They are not four different machines. Strip the branding and ChatGPT, Perplexity, Gemini and Amazon's assistant all run the same two tiers — a free crawl that earns you a citation and a link out, and a gated product feed that earns you a card with a price and a buy button — so the real question is never "which agent" but who holds the key to tier two. As of July 2026 only Google's key is sitting on the table: Merchant Center takes no application, and its free listings reach Gemini switched on by default.
We can be specific about this because it went badly. Getting independent WooCommerce stores through all four doors is work we did while building Contexta, and three of those doors turned out to be shut in ways no plugin can open — which is worth saying out loud, because the WordPress repository now sells plugins promising ChatGPT product-feed visibility in five minutes flat, while OpenAI's own documentation says feed onboarding is available to approved partners only. You can build a perfectly conformant endpoint and never get called.
What actually separates the four AI shopping agents?
Not the mechanic — the gatekeeper. Every one of the four splits your store into the same two tiers, and once you see the split, the marketing differences between them stop mattering.
Tier one — crawl to citation
- A search bot reads your live pages
- Governed by robots.txt, free, no application
- Produces a mention and a link back to your store
- Open on all four surfaces right now
Tier two — feed to product card
- A structured product feed you deliver
- Produces price, availability and a buy button
- Closed on three of the four
- The barrier is approval, not technology
Tier one is where most stores lose before they start, and usually over a naming confusion. OpenAI documents four crawlers, and the two that matter pull in opposite directions: GPTBot exists to gather training data, while OAI-SearchBot is the one that decides whether you can surface in ChatGPT's search and shopping answers. Blocking GPTBot costs you nothing in shopping visibility. Blocking OAI-SearchBot costs you all of it, and plenty of stores have done exactly that with a security plugin — the same bot-by-bot trade-off we mapped in which AI crawlers to allow.
Who holds the gate on each surface?
Four different doormen, and only one of them lets you in unaided.
| ChatGPT | Perplexity | Gemini | Alexa for Shopping | |
|---|---|---|---|---|
| Crawl tier open to you | ||||
| Feed tier open to you | ||||
| Who decides | OpenAI approval | Partner payment rails | Nobody — self-serve | Enterprise sales rep |
| Anyone asked you first | ||||
| What it charges you today | $0, ads sold separately | $0 | $0, ads sold separately | $0 "currently" |
One naming note that dates most competing coverage instantly: Amazon renamed Rufus to Alexa for Shopping on May 13, 2026, appending a note to that effect to its own original announcement. The recommendation logic carried over unchanged, so guides still written around "Rufus" are unrevised rather than wrong.
Can a WooCommerce store get into ChatGPT's or Perplexity's feed?
No to either, and WooCommerce isn't the reason for either. OpenAI's developer documentation states plainly that onboarding product feeds is currently available to approved partners, and the named partners are national chains and platform pipes: Target, Sephora, Lowe's, Best Buy, plus Shopify Catalog and Etsy. There is no feed URL to submit and no console to log into.
The frustrating part is that the specification itself is platform-agnostic. It asks for flat files over SFTP to an endpoint OpenAI issues at onboarding, and a WooCommerce store can produce those files perfectly. Approval is the obstacle, not your platform — which is precisely why no plugin can solve it, and why any product claiming to put your catalog in ChatGPT this afternoon is selling motion rather than access. The door that genuinely is open to you is the crawl, which we walk through in how ChatGPT shopping listings actually work.
Perplexity's gate is shut for a different reason, and it's worth naming because the paperwork looks so inviting. The Merchant Program is free and its application form is live, but Perplexity's own launch post scopes the program to large retailers, and the checkout path runs through PayPal Store Sync — which is request-access only, needs an existing PayPal merchant account, and names Wix, Cymbio, Commerce and Shopware as its platform partners. WooCommerce appears nowhere in any of it. A form existing is not the same as a path existing.
Is Google really the only open door?
Yes, as of July 2026 — and it's the least glamorous of the four, which is probably why the comparison pieces skip past it. Merchant Center has no application, no approval committee and no partner list. Free listings cost nothing, explicitly reach Gemini among their surfaces, and are turned on by default. WooCommerce has an official Google Product Feed extension carrying 50-plus attributes, and Google's automatic feeds re-check your site at least once every 24 hours.
Google is also the only one of the four with a genuine mechanical difference underneath, and it's easy to miss: retrieval runs against the Shopping Graph — Google's own index of over 60 billion product listings as of its May 19, 2026 update — rather than against a live read of your page at answer time. That inversion is what makes feed freshness outrank page copy on Google and nowhere else, and it's the detail behind what the Shopping Graph actually reads from your feed.
Be honest about where Google's welcome stops. Getting listed is open; getting a buy button is not. WooCommerce and WordPress appear nowhere among UCP's co-developers, announced January 11, 2026 alongside Shopify, Etsy, Wayfair, Target and Walmart — and the WooCommerce feature request asking for UCP support has sat open since January 13, 2026 with 31 votes and no response from Automattic. Google has also published no fee, commission or revenue-share terms for UCP checkout at all, and an absence of a published fee is not the same as no fee.
Is Amazon's AI already listing your products without asking?
Quite possibly, and this is the door nobody knocked on. Amazon's Shop Direct carries more than 100 million products from over 400,000 merchants who don't sell on Amazon at all, per Amazon's own March 11, 2026 announcement — sourced, in the words of its VP of Core Shopping Amanda Doerr, from publicly available data on the internet rather than from merchants. Participation is opt-out, by email, after the fact.
