Do You Need a Separate Product Feed for AI Shopping Assistants? Get Ready for the 2027 Season

A shopper asks ChatGPT for the best waterproof hiking boots under $200. The answer shows six products with prices, ratings and a "buy" link for each. Your boots are a better fit for the question, sell at $179 and have 2,000 reviews on your site. They are not in the answer. Your merchandising team asks the obvious question: do we need to build some new kind of feed for this? It is the right question, and Cognizo helps ecommerce teams answer it with data, because Cognizo tracks which products and brands AI assistants recommend for the shopping questions buyers actually ask.

The short answer is: one more feed, probably; a whole new feed strategy, no. Each AI shopping assistant reads product data from its own pipe. Google reads Merchant Center. Microsoft Copilot reads Microsoft Merchant Center. ChatGPT has its own product feed spec. Amazon's assistant reads your Amazon catalog. Most of them also read your product pages and what the rest of the web says about you. The brands that win treat this as one master product data set with several outputs, not as five separate projects. And because a feed only gets you into the candidate pool, they track whether their products actually appear in AI answers. That is where Cognizo gives ecommerce teams three things no feed dashboard can: daily visibility into ChatGPT Shopping and the other AI engines, the sources each assistant trusts for your category, and the content and technical fixes that win the recommendation.

This guide covers what each assistant reads as of November 2026, where a separate feed is genuinely needed, what to put in it, and how to tell whether any of it is working.

The short answer by assistant

The rest of this article explains each row, then the work that applies to all of them. One thing the table cannot tell you is whether your products actually show up once the data is in. That depends on the prompt, the engine and the day.

Where Cognizo helps: Cognizo tracks your shopping prompts across ChatGPT (including ChatGPT Shopping), Google AI Overviews and AI Mode, Gemini, Perplexity and the other major engines every day, so one dashboard shows which assistants recommend your products and which do not. Instead of five feed consoles that only report "accepted," you get one view of where you actually appear.

Why AI shopping assistants matter for product data now

AI-referred shopping traffic is still smaller than search, but it is growing fast and it converts. Adobe's analysis of more than 1 trillion visits to US retail sites found that AI-referred traffic grew 138% year over year in May 2026, and that those visitors converted 54% better than visitors from other sources, according to Digital Commerce 360. A year earlier they converted at about half the non-AI rate.

The assistants are also where product data gets judged most directly. A search results page shows ten links and lets the shopper decide. An AI assistant picks a handful of products, explains why, and often shows a price and a merchant. If your price, availability or key attributes are missing or wrong in the data the assistant reads, you are not in the shortlist, no matter how strong your SEO is. Our data study on whether a #1 Google ranking gets you into AI answers shows how weak that link can be.

Retail also dominates what these assistants cite. In Ahrefs' September 2026 analysis of Copilot citations, Amazon and Walmart alone took 29.4% of the citation share among the 50 most-cited sites, as we covered in our Copilot citation sources study. If marketplaces are the default source, your own product data has to be complete enough to compete with them. Where Cognizo helps: Cognizo's source classification and page type classification show, for each shopping prompt, whether an engine is citing Amazon, Walmart, a retailer, a review site or your own product page. Competitors and regions are unlimited on every plan, so you can benchmark against every brand and marketplace that sells in your category, in every market you serve.

ChatGPT shopping: the one place a separate feed is clearly on the table

ChatGPT is the assistant where a new feed is most likely to be worth building, because OpenAI publishes its own format and does not ask you to reuse a Google feed.

How ChatGPT picks products. OpenAI's help center says that when deciding which products to show, ChatGPT considers structured metadata from first-party and third-party providers, such as price and description, along with other third-party content. Merchants are ranked on factors including availability, price, quality, and whether the seller makes the item or is its primary seller. OpenAI states that product results are chosen independently and are not ads. In other words, ChatGPT can show your products without any feed from you, using data it gets elsewhere, but you do not control how accurate or current that data is.

