May 2026 may be remembered as the month AI search stopped pretending it was just about answering questions and openly became about buying things. Within a two-week window, Google used its I/O stage to wire a shopping cart into Search, Gemini, YouTube, and Gmail; OpenAI swapped in a new default model and handed product discovery in ChatGPT to a Klarna-powered app; Microsoft made autonomous, computer-using agents generally available in Copilot; and Perplexity quietly turned its shopping experience into a personalized stylist. The phrase everyone used was "agentic commerce," and for once the demos backed up the buzzword.
For Shopify and e-commerce brands, this matters because the surface where customers first encounter your products is shifting. The question is no longer only "do I rank on Google?" but "will an AI agent find, trust, recommend, and transact my catalog?" Below is a grounded recap of what actually shipped in May 2026 across the four major assistants — ChatGPT, Google AI Mode/Gemini, Perplexity, and Copilot — and what each change means for your visibility.
ChatGPT: a new default model and a pivot to product discovery
OpenAI made GPT-5.5 Instant the default model across every ChatGPT tier — Free, Go, Plus, Pro, Business, Enterprise, and Edu — early in May. The pitch was smarter, more accurate answers with fewer hallucinations and better personalization controls. For brands, a more reliable default model means AI-generated buyer guidance is more likely to cite real products and real specs rather than confabulate them.
The bigger commerce story was a strategic clarification. Earlier in 2026, OpenAI shut down Instant Checkout — the feature that let you buy inside ChatGPT — citing limited merchant adoption (only around 30 Shopify merchants ever went live) and missing infrastructure for sales tax and fraud. In May, OpenAI doubled down on the opposite end of the funnel: product discovery. Its shopping research experience asks clarifying questions about budget and criteria, researches across the web, and returns a personalized buyer's guide with multiple product matches and trade-offs — strongest in detail-heavy categories like electronics, beauty, home and garden, and sports.
To feed that discovery engine, OpenAI brought in a partner. On May 20, Klarna launched a Shopping Search app inside ChatGPT, built on Klarna's Product Search MCP (Model Context Protocol) server. Shoppers describe what they want and see visual results with up-to-date prices, availability, and offers drawn from Klarna's commerce data — more than 100 million products across 400 million listings in 13 markets. Crucially, instead of checking out in chat, the app redirects shoppers to the merchant's own site to complete the purchase. That is a meaningful signal: ChatGPT wants to be the place you decide, and send you to the merchant to buy.
Google I/O 2026: Universal Cart, UCP, and agentic AI Mode
Google's I/O (May 19–20) was the month's center of gravity. On the model side, Google launched Gemini 3.5 for frontier agentic and coding work and Gemini Omni, a model that generates output across modalities starting with video. The Gemini app gained a proactive "Daily Brief," a redesign, and a personal agent called Gemini Spark.
But the announcements that should grab every merchant's attention were on the commerce side. Google introduced the Universal Cart — a single, intelligent cart that follows you across Search, the Gemini app, YouTube, and Gmail, letting you add items wherever you encounter them. It is rolling out across Search and Gemini in the U.S. this summer, with YouTube and Gmail to follow.
Underneath it sits the Universal Commerce Protocol (UCP), an open standard Google co-developed with Shopify (alongside Etsy, Wayfair, Target, and Walmart, and endorsed by 20-plus others including Adyen, Mastercard, Stripe, and Visa). UCP gives AI agents a common language to connect and transact with any merchant, and Google paired it with the Agent Payments Protocol (AP2) for guarded, agent-initiated payments where you set the brands, products, and spending limits. Early checkout features are live across merchants including Nike, Sephora, Target, Ulta, Walmart, Wayfair, and Shopify merchants such as Fenty and Steve Madden. Google also began rolling information agents into Search to work in the background and brought Gemini 3.5 Flash's agentic coding into Search to build interactive, generative UI for shopping questions.
If you run a store, the takeaway is blunt: agentic checkout from inside an AI answer is no longer hypothetical, and the Shopify integration means many merchants get the plumbing without building it. This builds directly on the agentic-storefront trend we covered in Shopify agentic storefronts and ChatGPT discoverability, and dovetails with the broader feature set in the Shopify Editions 2026 and Summer '26 Horizons releases.
