Comparison
What are the best AI shopping agents and what does each one need from my store?
Compare the leading agentic commerce platforms — ChatGPT Operator, Perplexity Shopping, and Microsoft Copilot — and what each one needs from your store to recommend it.
Last updated:
Agentic commerce has stopped being a thought experiment. Three platforms now drive most of the AI-assisted shopping traffic hitting DTC stores: ChatGPT's Operator, Perplexity Shopping, and Microsoft Copilot. Each one picks products differently, and each one reads your store in its own way. If you only optimize for one, you'll get recommended on one.
Here's how the three best AI agents compare, and what your store needs to surface inside each of them.
ChatGPT Operator
OpenAI's shopping agent runs inside ChatGPT and completes purchases through Instant Checkout on Shop Pay. It reads your store like a careful human — but it also looks for a machine-readable manifest. The emerging convention is an agent.json file at your domain root that declares your checkout endpoints, return policy URL, and supported payment rails. Stores that publish one get parsed faster and recommended more often.
What it needs: Shop Pay enabled, guest checkout allowed, a clean PDP, and ideally an agent.json plus an llms.txt at the root. Heavy modals and pop-ups will get your store skipped entirely.
Perplexity Shopping
Perplexity leans hardest on structured data. Its shopping results are built almost entirely from schema.org markup — Product, Offer, AggregateRating, and Review in particular. If your PDPs don't emit valid JSON-LD with price, availability, and reviews, Perplexity quietly treats your catalog as low-confidence and recommends competitors instead.
What it needs: complete schema.org/Product markup on every PDP, valid Offer with currency and availability, and review data exposed as AggregateRating. Shopify's default theme covers the basics — custom themes usually don't.
Microsoft Copilot
Copilot blends Bing's shopping graph with live page reads. It rewards merchants already feeding the Microsoft Merchant Center — product feed, GTINs, and Microsoft Shopping Campaigns — but it also re-crawls the live page to verify price and stock. A mismatch between your feed and your live PDP is the most common reason Copilot silently drops a product.
What it needs: an up-to-date Microsoft Merchant Center feed, GTINs on every variant, and live PDP price/stock that matches the feed within a few minutes.
Side-by-side
| Agent | Primary signal | Checkout path |
|---|---|---|
| ChatGPT Operator | agent.json + llms.txt | Instant Checkout (Shop Pay) |
| Perplexity Shopping | schema.org JSON-LD | Deep-link to merchant PDP |
| Microsoft Copilot | Merchant Center feed + live crawl | Deep-link to merchant PDP |
How to be ready for all three
The good news: there's heavy overlap. Clean PDPs, complete schema.org markup, a published returns policy, and frictionless checkout cover the first 80% across every agent. The remaining 20% is platform-specific — an agent.json for Operator, a healthy Merchant Center feed for Copilot.
Daeri's Agent-Readiness Score grades your store against all three at once across five dimensions — Discoverability, Product legibility, Inventory, Action surface, and Trust — so you can see exactly which agent is most likely to skip you, and why.
Run a free audit and we'll email you a scored report in a few minutes.
Frequently asked questions
- Which AI agent drives the most shopping traffic right now?
- ChatGPT, by a wide margin — it has the largest user base and Instant Checkout finishes the purchase inside the chat. Perplexity is the fastest-growing on a percentage basis, and Copilot benefits from being embedded in Windows and Edge.
- Do I need a different optimization for each agent?
- Mostly no — clean PDPs, full Schema.org markup, and Shop Pay cover the first 80% across all three. The remaining 20% is platform-specific: an agent.json for ChatGPT Operator and a healthy Microsoft Merchant Center feed for Copilot.
- Does Perplexity buy products directly?
- Perplexity Pro completes some purchases via partner integrations and deep-links to the merchant PDP for others. Its recommendations are built almost entirely from Schema.org data, so weak JSON-LD = invisible to Perplexity shoppers.
- What kills my chances of being recommended by any of them?
- Pre-checkout modals, login walls, mismatched feed-vs-PDP price/stock, missing variant-level Offer JSON-LD, and policy pages rendered in JavaScript. All five are silent killers — the agent just picks a competitor without leaving a trace.