What the feed does. OpenAI invites merchants to send a direct product feed so that "ChatGPT always reflects the most up-to-date information" about their products. Its product feed specification lists nine required fields: item_id, title, description, url, brand, seller_name, image_url, availability and price. Files can be JSONL, CSV or TSV in UTF-8, uploaded through OpenAI's file upload process. OpenAI does not set a fixed refresh schedule; it asks merchants to update the feed whenever price or availability changes, such as when a sale starts or an item sells out.

Two flags matter. enable_search controls whether products are eligible to appear and defaults to true. enable_checkout opts into in-chat checkout, defaults to false, and only works if checkout is separately enabled for your integration. Eligibility does not guarantee display.

Shopify stores are already in. According to OpenAI, product data for Shopify merchants is integrated into ChatGPT through Shopify Catalog, and "no additional work is required from individual merchants." If you sell on Shopify, a separate ChatGPT feed is optional. Your job is making sure the Shopify product data itself is complete.

Feeds became the main input in July 2026. The clearest evidence that feeds matter comes from an industry analysis of 1,757,723 ChatGPT Shopping prompt runs, reported by Search Engine Journal. Before July 10, 2026, only 8.26% of ChatGPT's shopping recommendations came from feed-integrated products. On July 10 that share jumped to 61.54%, and by September 3 about 65% of recommendations came through feed retrieval. Recommendations also concentrated: the top 10 stores' share of references rose from 22.5% to 41.8%, and the number of unique merchants referenced fell from 13,524 to 10,607. The analysis links the shift to the release of GPT-5.6 on July 9; OpenAI has not confirmed that link. Either way, for ChatGPT a complete, current feed is no longer a nice-to-have.

Google's shopping data matters here too. A study of 43,000 ChatGPT carousel products across 10 retail categories, published in Search Engine Land in March 2026, found that 83% of them matched a product in Google Shopping's top 40 organic results for the same search, against just under 11% for Bing. Of those matches, 60% came from Google Shopping's top 10. In other words, a strong Google Merchant Center feed and good Google Shopping rankings help your ChatGPT visibility as well.

Checkout changed in March 2026. OpenAI launched Instant Checkout in 2025, then scaled it back. In a March 24, 2026 post, it said merchants can use their own checkout while OpenAI focuses on product discovery, explaining that the first version "did not offer the level of flexibility" it wanted, Modern Retail reported. Walmart's Daniel Danker told Modern Retail that conversion was three times lower for products sold directly inside the chatbot than for those that clicked through to Walmart's site. For most merchants, that makes the feed a discovery tool first: its job is to get accurate products into answers and send shoppers to your store. Where Cognizo helps: Cognizo's ChatGPT Shopping tracking shows whether that is happening: which of your products appear in ChatGPT's shopping results for the prompts you care about, which competitors appear when you do not, and how that changes day by day. That matters because ChatGPT's sources can change overnight: in the analysis Search Engine Journal reported, 450 of 687 tracked stores lost at least a third of their ChatGPT Shopping visibility between July 7 to 9 and July 10 to 12, 2026, while 67 gained that much. With daily tracking, a drop like that shows up the day it happens, not in next month's traffic report.

Do not confuse the organic feed with the ads feed. ChatGPT Ads also accepts product feeds for shopping campaigns, through CSV or TXT uploads, a hosted URL or SFTP. According to OpenAI's help center, products supplied through those ad campaigns are eligible for ads during the beta and do not appear in organic ChatGPT conversations. Uploading your catalog to the ads tool will not make your products show up in regular shopping answers; organic discovery runs through Shopify Catalog, the direct product feed or third-party data. Cognizo tracks organic ChatGPT Shopping results separately from ad placements, so the two do not get mixed up in reporting.

Our guide on optimizing your brand for ChatGPT covers the non-feed side, and ChatGPT Ads explains how paid placements sit next to these organic product results.

