How To Get Your Products Into ChatGPT, Google AI Mode, and Perplexity: A Merchant’s Guide

Preparing your product catalog for ChatGPT, Google AI Mode, and Perplexity channels

ChatGPT shopping sounded far-fetched even a handful of years ago. Now? There’s been a 200% growth where agentic search as a first step in the product journey is concerned. On top of that, U.S. retail sites saw a 393% YoY increase in AI traffic during Q1 2026.

In other words: People are increasingly comfortable asking their favorite robot to help them shop.

Agentic commerce is taking off, with the likes of ChatGPT, Google AI Mode, Perplexity,Microsoft Copilot, and others leading the charge. Instead of acting as mere answer engines, these platforms are serving up product suggestions and even helping people complete purchases.

Traditional ecommerce isn’t going away anytime soon, but ignoring these agentic channels is to stay hidden to the more than one-billion people using ChatGPT every week alone.

But, getting into Google AI Mode and others isn’t as easy as asking ChatGPT nicely. You need product data that’s complete, optimized, enriched, and more. Now, what’s that look like?

Read on to learn how agentic AI decides what to surface, and more importantly: How you can increase the chances it’s your products getting shouted from the rooftops. (Or phone screens.)

Psst. This new agentic hotness doesn’t mean you should ignore AI shopping assistants, either. Check out our complete guide to learn how you can utilize AI shopping assistants to the fullest.

How the agentic shopping journey works across AI channels

How does agentic shopping work? In four stages: a shopper describes what they need in plain language, the assistant retrieves and compares candidate products using live merchant data, the shopper checks out (most often on your own site) and the order lands in your store. AI shopping agents handle discovery and evaluation; you still own the transaction and the customer.

Stage 1: Discovery: the shopper asks, the assistant retrieves.

Instead of typing keywords, the shopper describes a need: “a quiet tower fan under $150 for a small bedroom.” The assistant retrieves candidates from merchant feeds, the Shopping Graph, or a web crawl; the mix differs by platform, and the next section breaks that down. The core shift: an agent can’t be swayed by slick design or ad spend. It reads your product data. Steps 1 and 3 below cover opening the door to crawlers and supplying the feed.

Stage 2: Evaluation: the agent compares live product data.

From those candidates, the agent compares attributes, real-time price and availability, shipping terms, and third-party reviews, then presents a short list — or a single pick. Here’s the part that stings: a stale feed or a feed-to-page price mismatch removes your product at this stage, not at checkout. Steps 2 and 3 cover the schema and feed hygiene that keep you in the running.

Stage 3: Checkout: three paths, only one of them fully in-assistant.

Once the shopper commits, checkout takes one of three paths:

1. A redirect to your own checkout: the default on ChatGPT and most journeys today (the ACP standard, covered in Step 4, still runs underneath).

2. In-assistant checkout inside Google AI Mode or Gemini via Google’s UCP (AP2 handles the payment authorization): Early Access and rolling out, not generally available.

3. Buy with Pro inside Perplexity: one-click checkout on PayPal rails for US Pro subscribers, in beta.

As of September 2026, end-to-end in-chat purchasing isn’t live across all three surfaces; Buy with Pro is the only fully in-assistant path today. Step 4 walks you through each program.

Stage 4: The order lands in your store.

Whichever path the shopper takes, you remain the merchant of record. The order, payment record, fulfillment, returns, and customer relationship all land — and stay — in your store. The assistant found the shopper; the business is still yours. (The measurement section below covers tracing these orders back to their AI source.)

Here’s the journey at a glance, channel by channel:

How ChatGPT, Google AI Mode, and Perplexity choose products

While all three major AI-powered answer engines surface products in a similar way where the user is concerned, behind the scenes is another story. Each of the three goes about their selection process in a different way (because of course they do), so it’s important you take a comprehensive approach.

Also worth calling out: Paid placements aren’t really a thing in the truest sense of the term. While ChatGPT and Google AI Mode allow businesses to pay for ad space, paid results are clearly labeled and shown separately from organic query results.