That contradicts a promise Amazon made in writing when Buy for Me launched on April 3, 2025: that brands would have the choice whether to participate. By January 2026, 145 brands had told one merchant's survey they believed their products were listed without consent, and CNBC reported that 180-plus businesses had contacted her — across Shopify, Squarespace, WooCommerce and Wix. Angie Chua of Bobo Design Studio described being turned into a drop shipper for Amazon against her will, after receiving orders from a program she'd never heard of.
The free thirty-second check is to search your analytics for the string utm_source=AMZN_US_ShopDirect. If it's there, you're already listed. Opting out runs through [email protected], with a stated two-business-day response — though the same inbox also handles merchants asking to opt in, which tells you how purpose-built the exit really is.
There's a sharp edge worth weighing before you call this free money. Buy for Me has Amazon's agent complete the purchase on your own checkout, injecting the shopper's encrypted payment details; you receive the name and address and own delivery, returns and service. Amazon's exact on-record phrasing is that it does not currently charge for this, on a beta whose eligibility it says is subject to change at any time. What it hasn't published is any merchant-facing documentation on liability — who owns the chargeback when a bot-placed order is disputed, and whether your fraud rules will kill the checkout. That silence matters more than the price, and it's the gap the signed-mandate approach in AP2 is built to close. Meanwhile Amazon sued Perplexity in November 2025 over an agent visiting Amazon without permission — a case the Ninth Circuit heard on June 11, 2026 and, as of this writing, hasn't ruled on.
Does missing the checkout gate actually cost you sales?
On the evidence so far, no — and the largest merchant that actually shipped in-chat checkout says the opposite of what you'd expect. Walmart put roughly 200,000 products into ChatGPT's Instant Checkout, and its EVP of product and design Daniel Danker went on record with WIRED in March 2026 saying it converted at about one-third the rate of simply clicking out to Walmart's own site.
OpenAI's behavior corroborates the quote rather than merely echoing it. Instant Checkout launched September 29, 2025; the 4% merchant fee arrived January 26, 2026; fewer than about thirty of Shopify's millions of merchants ever activated it; and OpenAI retreated from it in March 2026, roughly six weeks after starting to charge. The take rate didn't disappear out of generosity — the product underperformed a hyperlink, and the money moved to advertising instead. Run the arithmetic and the fee turns out to have been the small variable all along: 4% on top of card processing came to 7.0–8.1% per order, against a conversion gap worth roughly three times that much.
Read that against the tier-two anxiety the whole category runs on and it inverts. The supposedly second-class outcome — being cited, getting clicked, landing the shopper on your own store where your merchandising, upsells, email capture and pixel all still work — is the outcome that measurably converts better. The gate you can't get through leads to a room that performs worse, which is the same reason we keep arguing that chasing a checkout protocol is the wrong first move.
So what should a WooCommerce store actually do?
Work tier one on all four surfaces, walk through the single tier-two gate that's actually open, and stop buying access nobody is selling you. That's three concrete jobs, and none of them involves a protocol adapter.
Let OAI-SearchBot and PerplexityBot through, and audit whatever security plugin or Cloudflare toggle might be quietly refusing them. Put price, availability and identifiers in raw HTML — Google's Merchant Center documentation is explicit that structured data can't be generated by JavaScript after the page loads, and OAI-SearchBot doesn't run JavaScript either, so a price your theme injects client-side isn't there when it counts. Then set up Merchant Center, because it's free and it's the one gate with nobody standing at it.
All three jobs resolve to the same underlying thing: a product record an agent can read and trust. Doing that by hand across a few hundred SKUs is the task nobody finishes, which is why Contexta's Commerce Readiness audit checks each product for the fields these agents actually read — image, price, GTIN, brand, weight — and hands back a per-product fix link. The gatekeepers will keep changing their minds; three of the four already have this year. A catalog that reads cleanly is what survives every one of those reversals.
FAQ
Which AI shopping agent should a WooCommerce store target first?
Google, because it is the only one of the four whose product-feed tier an independent store can enter without approval as of July 2026. Merchant Center takes no application and no partner list, free listings cost nothing and explicitly reach Gemini, and they are turned on by default — plus WooCommerce has an official Google Product Feed extension. ChatGPT's feed onboarding is restricted to approved partners, Perplexity's checkout runs through PayPal rails that don't list WooCommerce, and Amazon's path runs through enterprise feed syndicators.
Is Amazon listing my products in its AI without my permission?
It might be — Amazon's Shop Direct carries over 100 million products from more than 400,000 merchants who don't sell on Amazon, and participation is opt-out rather than opt-in. In January 2026, 145 brands told one merchant's survey they believed their products were listed without consent, and CNBC reported 180-plus businesses contacting her across Shopify, Squarespace, WooCommerce and Wix. Search your analytics for the string utm_source=AMZN_US_ShopDirect to find out whether you are already listed, and email [email protected] to opt out.
Is ChatGPT Instant Checkout still available in 2026?
No — OpenAI retreated from Instant Checkout in March 2026, roughly six weeks after it began charging a 4% merchant fee on January 26, 2026. Fewer than about thirty of Shopify's millions of merchants ever activated it, and Walmart's EVP of product and design told WIRED that in-chat checkout converted at around one-third the rate of clicking out to Walmart's own site. The underlying Agentic Commerce Protocol survives, so any plugin still selling ChatGPT Instant Checkout for WooCommerce is marketing a feature that no longer exists.
Does blocking GPTBot remove my products from ChatGPT shopping?
No — this is backwards, and it is the most commonly botched robots.txt decision in AI commerce. GPTBot gathers training data, so blocking it costs you nothing in shopping visibility. OAI-SearchBot is the crawler that decides whether you can surface in ChatGPT's search and shopping answers, so blocking that one costs you everything — and security plugins and Cloudflare bot rules frequently block it without the store owner knowing.