Google AI Mode and Gemini: your Merchant Center feed already does the job

For Google, you do not need a separate AI feed. Product results in AI Mode, Gemini and Google's shopping surfaces come from the same Merchant Center data and Shopping Graph that power Google Shopping. Practitioner guides consistently note that AI Mode draws on free product listings, so having free listings turned on in Merchant Center is the entry ticket.

What Google added in 2026 sits on top of that feed rather than replacing it. In a January 11, 2026 announcement, Google introduced:

• The Universal Commerce Protocol (UCP), an open standard for agentic commerce co-developed with Shopify, Etsy, Wayfair, Target and Walmart. UCP powers a checkout feature on eligible Google product listings in AI Mode and the Gemini app, letting shoppers buy from eligible US retailers while they research. Retailers remain the seller of record.
• Dozens of new Merchant Center attributes built for conversational surfaces, covering things like answers to common product questions, compatible accessories and substitutes. Google says they "complement retailers' existing data feeds" and will roll out to a small group of retailers first.
• Business Agent, a branded AI agent that lets shoppers chat with a retailer directly in Search, activated through Merchant Center.

The practical work for Google is therefore feed quality, not a new feed. Fill optional attributes you have been skipping (material, size system, compatibility, product highlights), keep price and availability in sync, and add the conversational attributes as they become available to you. Merchant Center tells you whether items are approved, not whether AI Mode recommends them.

Where Cognizo helps: Cognizo tracks AI Overviews and AI Mode as separate engines, so you can see whether better attributes actually move your products into Google's AI answers, and which competitor or marketplace page Google cites when you are missing. Our guide to tracking your brand in Google AI Mode covers how to see whether your products show up.

Microsoft Copilot: Microsoft Merchant Center and UCP feeds

Copilot reads product data from Microsoft Merchant Center, plus the web. If you already import your Google Merchant Center feed into Microsoft Merchant Center for Microsoft Shopping, you are most of the way there.

In April 2026, Microsoft added support for UCP feeds in Merchant Center, generally available in the US, The Keyword reported. Any merchant can use them, and merchants already on UCP can appear in Copilot without building a separate integration. Shopify merchants can also sync their catalogs to Copilot in real time through Microsoft's commerce API, without custom feeds. At the same time, Copilot Checkout expanded from desktop to Microsoft's mobile apps; Microsoft says it covers more than 500,000 US merchants, who keep merchant-of-record status.

So for Copilot: no new feed if you are already in Microsoft Merchant Center or on Shopify. If you are in neither, importing your Google feed into Microsoft Merchant Center is a small job with a direct payoff. Copilot leans heavily on retail sources: in Ahrefs' data, retail and marketplace domains held about half of Copilot's top-50 citation share. Where Cognizo helps: Cognizo shows whether Copilot cites your own store, a marketplace listing of your product or a reseller for each tracked prompt. If a marketplace is taking the citation for your own product, you know to strengthen your product pages and feed rather than spend more on ads.

Perplexity: join the Merchant Program

Perplexity launched shopping and a free Merchant Program in November 2024, describing the program in its launch post as a simple way for merchants to share product data, with payment integrations for in-app purchases on select products. Modern Retail's launch coverage noted that merchants supplying detailed product data may be recommended more often.

Perplexity has not published a public feed field specification, and third-party guides disagree about the exact format and eligibility. Some describe Google Merchant Center-format feeds; others describe an application form rather than a self-serve upload. Shopify product data reaches Perplexity through platform integrations. Independent of the program, Perplexity's crawler reads your product pages, so structured data and crawler access matter here as much as any feed.

For Perplexity: apply to the Merchant Program, expect to reuse your Google-format data, and do not wait on it. Make your product pages readable in the meantime. Where Cognizo helps: Perplexity cites its sources openly, and Cognizo captures them. Its citation data shows which review sites, roundups and product pages Perplexity trusts for your category, and which of those mention competitors but not you, which gives your PR and outreach team a ready-made target list.

