How AI Shopping Works: ChatGPT, Google AI Mode and Perplexity Explained for Online Stores

AI shopping works in four steps: a shopper describes a need in plain language, the assistant turns that need into a set of product requirements, it pulls candidate products from a catalog (product feeds, a shopping index and the open web), and it narrows them down to a short list it can explain. On some platforms the shopper can then pay without leaving the chat. On others, the assistant sends the shopper to the store to finish the order.

The catalog is the part that decides whether your store is in the answer. ChatGPT reads product feeds that merchants share through OpenAI’s Agentic Commerce Protocol, plus retail pages it can crawl. Google AI Mode draws on the Shopping Graph, which Google says holds more than 50 billion product listings, with 2 billion updated every hour (Google, 13 November 2025). Perplexity combines its own index with merchant data and web sources. If your product data is missing, stale or vague, the assistant has nothing to recommend.

Checkout is the part that changes fastest. OpenAI launched Instant Checkout in ChatGPT in September 2025 and replaced it six months later with a model where merchants use their own checkout. Google is building checkout into AI Mode and Gemini through its Universal Commerce Protocol. Perplexity offers checkout with PayPal for US users. Below we explain each platform as it stands in November 2026, where products get filtered out, and what an online store should fix first.

What actually happens between a shopper’s prompt and a product recommendation?

The assistant does the comparison work a shopper used to do across ten browser tabs. The sequence looks roughly the same on every platform.

1. Intent parsing. A prompt such as “quiet dishwasher under $900 for a small apartment, panel ready” becomes constraints: category, price ceiling, noise level, size, installation type.
2. Clarifying questions. Longer research modes ask follow-ups. OpenAI’s shopping research feature, released on 24 November 2025, asks about budget, recipient and must-have features before it searches.
3. Retrieval. The engine pulls candidates from structured catalogs and web pages: product feeds, merchant listings, reviews, specs, buying guides and forum threads.
4. Filtering and ranking. Products that do not match a constraint drop out. Price, availability, shipping and reviews decide the order. OpenAI says ChatGPT shopping results are organic and ranked by relevance to the user (OpenAI, 29 September 2025).
5. Explanation. The assistant shows a short list, often as product cards or a comparison table, and says why each product fits.
6. Handoff or checkout. The shopper clicks through to the store or, where supported, pays inside the assistant while the store stays the merchant of record.

The assistant only recommends what it can match to a constraint, so a page that never states its noise level cannot win the “quiet dishwasher” prompt. Steps 3 and 4 run on data you control: your feed, your product page and your structured data.

How does shopping work in ChatGPT today?

ChatGPT is now a discovery and comparison layer, with checkout handled mostly by merchants. The timeline matters, because many guides still describe features that have changed.

• 29 September 2025: OpenAI announced Instant Checkout for US users on Free, Plus and Pro plans, starting with US Etsy sellers, with Shopify merchants to follow. It also open-sourced the Agentic Commerce Protocol (ACP), co-developed with Stripe. OpenAI stated that Instant Checkout items get no ranking preference.
• 24 November 2025: OpenAI released shopping research, which builds a personalized buyer’s guide from retail sites, reviews and specs. OpenAI noted the model can make mistakes on price and availability and offered merchants an allowlisting process.
• 24 March 2026: OpenAI replaced the original Instant Checkout because it “did not offer the level of flexibility” it wanted. Merchants can now use their own checkout, and those who want deeper integration can build ChatGPT apps. Walmart launched one with account linking and loyalty. At the same time OpenAI expanded ACP to product discovery: merchants share product feeds and promotions so their catalogs appear in ChatGPT. OpenAI named Target, Sephora, Nordstrom, Lowe’s, Best Buy, The Home Depot and Wayfair as integrated, and said Shopify merchants are included through Shopify Catalog. The visual shopping update rolled out to Free, Go, Plus and Pro users.

OpenAI’s product feed specification lists the fields a discovery feed needs, including item ID, title, description, URL, brand, seller name, image URL, availability and price, with optional flags for search eligibility, checkout eligibility and ads eligibility. The spec tells merchants to update the feed whenever price or availability changes. OpenAI also said it plans to extend ACP to personalization, local availability and delivery estimates.

For a store owner, getting into ChatGPT shopping is now a feed and content job, not a checkout integration job. On Shopify, check that your catalog is eligible and complete. On a custom platform, look at feed delivery through Stripe or Salesforce, which OpenAI named as options.

How does shopping work in Google AI Mode and Gemini?