Perplexity: shopping that gets personal
Perplexity spent May making its commerce experience feel less like search and more like a personal stylist. New memory enhancements let Perplexity remember past threads, favorite brands, and active projects, so shopping results lean toward your aesthetic rather than whatever an advertiser paid to surface. The company framed this as "shopping that puts you first."
It also shipped universal tabs, letting you jump to Links, Images, Videos, Places, and a dedicated Shopping tab directly from the top of any thread — turning a single question into a shoppable surface. These sit on top of Perplexity's existing commerce stack of in-chat purchasing and a virtual try-on avatar feature. Perplexity's commerce ambitions were validated externally too: it landed at No. 31 on CNBC's 2026 Disruptor 50 list. For continuity on how Perplexity has evolved this year, see our Perplexity updates and March AI search recap.
Microsoft Copilot: agents that use computers
Microsoft's May updates were less about shopping and more about the agent layer that will eventually power it. Claude Opus 4.8 became available in Microsoft 365 Copilot, aimed at complex, multi-step tasks with better tool selection and instruction following. In Copilot Studio, computer-using agents reached general availability — agents that operate websites and desktop apps through the UI — alongside agent-to-agent (A2A) communication so agents can delegate and collaborate. Microsoft also expanded federated MCP connectors, pulling live data from partners like Canva, HubSpot, Linear, and Notion into Copilot.
For brands, Copilot's direction reinforces a theme: the same MCP standard powering Klarna's ChatGPT app and Google's UCP feeds is becoming the universal way agents read commerce data. If your product information is exposed cleanly via these protocols, agents across every assistant can act on it.
May 2026 AI search updates at a glance
| Platform | Key May 2026 update | Why it matters for brands |
|---|---|---|
| ChatGPT (OpenAI) | GPT-5.5 Instant default; Klarna Shopping Search app via MCP (May 20) | More reliable discovery; live multi-merchant prices, redirect-to-merchant checkout |
| Google AI Mode / Gemini | Gemini 3.5 + Omni; Universal Cart on UCP; AP2 payments (I/O, May 19–20) | Agentic checkout inside AI answers; Shopify co-built the standard |
| Perplexity | Shopping memory/personalization; universal tabs with Shopping tab | Recommendations skew to user taste, not ad spend; new shoppable surface |
| Microsoft Copilot | Claude Opus 4.8; computer-using agents + A2A GA; federated MCP connectors | Standardized agent layer; MCP becomes the common commerce data pipe |
What this means for brand visibility
The connective tissue across all four platforms is the Model Context Protocol and UCP: open, structured ways for agents to read and act on commerce data. That changes the optimization job. Ranking blue links still matters, but increasingly you are optimizing to be the product an agent confidently recommends and can transact. Three practical implications:
- Structured product data is now revenue infrastructure. Klarna's app, Google's UCP, and Copilot's connectors all depend on accurate price, availability, and attribute feeds. Stale or messy catalog data means agents skip you. On Shopify, much of this is wired in through UCP, so your job is keeping the underlying data clean — if you are still planning your store, our guide to building a Shopify store with AI is a useful starting point, and you can start a free Shopify trial to test the agentic features firsthand.
- Buyer-guide content earns citations. ChatGPT's shopping research and Perplexity's personalized answers reward content that genuinely compares options and explains trade-offs. Thin, salesy pages get ignored; thorough, honest comparisons get cited.
- Trust signals decide recommendations. With OpenAI emphasizing reliable, low-hallucination answers and Perplexity weighting user preference over ad spend, reviews, clear specs, and consistent data across the web increasingly determine whether you make the shortlist.
Not sure where you stand across these assistants? Run a free AI visibility audit to see how ChatGPT, Gemini, Perplexity, and Copilot currently represent your brand.
The Bottom Line
May 2026 was the month agentic commerce got real. Google put a cart inside its entire ecosystem on a standard it built with Shopify, OpenAI clarified that ChatGPT is the place shoppers decide before being sent to your store, Perplexity made discovery personal, and Microsoft hardened the agent layer underneath it all. The unifying lesson for brands is that AI assistants are becoming a primary discovery and decision surface, and the protocols feeding them — MCP and UCP — run on clean, trustworthy product data and genuinely helpful content. Get those right now, and you will be the brand the agents recommend when the rest of this rolls out at scale.