And another thing worth noting: All three answer engines, as well as many others, rely on a quality product feed. That is, an optimized and enriched file that contains structured product data for all of your products. (Okay, that’s it, promise.)

ChatGPT: merchant product feeds plus web crawl.

With ChatGPT shopping, the merchant-submitted product feed is the ultimate source of truth. To do this, you can apply through OpenAI’s merchant portal, followed by delivering a feed that meets the OpeNAI product feed spec.

Beyond this, the OAI-SearchBot will crawl the web, filling in any gaps it can while verifying details. So, while you have some chance of showing up in ChatGPT without submitting a feed, submitting a feed is your chance to stay in control of the data.

Google AI Mode: the Shopping Graph and Merchant Center.

Google AI Mode exists right inside the traditional Google search experience, making it an easy way for people to engage with your brand in a quick, meaningful way.

To make all this AI magic happen, Google AI Mode pulls product info from the Shopping Graph. The Shopping Graph is built on the foundation of your Google Merchant Center feeds and free listings, not your Shopping ads. (Again: Ad spend doesn’t buy AI Mode placement.)

As long as you have your Merchant Center up and running, the most important elements are:

• Complete attributes for every product

• Updated price and availability

• Consistency between your feed and site across the board

As long as your products are already in the Merchant Center with free listings enabled, you’re eligible and off to a great start. If free listings are off, no amount of ad budget or magic will get you mentioned.

Perplexity: merchant feeds, the PayPal network, and cited reviews.

Perplexity shopping is powered by a free merchant program, which utilizes Google Shopping-formatted feeds. On top of this, if you’re already accepting PayPal, you can surface through the PayPal merchant network, thanks to their partnership with Perplexity.

Unlike the other two engines mentioned, where your own feed is the be-all and end-all, Perplexity cross-checks product claims against independent third-party reviews. So, if you’re not staying on top of your social proof game, now’s the time.

How the three platforms compare.

Step 1: Open your site to AI crawlers

Step one is inarguably the most important, as nothing else in this guide works if an AI crawler can’t reach your product pages. Fortunately, the process for opening your site to AI crawlers is fairly straightforward, as long as you know which bots to allow.

Which bots to allow, and what each one controls.

There are a few different bots to allow, each playing a similar role for ChatGPT, Google AI Mode, and Perplexity. Keep in mind, allowing a bot for one doesn’t give the other AI engines access, so it’s important you give all three platforms access if you want to maximize your AI search presence.

The three OpenAI user agents do different jobs, and merchants should decide on each separately:

• OAI-SearchBot: controls ChatGPT search and shopping visibility.

• GPTBot: collects data for model training only. Blocking it does not affect ChatGPT shopping.

• ChatGPT-User: handles user-triggered page fetches when someone asks ChatGPT to look at a specific URL.

For Perplexity, allow PerplexityBot for citations. For Google, standard Googlebot indexing plus snippet eligibility governs AI Mode inclusion.

Worth noting: blocking Google-Extended does not remove your site from AI Mode or AI Overviews. Google-Extended only opts your content out of Gemini training and grounding.

Use this copy-ready robots.txt block to streamline your efforts:

# ChatGPT search and shopping visibility User-agent: OAI-SearchBot Allow: / # ChatGPT user-triggered fetches User-agent: ChatGPT-User Allow: / # OpenAI model training (your call — does not affect shopping visibility) User-agent: GPTBot Disallow: / # Perplexity citations User-agent: PerplexityBot Allow: / # Google Search, AI Overviews, and AI Mode User-agent: Googlebot Allow: /

Check your CDN and bot protection.

Robots.txt is a request, not a wall. If Cloudflare, Akamai, or your WAF challenges these user agents, the crawlers get a CAPTCHA page instead of your product data, and your robots.txt edits change nothing.

