AI Search Traffic Tripled. Organic Still Sends More.

Shopify’s second-quarter commerce data gave the AI search argument a real number to sit on.

AI-referred sessions to merchant storefronts grew 197 percent year over year. Organic search traffic grew 12 percent over the same period. And organic search still sent more traffic to Shopify merchants than every tracked AI platform put together.

Both halves of that are true at once, and the second half is the one most teams are skipping past.

A 197 percent rise on a small base is still a small number. A 12 percent rise on a very large base is a lot of visits.

The honest read is not that AI search is taking over, and not that it does not matter. AI referrals are a small, fast-growing, unusually high-intent slice of traffic, and the work that earns them is the same work that earns the big slice.

What Shopify actually reported

The analysis compared AI-referred and organic search traffic across merchant storefronts using sessions, orders and conversion rates by product category. The headline figures:

• AI-referred sessions grew 197 percent year over year in the second quarter.
• Organic search traffic grew 12 percent in the same period.
• Organic search still sent more traffic than all tracked AI platforms combined.
• In specification-heavy categories, AI-referred shoppers converted at roughly twice the rate of organic visitors.
• In broader categories, organic remained the larger discovery channel, while AI brought in about 1.3 times more first-time customers.

One caveat belongs next to all of that. Shopify did not disclose how many merchants or how many transactions the analysis covered. The direction is credible. The precision is not something anyone outside Shopify can check.

A growth rate is not a volume

Percentages flatter small channels and punish large ones, which is exactly why growth rates are the figure that gets quoted.

Shopify’s own earlier reporting makes the point sharper. On Q1 2026 data the company described AI chatbot referral sessions as growing more than eightfold year on year. The Q2 figure is 197 percent. Whether that is a genuine slowdown, a different set of tracked platforms, or simply a base that has started to mature, neither publication says. Nobody should build a plan on the second derivative of a number this young.

What has held steady across both quarters is the ranking. Organic search refers more sessions than every assistant combined. That is the sentence to write on the wall before anyone reorganises a content calendar around getting cited, which is a separate discipline with its own set of moves.

The conversion gap is a clue, not a scoreboard

AI-referred shoppers converting at twice the rate looks like proof that AI traffic is better traffic. It is really proof that AI traffic is later traffic.

Shopify’s Q1 read on the same behaviour is the tell. More than half of AI-referred sessions began on a product detail page, against roughly a fifth of organic search sessions. People are not arriving at the front door and browsing. They are landing on the shelf, holding the thing they already decided to look at.

The assistant did the comparison work that a shopper used to do across five tabs. So the visit that arrives has already survived it.

High intent and low volume are the same fact

This is the part that gets read as bad news and should not be.

An assistant filters. It reads twenty options and names two or three. Everyone it did not name gets no session at all, and everyone it did name gets a visitor who has skipped most of the deciding. Fewer clicks arriving further down the funnel is not a channel underperforming, it is a channel doing its job.

Which means AI referral volume can never look like search volume, no matter how good the underlying content gets. Judging it against organic search on raw sessions is judging a shortlist against a phone book.

Read the number as a quality signal about intent. Do not read it as a channel waiting to scale.

This is e-commerce data, and that limit is real

The figures describe shopping. They come from Shopify merchant storefronts, they measure sessions and orders, and the categories in question are product categories where a specification comparison is a natural thing for an assistant to run.

A B2B services firm, a football club, a publisher or a local operator should treat this as directional evidence, not as a measurement of their own market.

There is a second limit worth naming. A referral only exists when someone clicks. An assistant that answers a question fully, names your brand and sends nobody anywhere has still influenced a buyer, and left no trace in the referral column at all. Referral counts are a floor on AI’s commercial effect, never the full picture.

What most teams are about to get wrong

The panic move is to rewrite the content plan for machines. Shorter answers, more question-shaped headings, pages built to be quoted rather than read, and a quiet drop in the standard of everything else because the budget went somewhere.

That trade sacrifices the channel sending most of the traffic to chase the one sending least.

It is also unnecessary, because the inputs overlap almost completely. Clear structure, specific claims, a real point of view, consistent naming, evidence a reader can check: that is what ranks, and it is what gets cited. The system produces both. A strategy and audit pass that fixes what you publish and how it is structured will move organic search and AI citations at the same time, because they are reading the same signals for different reasons.

What good looks like

A team reading these numbers well does five things.

1. Keeps organic search as the primary traffic assumption, because on this evidence it still is.
2. Tracks AI referrals separately in analytics, so the number exists and can be watched rather than guessed at.
3. Judges AI traffic on conversion and revenue per session, never on session count.
4. Fixes clarity, structure and specificity once, and lets both channels benefit.
5. Reviews the split quarterly, not weekly. A channel this small moves violently on tiny absolute changes.

The discipline is refusing to let a fast percentage set the agenda for a slow, larger asset.

How NBK thinks about AI referrals

NBK’s position is that discovery keeps changing shape and the underlying job does not. Brands get named by people, by algorithms and now by assistants for the same reason: they are legible. They say clearly what they do, they say it consistently everywhere, and they publish enough specific material that the answer is easy to assemble.

Social sits inside that. Assistants and search engines both lean on how often and how consistently a brand shows up in places they can read, which is why being known in the first place does more for citation than any formatting trick.

None of that is bought. It is built by an operating system: positioning, content pillars, cadence, approvals that do not strangle good ideas, and reporting that tells you which of those is broken. Get the system right and the citations arrive as a consequence, not as a project.

Next step

If your social output feels busy but not effective, and you are being asked whether the plan needs rebuilding for AI search, the useful first move is to find the actual constraint before changing anything. Start with an audit of what you publish now, how it is structured and where it shows up. If the answer is that the content is thin rather than badly optimised for assistants, that is the work, and it pays on both channels.