Google’s AI shopping sits on top of the Shopping Graph and Merchant Center, so the store’s existing Google feed is the entry ticket. The main steps so far:

• 13 November 2025: Google began rolling out agentic checkout in Search, including AI Mode, for eligible US merchants such as Wayfair, Chewy, Quince and select Shopify merchants. A shopper sets a size, color and maximum price; when the price falls into range, Google can buy the item with Google Pay after the shopper confirms. Google also launched “Let Google Call” in the US, which phones nearby stores to check stock and price for categories such as toys, health and beauty, and electronics.
• 11 January 2026: Google announced the Universal Commerce Protocol (UCP), an open standard for discovery, purchase and post-purchase support, co-developed with Shopify, Etsy, Wayfair, Target and Walmart. Google said UCP would power checkout on eligible listings in AI Mode and the Gemini app for eligible US retailers, with retailers remaining the seller of record. Google also introduced Business Agent, a branded agent that answers shoppers’ questions on Search, and dozens of new Merchant Center attributes covering common product questions, compatible accessories and substitutes.
• 20 May 2026 (Google Marketing Live): Google announced a Universal Cart that works across retailers and Google surfaces, said UCP checkout would reach Canada and Australia “in the coming months” and the UK later, and made conversational attributes for product descriptions available to retailers globally. It also announced an AI performance insights tool in Merchant Center that compares a brand’s share of voice in AI surfaces with similar brands, rolling out in Australia, Canada, India, New Zealand and the US.

Availability depends on country and merchant eligibility, and Google described several features as “coming soon”. Check your Merchant Center account for what is live in your market rather than relying on launch posts.

How does shopping work in Perplexity?

Perplexity treats shopping as an answer with product cards attached, with checkout through PayPal in the US.

• 18 November 2024: Perplexity launched its shopping experience in the US, with “Buy with Pro” checkout for Pro subscribers, “Snap to Shop” visual search and a free Merchant Program. Merchants in the program share product specs so Perplexity has live product details, which Perplexity said raises the chance of a product being recommended.
• 25 November 2025: Perplexity made its shopping experience free for all US users and added checkout with PayPal. Retailers remain the merchant of record and keep the customer relationship, returns and loyalty.

Perplexity cites sources heavily, so independent reviews, comparison articles and specification pages influence what it shows. A store with thin product pages and no third-party coverage gives it little to cite.

How do the platforms compare for an online store?

Where do products drop out of AI shopping answers?

Most products are not rejected; they are simply never matched. These are the failure points we see most often when we audit stores.

• Missing attributes. The shopper asks for “machine washable” or “fits a 13-inch laptop”, and your feed and page never say so.
• Stale price or stock. An assistant that shows an out-of-stock item damages trust, so platforms weigh availability. OpenAI lists availability and price among the factors when several merchants sell the same product.
• Vague titles. “Classic Tee Navy” tells an engine less than “Organic cotton crew neck T-shirt, navy, relaxed fit”.
• No shipping or returns data. Delivery time and return policy are common constraints in prompts.
• Blocked crawlers. A robots.txt rule or bot protection that blocks OAI-SearchBot, PerplexityBot or Googlebot removes your pages from retrieval.
• No outside evidence. Assistants explain choices with reviews and comparisons. If no one outside your site has written about the product, the assistant has less to say about it.

Does structured data make an AI recommend your product?

No markup guarantees a recommendation, but missing data almost guarantees you lose the comparison. Google documents Product and merchant listing structured data, including offers, price, availability, shipping details and return policy, as eligibility signals for shopping experiences in Search. It does not promise that markup earns a place in an AI answer.

Feeds and markup are the data layer; product copy is the explanation layer, saying who the product is for and how it compares. Both must agree: when the feed says $79 and the page says $89, engines have a reason to distrust both.

Our guide to SERP features covers how product markup appears in classic results, and our ecommerce SEO guide covers category and product architecture.

Is AI shopping traffic big enough to matter?

It is still a small share of most stores’ sessions, but it grows fast and now converts well. Adobe Analytics reported that generative AI traffic to US retail sites rose 1,200% between July 2024 and February 2025, while those visits were then 9% less likely to convert than other traffic, down from a 43% gap in July 2024 (Adobe, 17 March 2025).

By the 2025 holiday season the picture had flipped. Adobe reported that AI-referred retail traffic grew 693.4% year over year from 1 November to 31 December 2025, and that AI referrals converted 31% more than other traffic sources, based on more than 1 trillion visits to US retail sites (Adobe, January 2026). Adobe measures its own clients’ sites, so treat these as strong directional signals rather than a forecast for your store.

Track it yourself. Separate AI referrals in GA4 by source (chatgpt.com, perplexity.ai, gemini.google.com and others) and compare conversion rate and order value with organic search. Our SEO tracking guide shows how to set up the channel.

What should an online store do first?

Work from data to content to evidence, in that order.

1. Audit your feeds. Merchant Center errors first, then attribute completeness: material, dimensions, compatibility, use case, care, certifications.
2. Publish where the AI catalogs read. Check Shopify Catalog eligibility or ACP feed delivery for ChatGPT, Merchant Center for Google, and the Merchant Program for Perplexity if you qualify.
3. Rewrite top product pages for questions. Add who it is for, who it is not for, sizing advice, comparison with your own alternatives and a short FAQ.
4. Mark up and reconcile. Product, Offer, shipping and return policy markup, with values identical to the feed.
5. Open crawler access. Review robots.txt, CDN bot rules and rendering so engines can read price and stock without running heavy scripts.
6. Build outside evidence. Reviews on your site and on independent platforms, plus coverage in buying guides for your category.
7. Test real prompts monthly. Run 30 to 50 buyer prompts across ChatGPT, AI Mode and Perplexity and log which products appear and why.