To avoid this happening, you need to:

• Allowlist OAI-SearchBot, ChatGPT-User, PerplexityBot, and Googlebot in your bot-management rules.

• Verify with server logs: look for successful 200 responses from those user agents on product URLs.

• Expect roughly 24 hours for OpenAI search-eligibility changes to propagate after you open access.

There’s an important caveat with Perplexity: Perplexity-User, which fetches pages on a user’s request, generally ignores robots.txt by design, and Cloudflare has documented undeclared Perplexity crawling. Treat robots.txt as permission signaling. If you want proper enforcement, it lives at the WAF level.

Where llms.txt fits.

An llms.txt file is an optional plain-text guide that points AI systems to your most important pages. It’s not required for any of the three platforms, and it does not replace crawler access or structured data. If you want to add that layer, our guide to llms.txt for ecommerce covers the format and where it helps.

Step 2: Get your product structured data right

Product structured data plays a crucial role in how AI assistants describe and rank products. When product structured data is enriched and identical to your feed, your products have a better chance of getting surfaced at the right time. This means complete Product, Offer, and AggregateRating markup with GTIN, price, and availability.

Adobe’s AI Content Visibility Checker found that US retail product pages score an average of 66%, meaning roughly a third of PDP content is invisible to the LLMs powering AI shopping. Product pages scored lower than homepages (75%) and category pages (74%). Structured data is the solution.

The Product schema fields AI engines actually read.

Structured data is only one part of the equation. In order for assistants to identify, price, and rank products, the right Schema.org fields are a must.

The core schema to focus on are:

• Product: name, description, brand, gtin / mpn, image, sku

• Offer: price, priceCurrency, availability, url, shippingDetails, hasMerchantReturnPolicy

• AggregateRating: ratingValue, reviewCount

GTIN and MPN matter more than most merchants assume. They let assistants match your listing to the same product across other retailers and independent reviews, which is exactly how Perplexity verifies claims.

A minimal, valid JSON-LD example:

{   "@context": "https://schema.org",   "@type": "Product",   "name": "Aero Quiet Tower Fan 42-inch",   "brand": {     "@type": "Brand",     "name": "Aero"   },   "gtin13": "0012345678905",   "sku": "AERO-TF42-WHT",   "image": "https://www.example.com/images/aero-tf42.jpg",   "description": "42-inch bladeless tower fan, 38 dB on low, 10 speeds, 8-hour timer.",   "offers": {     "@type": "Offer",     "url": "https://www.example.com/aero-quiet-tower-fan-42",     "price": "129.00",     "priceCurrency": "USD",     "availability": "https://schema.org/InStock"   },   "aggregateRating": {     "@type": "AggregateRating",     "ratingValue": "4.6",     "reviewCount": "1284"   } }

Match your on-page schema to your feed.

Accurate and complete product schema is a start, but next you have to make sure your on-page schema matches your feed. Why? Because consistency between on-page schema, your product feed, and Merchant Center listings is a major eligibility factor for Google AI Mode shopping.

For example, a price of $129 in the feed and a price of $119 on-page will read as unreliable data, which gets demoted or dropped entirely.

To prevent this, you can:

• Run price, availability, and GTIN through a single source of truth that populates both the feed and the page.

• Re-validate after every promotion, since sale pricing is where mismatches creep in.

• Check that availability updates when inventory hits zero. Otherwise you’ll get a stale “InStock” issue, which is a common flag.

This is also where a proper feed management solution can make a huge difference, automating and streamlining the above and more.

Make spec content machine-readable.

AI assistants can’t surface what they can’t find, and that includes spec content hidden in images. A product infographic that’s rife with dimensions, materials, compatibility, etc? Invisible.

Instead, move that kind of information to a plain HTML text box or spec table, while keeping the infographic as a visual aid.

Step 3: Syndicate your product feed to AI surfaces

With your various data types ready and visible, it’s time to syndicate your product feed to AI surfaces. While each AI answer engine requires different elements, rest assured there’s a lot of overlap as well.

