Orders From AI Grew 13x While Email Went Quiet

Two numbers landed this week that point in the same direction, and neither of them is about a new model.

Shopify published its Q1 2026 agentic commerce figures: AI-driven traffic to its stores grew 8x year over year, and orders originating from AI-powered searches grew nearly 13x. In the same stretch, analyses of Gmail's Gemini-powered inbox reported that up to 40% of emails that technically reach the inbox are being deprioritized before a human ever looks at them.

One channel is opening. Another is narrowing. Both are being decided by a machine that sits between you and your customer.

The 8 and the 13 are the whole story

It would be easy to read Shopify's numbers as "AI sends traffic now" and move on. The interesting part is the gap between the two figures. Traffic grew 8x. Orders grew almost 13x. Orders grew faster than the visits that produced them, which means people arriving from an AI answer buy at a higher rate than the average visitor.

That makes sense once you picture how the two journeys differ. A page of search results is a shelf: you scan ten options and most of them are noise. An AI answer is closer to a recommendation from someone who already listened to what you wanted. By the time that person clicks, half the deciding is done.

Orders grew faster than traffic. The visitors an AI sends you are not browsing. They have already been qualified by the thing that recommended you.

Shopify also did something structurally new. Agentic Storefronts give merchants centrally managed access to ChatGPT, Microsoft Copilot, Google AI Mode and the Gemini app from one admin screen. And through the Agentic Plan, brands that do not run Shopify stores at all can now add products to the Shopify Catalog to reach those same AI channels.

That last part is the one most people will skim past. Shopify is selling distribution into AI surfaces separately from its shop software, positioning itself as the rail between product catalogs and answer engines whether or not you use the rest of the platform.

Meanwhile, delivered stopped meaning seen

Email deliverability used to have three gates. Does the message authenticate, does it clear spam filtering, does it land in the primary inbox rather than a promotions tab. Gmail's AI features added a fourth gate, and it sits after all of those.

Your message can authenticate, clear the filters, land in the primary inbox, and still be quietly pushed down because a model decided it was not the most relevant thing in front of that person today. Reports put that at up to 40% of delivered mail.

The measurement side is where this gets genuinely awkward. Omeda's analysis of billions of emails found open rates rising to 45.6% after AI summaries launched, while click-through fell from 4.35% to 3.93%.

Opens went up and clicks went down at the same time. That is not a contradiction, it is a description of what summaries do. The summary counts as an open, and then it answers the reader's question well enough that they never click. Your dashboard gets greener while the actual outcome gets worse.

If you report on open rate, this change will look like a win. It is not one.

The same shift wearing two costumes

Strip away the detail and both stories are the same story. A model now stands between your business and a person, and it forms an opinion about whether you are worth surfacing.

In commerce, that model reads your catalog and decides whether to name you. In the inbox, it reads your email and decides whether it earns attention. In both cases the human is downstream of a machine that already made a judgement.

Think of it as the difference between a shop window and a concierge. For twenty years the job was dressing the window so passers-by would stop. Now there is a concierge at the door being asked "where should I go for this", answering from memory. Dressing the window still matters, but only after the concierge has a reason to say your name.

What actually changes on Monday

Three things, none of them dramatic.

• Write the first 200 characters for two readers. The emerging guidance for Gmail is direct subject lines, key information front-loaded, clear structure, no filler. That is no longer a copywriting preference, it is a delivery variable. Concrete and specific beats clever and vague, because one of your two readers has no sense of humour.
• Treat product data as marketing copy. One clean catalog now feeds ChatGPT, Copilot, Google AI Mode and Gemini. Vague titles, missing attributes and thin descriptions used to cost a little search traffic. Now they decide whether a machine can confidently recommend you at all.
• Re-anchor reporting on money. If opens inflate and clicks deflate for reasons that have nothing to do with your campaign, then opens and clicks have stopped being comparable across that boundary. Revenue per email and conversion per visitor still mean what they always meant.

The honest caveat

These are early numbers from interested parties. Shopify has every reason to publicise growth in a channel it has just started selling access to, and the Gmail figures come from third-party analyses rather than from Google. Multiples like 8x and 13x also flatter by construction when the starting base is small.

None of that makes the direction wrong. It means the size of the shift is less certain than the fact of it. The safest reading is that a new channel is real and growing quickly, an old one is getting quieter, and the metrics you have been using to watch both became less trustworthy at exactly the moment you need them most.