as a human agent, a phone number or an email address (WhatsApp business policy, last updated 23 Sep 2026, checked 7 Oct 2026).
What a small brand can do: write down your 20 most common questions and the exact answers. Publish your shipping, return and cash-on-delivery terms. An AI can only answer from what you have written, and a wrong refund promise costs more than a slow reply.
How is AI used for product content and catalogue work?
AI writes first drafts of product descriptions, titles and image text. It also fills gaps across a catalogue, the full list of products a store sells.
Shopify's assistant, Sidekick, takes details you provide, such as a product title and keywords, and suggests a description. Shopify advises giving at least a title and 2 features or keywords, and reading everything closely for accuracy before publishing. To use it, open a product in the Shopify admin on a desktop computer and click the Generate text icon in the description box (Shopify Help Center, checked 7 Oct 2026).
What a small brand can do: let AI draft, then add the facts only you know. Those are ingredients, materials, sizes, care and who the product suits. Check every claim. Google and AI assistants read the same facts when they describe your product, so a vague description costs you twice.
How is AI used in pricing and operations?
Behind the order, AI scores orders for risk, forecasts demand and suggests prices. One use built for India is predicting which cash-on-delivery orders will come back undelivered. The trade calls this RTO, short for return to origin.
Delhivery, a logistics company, offers one such tool. It describes RTO Predictor as an AI-led product built to reduce RTOs for D2C brands and ecommerce companies. The tool takes the buyer's mobile number and address, the product, the shipment value and the payment method. It returns a score from 0.0 to 1.0, where a higher score means a likelier return (Delhivery help centre, page updated 3 Apr 2025, checked 8 Oct 2026). The same page lists a charge of ₹2.49 plus GST for each cash-on-delivery order it scores as low or high risk. It is switched on in the Delhivery One dashboard under Services > Value Added Service > RTO Predictor.
Shopify does something similar for fraud on online card orders. It says its fraud recommendations come from machine learning trained on past transactions across all Shopify stores (Shopify Help Center, checked 8 Oct 2026). The same page says recommendations need the Grow plan or higher, or Shopify Payments. Orders at medium or high risk are flagged on the Orders page, and each order has an Order risk section.
Pricing tools change prices as demand, stock and rival prices move. What a small brand can do: start with risk, because the saving is easy to see. Call or message buyers on high-risk orders, or offer a prepaid option. With automated pricing, set a floor price and approve each change yourself at first.
Which AI use should a small Indian brand start with?
Start with discovery, because checking it is free and it shows where you stand. Then fix product facts, since every other use reads them. Add tools one at a time, where you can see the result within a month.
Keep a person on anything about money, health or refunds, because AI tools make mistakes. For where AI answers sit among your other channels, read D2C marketing in India.
- 1. Check discovery. Run the 5-question check above.
- 2. Fix product facts. Make price, size, ingredients and policies readable on every product page.
- 3. Draft with AI, edit by hand. Use it for descriptions and reply drafts, and read each one first.
- 4. Add risk scoring. It is worth the setup once cash-on-delivery returns are a visible cost.
- 5. Measure again. Repeat the discovery check every month with the same questions.