ChatGPT Shopping vs Perplexity vs Google: Three Different Checkout Architectures
In late 2025 and early 2026 the three biggest AI assistants moved from being recommendation engines to being in-assistant transaction layers. Each took a different path. ChatGPT shipped on the Agentic Commerce Protocol (ACP). Perplexity launched Instant Buy on PayPal infrastructure. Google rolled out the Universal Commerce Protocol (UCP) inside its Merchant Center stack.
AI Assistants Are the Decision Layer, Not a Discovery Channel
Traditional search returns results. AI assistants synthesize attributes, reviews, policies, pricing, and eligibility into a single answer. Incomplete or inconsistent product data means the assistant cannot confidently compare your SKU. That SKU gets excluded from the recommendation set before any human sees it.
How to Get Products Listed in ChatGPT Shopping
1. Build a Canonical SKU Truth
The ACP feed wants stable item_id values, complete title and description, canonical URL, primary image, accurate pricing, real-time availability, and proper variant grouping via a shared group_id. Inconsistencies between your storefront and the feed kill checkout eligibility.
2. Control Eligibility Flags Intentionally
Two flags decide everything: is_eligible_search and is_eligible_checkout. Missing policy URLs (returns, shipping, terms) structurally block checkout even when the product is otherwise complete. Most brands discover the gap only after the rejection.
3. Support the Session Lifecycle
Feeds need daily refresh at minimum. Intraday volatility on price or stock causes checkout failures mid-session, and ACP penalises merchants whose carts construct successfully and then collapse at confirmation.
Perplexity Instant Buy: Authority + Payment Compatibility
Perplexity weights authority more heavily than ChatGPT does. Selection mirrors editorial citation patterns: the more often credentialed sources discuss your product, the more often Perplexity surfaces it for buy-intent queries. Payment compatibility with PayPal infrastructure gates instant transactions; without it your product gets a link-out instead of an in-chat checkout, which collapses conversion.
Google UCP Merchant Guide: Eligibility Is a Compliance Problem First
UCP eligibility starts with Merchant Center account status, product approval, return policies, customer support information, matching feed and checkout API IDs, and the native_commerce attribute. Most rejections trace back to a single missing field that the merchant did not realise was structurally required.
The Core Problem: Data Architecture, Not Content Volume
You don’t lose because you ranked #3 instead of #1. You lose because the assistant couldn’t safely include you in the comparison set.
Building AI Commerce Readiness Across All Three Platforms
Four pillars hold the work together:
- Stable variant identity (GTIN, SKU, group_id) across catalog, feeds, and storefront.
- Structured policy infrastructure (returns, shipping, warranty) machine-readable and consistent.
- Review and performance signals attached at the variant level, not the parent product.
- Attribution engineering so AI-driven sessions stop landing in "Direct" traffic and start being measurable.
Where to Start
The fastest path forward: examine your variant mapping, your eligibility flags, your policy completeness, and your checkout compatibility against each protocol. Most brands are within reach of all three platforms. They just have not stress-tested the data layer that the protocols rely on.