ChatGPT: the OpenAI product feed spec.

Access to ChatGPT is application-based, done through the OpenAI product feed.

Once approved, you deliver the feed by SFTP, file upload, or a hosted URL. Per the official spec, every row needs:

• id

• title

• description

• link

• image_link

• availability

• price

• brand

In addition to the above, you may also need GTIN or MPN where applicable.

For the application process itself, you can simply:

1. Apply through OpenAI’s merchant portal and label your submission as Early Access participation.

2. Map your catalog to the required fields, adding optional enrichment (material, dimensions, use cases) where you have it.

3. Publish a full snapshot at least daily. More frequent updates help for fast-moving price or stock.

4. Monitor ingestion errors and fix rejected rows. (A rejected row is an invisible product.)

Once you’re done with the application, make sure to check your product count in your OpenAI Ads Manager panel. If it lines up with the number of products you submitted, it was successful. If not, retrace your steps and double-check that everything was done properly.

Google AI Mode: Merchant Center and free listings.

One of the easiest eligibility steps you can miss is enabling free listings. Google AI Mode reads free listings, not ads. This means your Merchant Center account can still be invisible to Google AI Mode if you’re only running Shopping Campaigns.

Fortunately, you can easily prevent this:

1. In the Merchant Center, go to Growth and then Manage programs to enable Free listings.

2. Complete every recommended attribute, not just the required ones.

3. Turn on automatic item updates so price and availability stay accurate in near real time.

4. Resolve every Diagnostics warning. Disapproved items are excluded from the Shopping Graph.

Much like the OpenAI process, complete data is a must, so take your time.

Perplexity: dashboard feed or the PayPal network.

Where Perplexity is concerned, you have two routes. First, the Perplexity merchant program, which is free with zero commissions, and accepts the Google Shopping feed format. But, if you use PayPal, there’s also the PayPal merchant network.

In either event, you can get into Perplexity feeds by:

1. Signing up for the Perplexity merchant program and connect your Google Shopping-format feed in the dashboard.

2. If you accept PayPal: confirming your products are also surfacing through the PayPal merchant network.

3. Auditing your review presence. Perplexity checks claims against independent reviews before recommending.

Just like the other two AI engines, complete and accurate data is a must, so make sure the foundational work is in place before you start this process.

One feed, many surfaces: where Feedonomics Surface fits.

When you’re trying to manage three different feeds, you have three times as many opportunities to make mistakes. Price mismatches, dated descriptions, and more can easily crop up and hurt your chances of being surfaced by AI. This is where Feedonomics Surface comes in.

Feedonomics Surface takes one catalog and syndicates it to Google, Meta, Microsoft, TikTok, and Pinterest today, with agentic-AI channels on the published roadmap alongside the Feedonomics–Perplexity product data partnership. AI-assistant syndication has also launched and is expanding, but not fully live everywhere.

For the basics, start with product feed management fundamentals. These best practices will make a solid foundation, while Feedonomics Surface can take those efforts even further.

Step 4: Opt into agentic checkout (Early Access)

There are a number of early access programs that fall within agentic checkout, making it possible for your products to not only show up in AI engines, but also be purchasable. Keep in mind the following programs are all early access and prone to changes, and can be potentially sunset as well.

ACP: OpenAI and Stripe’s open standard.

The Agentic Commerce Protocol (ACP) is an open standard, co-maintained by OpenAI and Stripe under an Apache 2.0 license, that lets AI shopping agents read merchant catalogs and, where enabled, place orders on a shopper’s behalf.

ACP launched in September 2025 alongside ChatGPT Instant Checkout. Then, in March 2026, OpenAI retired the standalone Instant Checkout button, saying the initial version lacked the flexibility it wanted, and pivoted ChatGPT shopping to discovery-first results that hand shoppers to the merchant’s own checkout or to retailer-operated ChatGPT Apps.

