ChatGPT, Google's AI search, and Perplexity all recommend products now — but they find, cite, and transact differently, so "optimise for AI shopping" isn't one job. As of mid-2026, ChatGPT is strongest at conversational discovery, Google's AI surfaces at shopping-data depth and checkout breadth, and Perplexity at cited, research-style comparisons. This is a neutral, criteria-based look at how they differ and what that means for a brand.
A note on scope: agentic-commerce features are moving fast and several have changed more than once across 2025–2026. The durable differences are in how each engine retrieves and cites; treat the specific shopping features as dated to mid-2026 and verify them against each platform's current announcements.
How do ChatGPT, Google AI search, and Perplexity differ for product discovery?
At a glance, by criteria rather than by verdict:
| Criterion | ChatGPT | Google AI search (AI Overviews / AI Mode) | Perplexity |
|---|---|---|---|
| Primary strength | Conversational discovery & recommendation | Shopping-data depth, checkout breadth | Cited, research-style comparisons |
| How it finds products | Live web retrieval + model knowledge | Google web index + Shopping Graph | Live web retrieval, source-first |
| Sourcing tendency | Established / authoritative sources | Google ranking signals + product feeds | Forums/community (e.g. Reddit) + cited web |
| Citation style | Selective, fewer links | Inline links + shopping modules | Citation-heavy, sources listed |
| Checkout rails (mid-2026) | Agentic Commerce Protocol (with Stripe) | Universal Commerce Protocol; checkout in AI Mode | In-assistant purchase for subscribers |
The table is a starting map, not a scoreboard — the sections below explain each row and where the figures come from.
How does each engine find products?
They retrieve from different places:
- ChatGPT blends its trained knowledge with live web retrieval, surfacing products conversationally as a shopper narrows intent ("a waterproof jacket for commuting under £150"). It reads the open web rather than a dedicated product feed.
- Google's AI surfaces sit on top of Google's web index and its Shopping Graph — which Google reported holds more than 50 billion product listings, with over two billion refreshed hourly (Google's own figures, attributed to its 2025–2026 announcements). That feed depth is Google's structural advantage for shopping specifically.
- Perplexity is retrieval- and source-first: it runs a live search and assembles a cited answer, which makes it well-suited to "compare X vs Y" style product research.
The practical takeaway: a clean, server-rendered product page with structured data helps all three, but Google additionally rewards a complete, accurate product feed.
How does each one source and cite?
Sourcing tendencies differ enough to change who gets named. As of mid-2026, studies reported that ChatGPT skews toward established, encyclopedic and authoritative sources; Perplexity leans heavily on community and forum content (Reddit prominent among them) alongside cited web pages; and Google's AI surfaces reflect Google's own web index and ranking signals plus its product data. Treat the exact percentages floating around as study-dependent and directional — the direction (ChatGPT authoritative-leaning, Perplexity community-leaning, Google index-leaning) is the stable, useful signal.
Why it matters: the same product page can be cited by one engine and missed by another, purely on sourcing policy and crawler access — not on quality. That's the core reason to measure presence per engine rather than assume it transfers. The mechanism is unpacked in why AI cites one brand and ignores a near-identical competitor.
What are the shopping and checkout features?
This is the fastest-moving layer, so read it as a mid-2026 snapshot:
- ChatGPT surfaces products in-conversation and uses the Agentic Commerce Protocol it co-developed with Stripe to handle checkout sessions and payment; the in-chat purchase experience has shifted across 2025–2026, so confirm the current flow.
- Google and Shopify back the Universal Commerce Protocol, and Google has brought checkout into AI Mode with select merchants, alongside visual and shoppable features. The two protocols are compared in WebMCP vs UCP.
- Perplexity offers in-assistant purchasing (notably for its subscribers), keeping discovery and buying in one cited flow.
The bigger shift behind all three is agentic commerce — software agents completing the purchase. Getting ready for it is its own checklist: agentic commerce readiness and preparing your product catalog for AI agents.
The engines don't share one mechanism, so they don't share one answer. The same product can win in ChatGPT, lose in Perplexity, and never appear in Google's AI Mode — not because of quality, but because each finds, cites, and transacts differently.
Which should a brand optimise for?
All three — because shoppers use all three, and presence in one doesn't guarantee the others. The even-handed playbook:
- Build the foundation once. A reachable, parseable, structured product page serves every engine.
- Feed Google specifically. Its shopping advantage runs on complete, accurate product data.
- Earn community and authoritative mentions. That's what Perplexity and ChatGPT respectively lean on.
- Measure each engine separately. Track who cites and recommends you per engine, over time — don't infer one from another.
There's no single engine to "win"; there's a share of voice to hold across all of them. Sampling the same product questions across ChatGPT, Google's AI surfaces, and Perplexity — repeatedly, and logging exactly who gets recommended where — is precisely what Buffy Intel is built to do.