Amazon's assistant: your Amazon listings are the feed

Amazon retired the Rufus name in May 2026 and launched Alexa for Shopping, combining Rufus's product knowledge with Alexa+'s personalization. It is available to US customers across the Amazon app, website and Echo Show devices, according to coverage of the launch. There is no external feed to submit: the assistant works from your Amazon listings, reviews, pricing and ads. Coverage of the launch reports that Amazon Sponsored Products campaigns are integrated into the assistant's conversations.

For Amazon sellers, the "feed" work is listing quality: complete attributes, clear bullet points that answer common questions, accurate variations, competitive and stable pricing, and steady review volume. Amazon's "Buy for Me" feature can also surface eligible products from brands' own websites, which is another reason to keep your own site's product data clean even if you sell on Amazon. Where Cognizo helps: Outside Amazon, the other assistants cite Amazon listings heavily. Cognizo shows whether ChatGPT, Copilot or Google's AI features point shoppers to your Amazon listing, your own store or a reseller, so you can decide where to invest: in your Amazon content, your own site or both.

So, do you need a separate product feed?

Put together, the answer for most ecommerce teams looks like this:

• You need one master product data set that is complete, accurate and updated as often as prices and stock change. Every AI channel draws from some version of it.
• You probably need one additional output: an OpenAI feed, unless you are on Shopify, where Shopify Catalog already covers ChatGPT.
• You do not need a separate Google or Microsoft AI feed. Merchant Center feeds are the source. Spend the effort on attribute completeness and the new conversational attributes instead.
• You should apply to Perplexity's Merchant Program and expect to reuse existing data.
• On Amazon, your listings are the feed.

The mistake to avoid is building five separate, hand-maintained feeds that drift apart. When ChatGPT shows one price, Google another and your site a third, assistants that compare sources lose confidence in your data, and shoppers lose trust. Generate every channel's feed from the same source of truth.

Where Cognizo helps: Feed rules can change without notice. When ChatGPT shifted to feed-based picks in July 2026, about two out of three stores in the tracked sample lost at least a third of their visibility within days, and only teams watching daily would have seen it as it happened. Cognizo re-runs your shopping prompts every day across the major AI engines, so a sudden drop on one engine shows up on the day it happens, along with the competitor or marketplace that took your place, which tells you where to act first.

What to put in your product data, whatever the channel

AI assistants answer questions, so the data that wins is the data that answers them. Beyond the required fields, these make the biggest difference:

Our guide on optimizing product descriptions for AI shopping assistants goes deeper on writing descriptions that match how people ask. Where Cognizo helps: Cognizo's Prompt Volumes module shows the questions shoppers actually ask AI in your category, with volume estimates, so you know which attributes and use cases your titles and descriptions need to cover. Its Content Optimization module then drafts AI-optimized buying guides and product content around those questions, and the content agent can publish them automatically.

The feed is only half of it: your product pages and the web

A feed gets a product into the candidate pool. What the assistant says about it, and whether it recommends it, often comes from elsewhere.

Product pages. ChatGPT, Perplexity and Copilot all read the web. Product pages with clean HTML, Product and Offer structured data, visible prices, specifications and reviews give assistants a second, consistent source. Our guide to schema markup for AI search covers the markup, and optimizing your site for AI crawlers covers making sure OAI-SearchBot, PerplexityBot and Bingbot can reach those pages.

Third-party sources. When a shopper asks "what are the best running shoes for flat feet," the assistant leans on reviews, roundups, forums and videos to decide what to recommend, then uses feeds to fill in price and availability. OpenAI's help center lists "other third-party content" among its inputs, and its review summaries are generated from public reviews. A brand that is absent from the roundups and comparison pages an assistant trusts will struggle to be recommended, however good its feed is. Our research on third-party AI citations covers how to earn those mentions, and Cognizo's citation data lists the review sites and roundups each engine cites for your prompts, including the ones that mention competitors but not you.