While Instant Checkout is gone, ACP itself continues. The April 2026 release added cart, feed, orders, and authentication support, and it remains the technical layer connecting merchant data to ChatGPT.

On fees, be precise:

• OpenAI’s official language is a “small fee on completed purchases” made inside ChatGPT.

• Specific percentages have been reported by press (The Information in January 2026), not confirmed by OpenAI. Treat them as reporting, not pricing.

• Purchases completed on your own site carry no OpenAI fee.

• OpenAI is not the merchant of record; you are.

UCP and AP2: Google’s commerce and payments layer.

Google’s Universal Commerce Protocol (UCP) was announced at NRF in January 2026, co-developed with Shopify, Etsy, Wayfair, Target, and Walmart, and endorsed by more than 20 partners. It covers the full shopping journey inside Google surfaces, including AI Mode. AP2 is the signed-mandate payments layer beneath it, which lets a shopper authorize an agent to pay within set limits.

Both are rolling out, not generally available. Participation requires an active Merchant Center account with checkout-eligible products — one more reason Step 3 comes first.

Perplexity: Buy with Pro on PayPal rails.

Buy with Pro offers one-click checkout inside Perplexity for US Pro subscribers, running on PayPal and Venmo rails. It is in beta. The terms are merchant-friendly:

• You remain merchant of record.

• Commissions are zero.

• Perplexity sponsors free shipping on eligible orders.

What to do while these programs are in beta.

You cannot control the rollout schedule, but you can make sure enrollment is trivial when it opens. This quarter:

1. Join the OpenAI merchant and Google UCP early-access lists.

2. Confirm your payment provider supports ACP or AP2 flows — Stripe and PayPal are furthest along.

3. Keep feed, returns-policy, and shipping fields complete and checkout-ready in every syndicated feed.

4. Read the concept background in what is agentic commerce and agentic AI in ecommerce so your team speaks the same language as your PSP.

Step 5: Build answer-ready content and third-party proof

The right feeds and schema will get you eligible for AI engines, but it’s content that decides whether an assistant surfaces your product and how it describes it. Like schema, this is another area where you have control.

Write PDP copy assistants can quote.

Complete schema acts as a blueprint, but quotable product detail page (PDP) copy that assistants can surface is as good as gold.

For example, lean in on use-case-driven sentences that an assistant can repeat, like, “weather-resistant material that’s suitable for temperatures below 32-degrees.”

Once you have a use-case-driven sentence, support it with:

• A plain-text spec table (dimensions, materials, compatibility, certifications)

• Short Q&A blocks answering the three questions shoppers ask before buying

• Comparison language that positions the product against its own siblings (“opt for the lighter material if you’re hiking in temperatures above 32-degrees.”)

It’s also a great idea to follow product description best practices in general, as these will help you with AI queries and driving conversions on your actual product pages.

Earn the reviews assistants cross-check.

Reviews and other forms of social proof are a general win for ecommerce businesses, but they’re also useful for AI assistants. For AI assistants, sustained review generation is a ranking input that sends positive trust signals.

For instance, Perplexity explicitly cross-checks product claims against independent reviews, while off-site mentions in publications and community threads feed retrieval across all three assistants.

Beyond on garnering social proof, you can make sure AI assistants factor in reviews by:

• Automating post-purchase review requests and keeping them running, not campaign-based.

• Seeding products with expert and publication reviews where the category supports it.

• Answering questions in community threads or forums under your brand name, as these often get cited.

Half the battle is getting people to review your products, but keep in mind you can be more proactive by engaging with social media. Again, threads, forums, and other social posts are frequently cited by AI assistants.

Publish buying guides that win AI citations.

As much as AI assistants love forum replies and social proof, they also love buying guides. To capitalize on this, write question-first buying guides or comparison articles that map to each of your top product categories.

It’s uncommon for a PDP to win a “which one should I buy” query. A buying guide? It’s made for this. With a well-structured guide that answers a comparison question within the first 100 words, your chances of winning a query are great.

