AI and Your Online Store: What Changes for Shops in India

AI assistants now shortlist shops before a customer opens a website. Ask ChatGPT, Gemini, Perplexity or Google's AI Mode where to order groceries, and the engine reads the open web, picks a few names and cites a few pages. Whether your shop is named depends on how clearly and consistently it is described across the web, and on whether your products, prices and policies are readable by a machine.

We tested this on ourselves. On 5 October 2026 we asked two AI engines eight questions a buyer of store software might ask. Across 16 answers, Ordrio was named zero times. Below: what we did, what the engines cited instead, and how to run the same test on your shop in 30 minutes.

What has changed, and what we know for sure

AI answers sit above the links. Google began rolling out AI Mode to everyone in India, in English and with no sign-up, on 8 July 2025, and added Hindi on 8 September 2025. ChatGPT, Gemini and Perplexity answer shopping questions with a written answer and a few sources.

Fewer people click through. Pew Research Center studied the browsing of 900 US adults in March 2025. When a Google search showed an AI summary, people clicked a traditional result on 8% of visits, against 15% when there was no summary. They clicked a link inside the summary on 1% of visits. Ahrefs compared 300,000 keywords in December 2023 and December 2025 and found that an AI Overview correlates with a 58% lower click rate for the top result on informational searches. Both are US data on mostly informational searches. We found no comparable numbers for shopping searches in India: the direction is known, the size for your shop is not.

The answer is a shortlist. A results page shows ten links; an AI answer shows a handful of names. Microsoft puts it this way: in AI search, ranking still happens, but it is less about ordering whole pages and more about which pieces of content earn a place in the final answer.

How AI engines choose what to cite

None of the companies publishes its ranking rules. What they do publish:

Two points follow. An engine can only quote what it can read, so a price list in a photo or a menu in a PDF is close to invisible. And an engine that finds conflicting facts about you, such as two phone numbers, may repeat the wrong one.

Our test: does AI know Ordrio?

We sell store software, so we asked what our buyers ask. The next section adapts the method for a shop.

Method

• Date: 5 October 2026, one run.
• Engines: Perplexity (sonar model) and Gemini (3.5 flash lite model), both with web search on, called through DataForSEO's LLM responses API with a 600 token output limit per answer. We also called ChatGPT (o4-mini); those calls returned no text, so they are not counted.
• Prompts: eight questions, written as a buyer would type them, covering India, the Gulf and Australia.
• Scoring: whether the answer named Ordrio, which other platforms it named, and which domains it cited.

We then asked Perplexity a direct question: "What is Ordrio?"

Results

What it tells us

The engines know Ordrio when asked by name, so our own pages are readable and correct. But for open questions they built the shortlist from comparison lists and directories, and we are not on the lists they read. The gap was our absence from other people's pages.

Its limits

Two engines, eight prompts, one day, one set of settings, through an API rather than a logged-in app. Answers vary between runs and users, so a second run could name different platforms. It measures a software brand, not a shop. It shows a pattern worth checking, not a ranking.

Run the same test on your shop in 30 minutes

Steps

1. Write six prompts (5 minutes). Two open questions a new customer would ask, two specific needs, one comparison and one about your shop by name. Use the area and city where you deliver. Examples for three kinds of shop are below.
2. Pick four engines (2 minutes): ChatGPT with search on, Gemini, Perplexity and Google AI Mode. Use a private browser window where you can, so your own history does not shape the answer.
3. Ask each prompt on each engine (15 minutes). That is 24 answers. Paste each prompt exactly as written and do not ask follow-ups.
4. Log every answer in the template below (5 minutes, as you go).
5. Count and read (3 minutes), using the guide after the template.
6. Repeat monthly with the same prompts, so you compare like with like.

Prompts to adapt

Replace the names in square brackets with your own.

Use the four prompts in your row and add one more open question and one more specific need.

Log template

Copy these columns into a spreadsheet, one row per answer.

How to read the result

• Named, with correct facts. Your description is working. Keep every listing in step with it.
• Named, with wrong facts. An engine has found a stale or wrong page. Open the cited sources, find the old phone number, hours or address, and fix or update it.
• Not named, but other shops are. Look at the sources cited for those shops. That list of sites is your to-do list: they are the pages the engines read for your area.
• The name question gets a vague or wrong answer. Your shop has a thin footprint on the open web. Start with steps 1 and 2 of the action list below.
• Different engines give different shortlists. Normal. Start with the sources that more than one engine cites.