Consistency. The same product should have the same name, price and key facts in your feed, on your page and on the marketplaces that list it. When the facts disagree, an assistant has less reason to prefer your listing over a reseller's.

Where Cognizo helps: Cognizo's technical site audits flag crawl and rendering issues that keep AI assistants away from your product pages, and its AI crawler analytics show which bots, such as OpenAI's and Perplexity's, actually visit your site and which pages they request. Together with the citation data, that tells you whether a missing recommendation is a data problem, a page problem or a third-party coverage problem.

How to tell whether your feed is working

Submitting a feed does not tell you whether your products appear in AI answers. Feed tools report whether items were accepted. They do not show whether ChatGPT recommended your boots or a competitor's when a shopper asked. That is the measurement gap most ecommerce teams have today.

Cognizo closes it. Its Answer Engine Insights module tracks the shopping questions your customers ask across the major AI engines every day and records which brands and products are mentioned, which pages are cited, and how your visibility and share of voice change. It includes ChatGPT Shopping tracking, so you can see how your products show up in ChatGPT's shopping results, not only in text answers. Source and page type classification show whether an engine is leaning on marketplaces, review sites, retailers or your own pages for each prompt, and the YouTube and Reddit breakdown shows how much community content is shaping recommendations.

That turns feed work into something you can measure:

1. Set a baseline before changing feeds. Track 50 to 150 shopping prompts that match your catalog, such as "best waterproof hiking boots under $200" or "hiking boots for wide feet," with your main competitors.
2. Ship the feed and data changes. Add the OpenAI feed if you need one, fill missing attributes, fix price sync.
3. Watch the daily trend. Check whether your products start appearing in ChatGPT Shopping and AI Mode for the tracked prompts, and whether a competitor or marketplace listing drops out.
4. Read what is cited when you lose. If a review site or a marketplace listing wins, the fix is coverage or content, not another feed field.
5. Connect it to revenue. Cognizo's AI Traffic Analytics shows AI referral traffic and crawler activity next to the visibility data, so you can see whether more AI visibility brings more visits. Our guide to tracking AI traffic in GA4 covers the analytics setup.

When the gap is content rather than data, Cognizo's Content Optimization module drafts AI-optimized buying guides and comparison content aimed at the prompts where you are missing, and its technical site audits flag crawl issues that keep assistants away from your product pages. For paid placements alongside organic results, the AI Ads module tracks competitor ads and targeting opportunities.

Ecommerce is where this measurement pays off most directly. For a real example, read how Hat Club turned AI visibility into 20x revenue with Cognizo. Our roundup of the best AI visibility tools for ecommerce compares the options.

A practical rollout plan

Week 1: audit the source of truth. Pull your master product data and check required fields, GTINs, attribute fill rates and price accuracy against your site. Fix the gaps at the source, not in each feed. In parallel, start tracking your top shopping prompts in Cognizo so you have a baseline.

Week 2: cover the channels. Confirm free listings are on in Google Merchant Center and that your feed is imported into Microsoft Merchant Center. If you are not on Shopify, build the OpenAI feed from the same source data and submit it. Apply to Perplexity's Merchant Program. Note the date in Cognizo so the before-and-after comparison is clean.

Week 3: fix the pages. Add or correct Product and Offer structured data, make sure prices and stock on the page match the feeds, and confirm AI crawlers are allowed in robots.txt and at your CDN. Run Cognizo's technical site audit and check its crawler analytics to confirm AI bots are reaching your product pages.

Week 4 onward: measure and iterate. Watch daily visibility for your tracked prompts per engine. Where you still do not appear, check what the assistant cites instead and decide whether the fix is data, content or third-party coverage. Add Google's conversational attributes as they become available to you. Use Cognizo's content agent to fill content gaps with buying guides and comparison pages, and its outreach targets to go after the review sites that cite competitors.

