A Canadian grocer moved shopping into ChatGPT: what Loblaw's move means for AI search

For most brands, AI search is still a visibility question: are we mentioned, are we cited? A Canadian grocer just turned it into a commerce question. When the checkout lives inside the conversation, being named by the AI stops being marketing and starts being distribution.

There is a line between being talked about and being transacted with, and AI search has mostly sat on the talked-about side. A brand gets mentioned in an answer, a reader maybe clicks through, commerce happens somewhere else. A Canadian grocer just moved that line.

What Loblaw did

On 12 February 2026, Loblaw announced a first-of-its-kind PC Express shopping app inside ChatGPT. Canadians can explore menu ideas, curate the ingredients for a meal and pick suggested products from their local store, then add those items to a PC Express cart and complete the purchase in the PC Express app. Loblaw also said it would equip its staff with ChatGPT Enterprise. It is one of the earliest large-scale grocery integrations directly into a conversational AI environment.

Strip away the grocery specifics and the pattern is what matters: the store moved into the conversation. The AI is no longer pointing at where you can buy; it is the surface where buying begins.

"When the store lives inside the answer, being named by the AI is the new version of being on the shelf."

Why Canada, and why now

Canada is fertile ground for this. One 2026 study found 56% of Canadians have tried ChatGPT, ahead of the equivalent US figure, and that a growing share of younger Canadians reach for an AI assistant as their first search instinct rather than a traditional engine. Consumer adoption is running ahead of business adoption - a gap that is precisely where early movers win.

That gap is the whole story for brands. The customers are already in the AI. Most competitors are not yet optimising for it. The distance between those two facts is opportunity, and it does not stay open forever.

The shift this represents

AEO has been described, fairly, as being about mentions and citations. Loblaw's move is a reminder that those are a means, not the end. The end is being present at the moment of choice.

• Naming becomes distribution. When a purchase can complete inside a conversation, the brand the AI names is the brand that gets bought. Being unnamed is not a ranking problem; it is a lost sale.
• The shelf is now an answer. Shelf placement used to be a physical and merchandising fight. Its AI-era equivalent is whether an engine surfaces you when a shopper describes a need in their own words.
• Consumer adoption leads. Canadians are adopting AI search faster than businesses are adapting to it. That asymmetry rewards brands that move before the behaviour is fully mainstream.
• Citability is the prerequisite. An engine can only name you in a shopping context if it has clear, trustworthy, consistent information about what you sell. Thin brand data is a barrier to the transaction, not just the mention.

The takeaway

Loblaw did not just add a feature. It demonstrated where AI search is going: from a place brands hope to be mentioned to a place where the buying actually happens. In a market where most consumers have already tried these tools, the brands that make themselves easy to name and cite are positioning for the moment of purchase, not just the moment of research. The ones that treat AI search as an afterthought are, increasingly, absent from the shelf.

What a brand should actually do about it

The Loblaw story is easy to read as a headline and hard to read as a to-do list. Most brands will not build their own app inside ChatGPT this quarter, and they do not need to. What they need is to be the answer the engine reaches for when a shopper describes a need in their own words, because that is the surface Loblaw just validated. The work is less about a single launch and more about making your product information legible to a machine that has to decide who to name.

Start with the plumbing. An engine can only recommend a product it can understand, so the first move is making your catalogue clean and consistent everywhere it appears: your own site, your retailer listings, and the third-party sources engines lean on. If your product name, size, price band and use case say three different things in three places, the model has three reasons to hedge and name someone clearer instead.

• Fix the product facts first. Titles, specifications, ingredients or materials, and "who it is for" should read the same on your site as they do on the retailers who carry you. Contradiction is the enemy of citation.
• Write for the question, not the keyword. Shoppers ask engines things like "what should I make for a nut-free birthday dinner for eight". Content that answers real intents in plain language is far easier for a model to lift than a page stuffed with search terms.
• Earn third-party corroboration. Engines trust a claim more when independent sources repeat it. Reviews, comparison pages and reputable mentions are not vanity metrics here; they are the evidence a model uses to feel safe naming you.
• Keep it current. Discontinued lines, stale prices and dead links teach an engine that your data cannot be trusted at the moment of purchase, which is exactly the moment that now matters.

How to know whether it is working

Traditional search gave you a rank and a click. AI commerce gives you neither in the old shape, so the measurement has to change with it. The question is no longer "where do we rank" but "when a shopper in our category asks, does the engine name us, and does it name us accurately". That is answerable, but only if you go looking for it deliberately rather than waiting for a dashboard to tell you.

Build a set of the buying questions your customers actually ask and run them across the major engines on a regular cadence. You are watching for three things: whether you are named at all, whether the description the engine gives is correct, and whether a competitor is named in your place. A brand that is mentioned but mis-described has a data problem; a brand that is never mentioned has a presence problem; and a brand that watches a rival get named for its own strength has a competitive problem. Each points at different work.

"In AI commerce you are not tracking a rank. You are tracking whether the engine trusts you enough to name you when money is on the line."

Treat this as a running measure, not a one-off audit. Engines change their models, their sources and their behaviour often, and the answer you got last month is not a promise about next month. The brands that hold their place will be the ones that check, correct, and check again, the same way they once watched their shelf position in a physical store.

What this does not mean

It is worth being clear about the limits, because the temptation with a story like this is to over-read it. Loblaw putting groceries inside ChatGPT does not mean traditional search is finished, that every category will move into a conversation at the same pace, or that a single integration has rewritten how people shop overnight. Habits move slowly even when tools move fast, and most purchases will still begin in familiar places for some time yet.

The lesson is narrower and more useful. When buying can happen inside an answer, the brands with clear, consistent, well-corroborated information are the ones an engine can safely name, and the ones with thin or contradictory data are the ones it quietly skips. That holds whether or not you ever build anything inside ChatGPT yourself. The right posture is neither panic nor dismissal: treat AI visibility as infrastructure you maintain rather than a campaign you run once, so that when the moment of choice moves into the conversation in your category, you are already the brand that is easy to name.

Related: what ChatGPT shopping apps mean for how brands get discovered →