What to do: a ranked action list

Ranked by likely help for the effort. The order is our judgement from what the engines publish and what our test showed, not a proven ranking.

1. Describe your shop the same way everywhere

Use one spelling of the shop name, one address, one phone number and one set of hours on your website, maps listing, directories, delivery apps and social pages. If you moved or changed numbers, update the old listings too.

2. Publish the facts a customer asks about, as text

One page, in plain HTML text: delivery areas, hours and slots, minimum order, delivery fee, payment methods and the return and refund policy. Write each as a short sentence that stands on its own, such as "We deliver to Kadri, Bejai and Kodialbail in Mangaluru between 7 am and 9 pm." That is the sentence an engine can quote.

3. Name products the way people ask for them

"Toor dal, unpolished, 1 kg" can be matched to a question; "TD-UP-1" cannot. Put the item, the variety and the unit in the product name, and other names customers use in the description. Our grocery store setup guide has a product data sheet you can copy.

4. Add structured product data

Structured data is a small block of code that states a product's name, price, currency and stock in a fixed format. Google uses Product and Offer markup to make a product eligible for its merchant listings, such as product snippets and popular products. Microsoft says schema helps search engines and AI systems understand a page. Google also says no special schema is needed for AI Overviews, so this helps shopping features rather than guaranteeing an AI mention.

A developer can paste this on a product page and fill in the real values:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Toor dal, unpolished, 1 kg",
  "sku": "TOOR-UNP-1KG",
  "image": "https://yourshop.example/images/toor-dal-1kg.jpg",
  "brand": { "@type": "Brand", "name": "Your Shop" },
  "offers": {
    "@type": "Offer",
    "price": "165.00",
    "priceCurrency": "INR",
    "availability": "https://schema.org/InStock",
    "url": "https://yourshop.example/products/toor-dal-unpolished-1kg"
  }
}
</script>

Keep price and availability in step with the shop; markup that says "in stock" for an item you cannot deliver is worse than none. Check a page with Google's Rich Results Test.

5. Collect genuine reviews

Ask every customer after delivery, with a direct link to your maps listing. Do not buy reviews or offer a discount for one: Google's Maps content policy removes paid or incentivised reviews, and India's IS 19000:2022 standard sets out how platforms should verify and moderate online reviews.

6. Get on the lists the engines already cite

Your test log shows which directories and comparison pages the engines read for your area. Check your shop is on them with the same facts as step 1, and ask the rest to include you. Our test suggests the shortlist comes from these pages; that a listing leads to a mention is likely, not proven.

7. Let the search crawlers in

robots.txt is a text file at the root of your site that tells crawlers what they may read. Search crawlers and training crawlers are separate, so you can allow one and refuse the other:

# Search crawlers: allow them, or you cannot be linked in their answers
User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

# Training crawlers: your choice (see the notes below on Google-Extended)
User-agent: GPTBot
Disallow: /

User-agent: Google-Extended
Disallow: /

Three honest notes. Blocking GPTBot tells OpenAI not to train on your content. Blocking Google-Extended stops Gemini using your pages for training and for grounding its answers, so it can cost you mentions in Gemini; Google says it does not affect inclusion or ranking in Google Search. If being named in Gemini matters to you, leave Google-Extended allowed. ChatGPT-User and Perplexity-User, which fetch pages when a person asks, may not follow robots.txt, by their own documentation. And User-agent: * followed by Disallow: / blocks every crawler, Google included; check your file for it.

AI shopping agents: what has launched and what a store needs

An agent goes further than an answer: it finds a product and places the order for the customer.

OpenAI, Google and Perplexity have all launched agent checkout, nearly all of it US only, and OpenAI scaled back Instant Checkout within six months because, in its words, it "did not offer the level of flexibility that we aspire to provide". For a shop in India, agent ordering is something to prepare for, not to sell through today.

What an agent needs from a store

OpenAI asks merchants for a structured product feed with up to date price and availability. Google asks for complete product feeds, return policies and business contact information in Merchant Center. Put together, this is the readiness checklist:

A shop that gets these right for human shoppers is most of the way there for agents.

