AI shopping agents completed more than 120 million transactions in a single week in China, according to a new NielsenIQ report. Analysts now size agentic commerce at US$3–5 trillion by 2030. For Canada’s natural health trade, the immediate stakes sit in AI recommendations rather than automated checkouts.
The week the numbers changed
For two years, agentic commerce lived in conference keynotes. NielsenIQ’s new global report, The Commerce Revolution: Where East Meets West, moves it into the operating column.
Indeed, the evidence comes from Asia. Ant Group reports that Alipay’s AI Pay processed more than 120 million transactions during a single week in February. Alibaba’s Qwen assistant reached 100 million monthly active users within two months of launch.
As NIQ’s Emilie Darolles put it, in parts of Asia AI is “quietly completing purchases at a scale most boardrooms haven’t registered.”
Meanwhile, capital has registered it. McKinsey values the global opportunity at US$3–5 trillion by 2030. Morgan Stanley expects agents to drive US$190–385 billion of American e-commerce by the same year, or up to a fifth of the total.
NIQ’s consumer data supplies the brake. Roughly 34 per cent of shoppers now use AI to research products, and 23 per cent use it to summarize reviews. Only about 8 per cent have ever let an AI complete a purchase.
Influence therefore runs well ahead of automation. Everything a Canadian retailer or brand should do next follows from that gap.
Discovery is where supplements are won
Natural health sits closer to this shift than most categories. That is because a supplement purchase begins as a research question.
Which form of magnesium absorbs best? Does an ingredient interact with a medication? What separates one omega-3 from another? Consumers now hand exactly this homework to AI tools first.
In fact, the traffic already shows it. Adobe Analytics measured a 1,200 per cent rise in visits from generative AI sources to US retail sites between July 2024 and February 2025.
Those visitors also behave differently. They read more pages and bounce less than almost any other channel, though they convert less for now. AI is sending people who are still deciding.
In a category that sells on evidence, that matters. Whoever the AI cites during the decision gets the sale.
Why a citation is not a ranking
Citation works differently from a search ranking. An assistant answers “best vitamin D for someone on a statin” with three products. The old path from impression to click to cart collapses into a single answer.
As a result, vague specifications carry a steeper penalty here. A brand with inconsistent listings, or with details locked inside PDFs, does not simply rank lower. It never appears at all.
Canadians are drawing a firm line
No Canadian data yet measures agentic commerce in natural health specifically, and that gap is worth stating plainly. Even so, the general numbers tell a coherent story.
Visa’s consulting arm finds that only 27 per cent of Canadians know what agentic AI is. Fifteen per cent use generative AI for shopping. Just half trust the AI features online retailers offer.
KPMG in Canada found that 78 per cent of consumers leaned on AI tools during holiday shopping. Yet 86 per cent insist on approving every step before an agent acts.
Read commercially, that is a sequencing instruction rather than a rejection. Canadians are adopting the research layer quickly. They are withholding the transaction layer deliberately, which makes the answer layer the battleground for 2026 and 2027.
However, independents face a specific catch. Their traditional edge is floor staff who genuinely know the products. That edge carries no weight in a channel that reads structured information instead of holding a conversation. Translating expertise into a form machines can parse is the work of the next eighteen months.
A licence becomes a language machines read
Canada’s natural health products framework hands the domestic industry an unusual asset. An NPN licence carries claims that are documented, bounded and verifiable.
In other words, those are the qualities an AI system looks for when it weighs which products to trust. In a market where machines score credibility, a monograph-backed claim beats a superlative every time.
The same mechanism cuts the other way. AI systems train heavily on American content, and Canadian rules do not permit many of the structure/function claims that circulate there.
Marketing copy that drifts beyond a licence no longer sits harmlessly on a forgotten product page. An assistant repeats it verbatim at the moment of recommendation. The same applies to “detox” or “cleanse” language, which regulators and increasingly consumers discount.
In short, compliance and visibility have long occupied separate budgets. They are now converging into one.
Time to prepare, and how to spend it
The 8 per cent figure buys the industry something rare: time, with a clear to-do list attached.
First, publish complete specifications, dosage forms, sourcing and NPN numbers as text a crawler can parse. Images and downloads do not count. Reconcile every public claim against the licence behind it.
Then add plain, self-contained answers to the questions shoppers actually ask AI tools. Finally, split out AI-referred visits in analytics. A channel nobody measures is a channel nobody defends.
Similarly, retailers face the same test one level up. In this channel, product pages and category copy stand in for the floor team.
The East has proven that consumers will eventually hand agents the transaction. Canada is still deciding which products those agents may trust. Operators who already sell on proof rather than price start that contest in front.
Frequently asked questions
What is agentic commerce?
Agentic commerce is retail in which AI agents act on a consumer’s behalf. In other words, they research, compare and, in mature markets, complete purchases. Ant Group reports Alipay’s AI Pay processed over 120 million transactions in one February 2026 week in China. In North America, agents mostly shape decisions instead of executing them.
Are AI agents buying supplements in Canada yet?
No, not at measurable scale. Canadian surveys show heavy use of AI for shopping research alongside firm resistance to autonomous purchasing. Some 86 per cent of consumers want to approve every step. The current battleground is the AI recommendation, which shapes what shoppers consider before they buy.
How does a natural health brand get recommended by AI shopping tools?
By being verifiable. Publish complete specifications as parseable text. Match claim language to the product’s NPN licence. Keep evidence-based content in extractable formats and details consistent across every listing. AI systems favour products they can confirm, and contradictory information tends to exclude a brand entirely.
What should retailers do first?
Review how product pages read to a machine rather than a person. That means accurate specifications, licensed claim wording, and self-contained answers to common questions. Then segment AI-referred traffic in analytics. Adobe recorded 1,200 per cent growth in that channel in under a year, so those visits are likely already arriving.
This is independent editorial analysis. It references NielsenIQ’s report The Commerce Revolution: Where East Meets West and other third-party research as market evidence, not as an endorsement of any company, platform, or product. Cited figures are the originating firms’ own.