Focus on giving a direct answer right away, number steps within the guide, include a comparison table where it makes sense, and keep paragraphs short. And then, err, profit?

Secondary CTA: Commerce AI Labs (Spreadsheet Link)

How to measure AI shopping visibility

So you’ve done all the things to make your products show up in AI assistants. Now, how can you be sure they’re actually showing up and delivering any kind of results?

Track AI referrals in analytics.

Within your analytics platform, like Google Analytics, you won’t find AI referrals by default. But, you can still track this information with a little handiwork.

ChatGPT referrals show up with a “utm_source” tag in analytics. Meanwhile, Perplexity and Google AI both show up as separate referrals. Within analytics, create an “AI channel” segment that groups both “utm_source” and the Perplexity and Google AI referrals.

Once you have a segment built, compare conversion rates and revenue per visit against your search and email baselines to determine both your overall AI visibility and its impact on your business.

Prompt-test your own products.

Once a month, run your category’s top buying prompts in ChatGPT, Google AI Mode, and Perplexity as if you were a customer on the hunt. Then, log four things per prompt:

1. Does your brand appear at all?

2. Which product is cited for each query?

3. Which URL does the assistant link: your PDP, a marketplace, or a competitor?

4. Is the price and availability it quotes correct?

While a single run can hint at potential problems, trends across three months or more will tell you far more.

Benchmarks to expect.

AI-referred business is still relatively new in the grand scheme of things, so benchmarks aren’t as clearcut as they are in something like SEO. But, there are still a handful of stats you can keep in mind when reviewing your own performance after getting your products into AI feeds:

• AI-referred traffic converts 42% better than non-AI traffic

• AI traffic had a 37% higher revenue per visit than non-AI traffic

• As of July 2026, AI traffic to retail sites is up 62% YoY

The above shows promise for AI traffic, hinting at major gains for retail sites in particular. Now, are ecommerce businesses ready for any of it? Our AI Pulse Survey paints a messy picture.

Common mistakes that keep products out of AI results

If your products don’t show in ChatGPT, the cause is almost always one of three things: the crawler is blocked, the feed is missing or stale, or the data on your page contradicts the data in your feed. Work through this list in order.

• Blocking OAI-SearchBot while assuming ChatGPT can still see the site. Fix: allow OAI-SearchBot in robots.txt and your CDN; block GPTBot separately if you want to.

• Assuming Shopping ads buy Google AI Mode placement. Fix: AI Mode reads free listings and the Shopping Graph, not bids. Invest in attribute completeness.

• Never enabling free listings in Merchant Center. Fix: Growth → Manage programs → Free listings. Check it today.

• Price or availability mismatches between feed, schema, and site. Fix: drive all three from one source of truth and re-validate after every promotion.

• Product specs locked inside images. Fix: move specs into plain HTML text or a table; keep the graphic as a visual aid.

• Expecting a Google-Extended block to remove the site from AI Mode. Fix: it doesn’t. Google-Extended only affects Gemini training and grounding.

• Letting feed data go stale. Fix: publish a full snapshot daily and update price and stock more often where you can.

• Treating press-reported beta checkout fees as confirmed pricing. Fix: rely on OpenAI’s official language and label everything else as reporting.

The final word

Getting into ChatGPT shopping, Google AI Mode shopping, and Perplexity shopping is a data job, not a marketing spend. The protocols will keep changing, but the fundamentals will not.

To ensure your products are prepared for this new reality of conversational shopping, start with these three things:

1. This week: audit robots.txt and CDN rules for OAI-SearchBot, PerplexityBot, and Googlebot, and confirm each returns a 200 on your product URLs.

2. This month: validate your product feed against the OpenAI spec and enable free listings in the Merchant Center.

3. This quarter: join the ACP and UCP early-access lists so checkout is a switch-flip when the programs mature.

If you want one clean catalog syndicated across shopping and AI surfaces without managing three feeds by hand, BigCommerce with Feedonomics Surface is the fastest route there.