What not to do

• Keyword stuffing. Repeating "best grocery store in Mangaluru" down a page is a named spam practice in Google's policies.
• Fake or paid reviews. They break Google's Maps policy and go against India's review standard.
• Facts in images and PDFs. A price list photo or a menu PDF is hard for an engine to read reliably. Put the same facts in text.
• Hiding answers in collapsed tabs. Microsoft notes that AI systems may not render hidden content. Keep delivery and return facts visible.
• Blocking every bot by accident. A blanket Disallow: /, or a security setting that blocks unknown crawlers, can remove you from search and AI answers alike.

Known, likely and speculation

What we do not know yet

• How engines weigh sources. No company publishes it. Our test shows what was cited, not why.
• How much answers vary. One day and one set of settings cannot show how stable a shortlist is. Monthly reruns will.
• When agent checkout reaches Indian shops. The ChatGPT payments pilot was announced in October 2025. By August 2026 Razorpay said around 50 brands were in pilots, but we found no public date for it opening to every merchant. Separately, OpenAI stepped back from its US Instant Checkout feature in March 2026.
• Whether Indian shopping searches lose clicks at the US rate. The click studies are US data.

Where Ordrio fits

Nadi gives your business its own website and mobile app, with payments, orders and delivery in one place.

Pili AI is the AI assistant built into Nadi, included in every plan. On your store, it is a shopping assistant for your own customers, by text or voice: it suggests products from your own catalogue, fills the cart and takes the shopper through checkout inside the chat, and nothing is bought until the shopper confirms and pays. Catalog Studio turns a product list or billing export into a clean catalogue that you review before anything is created, which helps with step 3 above. Pili AI works in English and Arabic. Ordrio does not use your store's data to train AI.

What we do not claim. We have not measured whether stores on Ordrio are named by AI engines, and we do not say they are. No platform can make an engine choose you; the action list is yours to do, on any platform. See pricing for what each plan includes.

Sign up for free

Book a demo

Sources

1. Google Search Central, AI features and your website, updated 10 December 2025.
2. Google Search Central, Google's common crawlers: Google-Extended, updated 14 July 2026.
3. Google Search Central, Merchant listing structured data, accessed 5 October 2026.
4. Google Search Central, Spam policies for Google web search, updated 28 August 2026.
5. Google India blog, AI Mode in Google Search rolling out in India, 8 July 2025.
6. Google India blog, AI Mode now available in Hindi, 8 September 2025.
7. Google, New tech and tools for retailers to succeed in an agentic shopping era, 11 January 2026.
8. Google Merchant Center Help, How to onboard to the UCP integration hub in Merchant Center, accessed 5 October 2026.
9. Google for Developers, Universal Commerce Protocol FAQ, accessed 5 October 2026.
10. Google, Maps user contributed content policy, accessed 5 October 2026.
11. OpenAI, Overview of OpenAI crawlers, accessed 5 October 2026.
12. OpenAI, Agentic Commerce Protocol, accessed 5 October 2026.
13. TechCrunch, OpenAI takes on Google, Amazon with new agentic shopping system, 29 September 2025.
14. TechCrunch, India pilots AI chatbot-led e-commerce with ChatGPT, Gemini, Claude in the mix, 9 October 2025.
15. TechCrunch, OpenAI's plans to make ChatGPT more like Amazon aren't going so well, 24 March 2026.
16. Perplexity, Perplexity crawlers, accessed 5 October 2026.
17. PayPal Newsroom, PayPal and Perplexity Launch Instant Buy Ahead of Black Friday, 25 November 2025.
18. Microsoft Advertising, Optimizing your content for inclusion in AI search answers, 8 October 2025.
19. Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results, 22 July 2025.
20. Ahrefs, Update: AI Overviews Reduce Clicks by 58%, 4 February 2026.
21. Press Information Bureau, Government of India, Centre launches framework for safeguarding consumers from fake and deceptive reviews in e-commerce (IS 19000:2022), 21 November 2022.
22. Inc42, Can Razorpay turn ChatGPT into India's next commerce channel?, 14 August 2026.
23. Ordrio, AI visibility test, run 5 October 2026 through the DataForSEO LLM responses API (Perplexity sonar and Gemini 3.5 flash lite, web search on). Prompts and results as listed in this article.