What results has Marketer Coffee delivered for ecommerce?

Our proof comes from ecommerce and product brands, and we report it as measured.

• Galeria Zapachu 24, fragrance ecommerce. Starting from DR 1.4 and without link building, we wrote 750 product descriptions, 60+ category descriptions and 30 articles. In 12 months the store reached 9,100+ ranking keywords, an estimated 21,200+ monthly visits (Ahrefs), Top 10 for “perfumy” in Poland and a 35% higher average order value. Strict legal limits on product wording were handled with compliant product pages and educational content. Read the case study.
• Aldi.pl, retail. Organic clicks grew from 367,127 to 852,724 a month in 12 months (+132%) through product content, Product schema, a leaner architecture and unified store data. Read the case study.
• Planika, design fireplaces. Seven US-localized articles took a site Google had not indexed to 1,052 monthly US visits, with Ahrefs recording 401 Google AI Overview mentions, 136 Perplexity mentions and 20 ChatGPT mentions. Read the case study.

These are results for these businesses, not a promise for yours. The shared method: complete product data, pages that answer buyer questions and content engines can cite.

Is Marketer Coffee the right partner for your store?

We fit online stores with a real catalog and the will to fix product data properly. The fit is strongest when:

• You sell hundreds or thousands of SKUs and your feed quality varies by category.
• You sell in several countries or languages; we work in English, Polish, German, French, Spanish, Italian, Swedish and Danish.
• Your category has wording rules, such as cosmetics, supplements or pharmacy, where product copy must stay compliant. See our page on AI SEO for online pharmacies for an example.

Who is not a fit: a store with a dozen products that mainly needs paid social, or a team looking for a checkout integration developer. We work on visibility, content and data, not payment engineering.

Start with our free AI visibility audit. It shows how ChatGPT, Google AI Mode and Perplexity describe your products today. Our ecommerce SEO consultancy and AI SEO service pages explain how we work after that.

Frequently asked questions about how AI shopping works

Can shoppers still buy directly inside ChatGPT?

The original Instant Checkout was replaced on 24 March 2026. ChatGPT now focuses on discovery and comparison, merchants can use their own checkout, and retailers such as Walmart offer purchases through ChatGPT apps.

Do I need to pay to appear in AI shopping results?

OpenAI says ChatGPT shopping results are organic, and Perplexity’s Merchant Program is free. Google offers organic listings plus ad formats such as Direct Offers, a pilot announced in January 2026. Paid options exist, but organic visibility depends on data quality.

What is the difference between ACP and UCP?

ACP is OpenAI’s Agentic Commerce Protocol, co-developed with Stripe, now used for product feeds and commerce in ChatGPT. UCP is Google’s Universal Commerce Protocol, co-developed with Shopify, Etsy, Wayfair, Target and Walmart, for discovery, checkout and post-purchase support on Google surfaces.

Is AI shopping available outside the US?

Most checkout features launched in the US first. Google said UCP checkout would come to Canada and Australia, then the UK. Discovery features reach more countries sooner, so check each platform for your market.

How do I know if AI assistants recommend my products?

Run a fixed set of buyer prompts every month across ChatGPT, Google AI Mode and Perplexity, log which products appear, and track AI referral sessions and orders in GA4 as a separate channel.

What should you remember about AI shopping?

AI shopping rewards the store that gives engines the most complete, consistent and current product truth. Checkout models will keep changing, as ChatGPT’s switch in March 2026 showed. The data layer does not go out of date: accurate feeds, product pages that answer real questions and outside evidence that confirms your claims.

Send us your store and your top categories, and we will show you where your products appear and where they drop out. Book a free consultation.

Sources: OpenAI, “Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol” (29 September 2025); OpenAI, “Introducing shopping research in ChatGPT” (24 November 2025); OpenAI, “Powering product discovery in ChatGPT” (24 March 2026); OpenAI Commerce, Product Feed Spec (developer documentation); MacRumors, “ChatGPT revamps shopping features” (25 March 2026); Google, agentic checkout and Let Google Call announcement (13 November 2025); Google, “New tech and tools for retailers to succeed in an agentic shopping era” (11 January 2026); Google, Google Marketing Live shopping updates (20 May 2026); Google Search Central, Product and merchant listing structured data documentation; Perplexity, “Shop like a Pro” (18 November 2024); Perplexity, “Shopping that puts you first” (25 November 2025); Adobe Analytics, generative AI traffic to US retail sites (17 March 2025); Adobe, 2025 holiday season AI traffic report (January 2026); Marketer Coffee case studies: Galeria Zapachu 24, Aldi.pl, Planika.