Common mistakes to avoid

• Treating the feed as the whole strategy. A perfect feed will not win a recommendation if every review roundup lists your competitors. Cognizo's citation data shows which roundups those are.
• Maintaining separate feeds by hand. They drift. Generate every output from one source.
• Letting prices go stale. OpenAI asks for updates whenever price or availability changes. Stale data is worse than missing data, because it can put a wrong price in front of a buyer.
• Blocking AI crawlers by accident. Some CDN bot settings block AI user agents by default, which hides your product pages from assistants that read the web.
• Assuming the ads feed covers organic results. Product feeds uploaded to ChatGPT Ads campaigns serve ads only during the beta, not organic shopping answers.
• Building for in-chat checkout first. Checkout programs are changing quickly, and OpenAI scaled its version back within a year. Discovery is where the reliable value is today.
• Not measuring. If you only track feed approval, you cannot tell whether any of the work put your products in front of shoppers. Track the shopping prompts themselves, per engine, with a platform like Cognizo.

FAQ

Do I need a separate product feed for ChatGPT shopping?

Only if you want to submit product data directly and you are not on Shopify. OpenAI publishes its own product feed specification with nine required fields, including title, price, availability and product URL. Shopify merchants are already integrated through Shopify Catalog, with no extra work required, according to OpenAI. ChatGPT can also show products using third-party data, but feeds now drive most recommendations: according to an analysis reported by Search Engine Journal, about 65% of ChatGPT Shopping recommendations came through feed retrieval by September 2026, up from 8% before July 10. Cognizo's ChatGPT Shopping tracking shows whether your products appear after you submit it.

How do I get my products into ChatGPT shopping?

Make sure your product data is complete and accurate, then either rely on Shopify Catalog if you use Shopify or submit a product feed through OpenAI's merchant program. Keep prices and availability updated whenever they change. Your product pages, structured data and third-party reviews also influence which products ChatGPT recommends. Cognizo's ChatGPT Shopping tracking shows whether your products start appearing and which competitors you are up against.

Does Google AI Mode need a separate product feed?

No. Google's AI Mode and Gemini shopping results draw on Merchant Center data and the Shopping Graph. Keep free listings on, fill optional attributes, and add Google's new conversational attributes as they become available. Checkout in AI Mode runs on the Universal Commerce Protocol for eligible US retailers. To see whether your products are recommended in AI Mode, track the shopping prompts there separately, which Cognizo does daily.

Do I need a feed for Microsoft Copilot shopping?

Copilot uses Microsoft Merchant Center, which supports UCP feeds in the US. If you already import your Google feed into Microsoft Merchant Center, or you sell on Shopify, you do not need anything new. Copilot Checkout is live for US merchants on desktop and mobile. Cognizo shows whether Copilot cites your store or a marketplace listing for your products.

Does Perplexity have a product feed?

Perplexity runs a free Merchant Program for sharing product data but has not published a public field specification. Shopify stores reach Perplexity through platform integrations. Perplexity also reads product pages directly, so structured data and crawler access matter regardless of the program. Cognizo's citation data shows which pages Perplexity trusts for your category.

What happened to Amazon Rufus?

Amazon replaced Rufus with Alexa for Shopping in May 2026. It works from your Amazon listings, reviews and pricing, so there is no external feed to submit. Listing completeness and pricing are what matter.

How do AI shopping assistants choose which products to show?

They combine structured product data, such as price, availability and attributes, with what reviews and other web sources say about the product. OpenAI says ChatGPT ranks merchants on availability, price, quality and whether the seller makes or primarily sells the item. Feeds get you into the candidate pool; reviews, roundups and clear product pages influence whether you are recommended. Cognizo shows which of those sources each engine cites for your category, so you know where to focus.

How can I tell if my products appear in AI shopping answers?

Feed dashboards only show whether items were accepted. To see whether your products are recommended, you need prompt-level tracking across AI engines. Cognizo tracks your shopping prompts daily, including ChatGPT Shopping, and shows which brands and products appear, which sources are cited and how your visibility changes over time.