AI Commerce Is Creating a Measurement Gap Most Brands Can’t See

There's a number getting passed around a lot right now. Adobe tracked more than a trillion retail site visits in Q1 2026 and found AI-referred traffic converting 42% better than every other channel. Revenue per visit was 37% higher and shoppers stayed 48% longer on site. The pitch that follows is usually the same: restructure your commerce strategy around this now.

Adobe’s number is real. But the conclusion is where brands need to be careful.

What those conversion rates measure are the people shopping through AI agents today, not how well the channel performs. Shoppers using ChatGPT or Perplexity are more digitally sophisticated and more purchase-ready than the average visitor.

Many were already close to a decision before they arrived. As adoption widens and AI shopping moves from early adopters to everyone, that early-adopter premium fades. It happened with mobile commerce. It happened with social commerce. It’ll happen here too.

That's not a reason to ignore AI commerce. It's a reason to measure it more honestly. As AI becomes part of more shopping journeys, the bigger question isn’t whether AI-referred traffic performs well, but whether your current model can see how AI influenced the sale.

What AI Commerce Does to Your Measurement and Attribution Model

Traditional attribution connects touchpoints to outcomes. A shopper sees an ad, clicks a listing, visits a product page, comes back through email, and buys. The model does its best to assign credit across that path.

AI changes the shape of that path. A shopper might ask ChatGPT, Perplexity, or Google AI Mode for the best smart thermostat compatible with Google Home under $150 that ships in two days. The agent compares product data, compatibility attributes, reviews, pricing, availability, and marketplace listings across multiple sources. It surfaces a recommendation, explains why that product fits, and sends the shopper to a retailer to complete the purchase.

Your reporting sees the last step. It may detect a referral from an AI platform. It may see a direct visit. Or a sale on Amazon, Walmart, or your DTC site. What it usually won't show is the full set of signals that helped the agent decide your product belonged in the recommendation set.

GA4 can report the session it sees. Last-click attribution can credit the final touchpoint. Marketplace dashboards can tell you where the transaction happened. None of those views are useless. They were just built around channel boundaries that AI shopping agents simply don’t follow. The influence may happen across channels while the credit goes to whichever channel closes the sale.

Three Gaps Your Current Tools Can’t See

1. The influence that happened before the visit

Standard analytics platforms measure visits, pages, events, and conversions. AI-influenced commerce often starts before any of those things happen.

The agent does the comparison work. It reads the product data, evaluates the reviews, checks availability, and weighs price across channels. Then the shopper arrives already informed, already narrowed down, and often already leaning toward a choice.

By the time that shopper reaches your site, your analytics platform is measuring the end of the decision, not the full decision. This is partly why AI-referred traffic can look unusually strong. The shopper who arrives from an AI recommendation may have already done the equivalent of several comparison visits inside the AI experience. Your site gets the conversion, but your reporting misses the evaluation that created it.

Here’s the implication: if paid media performance looks flat while overall revenue holds, the answer may not be that your campaigns became less efficient. It may be that more of the decision-making is happening in a layer your current reporting only partially captures.

2. The first-party data that never gets collected

 

When an AI-mediated purchase bypasses your product pages and checkout flow, the brand loses more than a visit. It loses the behavioral data that normally comes with it. No email capture. No browsing behavior. No preference signal. No loyalty touchpoint.

For brands that have spent years building direct relationships with customers, this is the quieter threat. It's not that AI commerce is performing badly. It's that it's performing well in a way that leaves no residue in your CRM, your loyalty program, or your retargeting audiences. The relationship you built disappears at the moment of transaction. 3. The channel definitions that no longer hold

Most commerce reporting is organized around channels: Amazon, Walmart, DTC, paid search, email, etc. Each has its own dashboard, its own metrics, and its own definition of what a conversion means. AI agents don't care about those definitions. They evaluate products across all of them simultaneously and surface whatever clears the confidence threshold.

When a purchase happens because an agent evaluated your Amazon listing, your Walmart pricing, and your DTC schema all at once and made a recommendation, which channel gets credit? Under current reporting structures, the answer is arbitrary. It depends on where the transaction completed, not on what actually influenced it.

What Good Measurement Looks Like Now

AI commerce measurement is still catching up to how AI commerce actually works. But you don’t have to wait for the platforms to solve this problem to get ahead of it. There are concrete things that produce better signal right now.

Connect Commerce Data Across Channels Before Trying to Attribute It

The root cause of most AI attribution problems is the same root cause as most commerce measurement problems: data lives in too many places with too many conflicting definitions. Amazon defines a conversion one way. Walmart defines it another. Your DTC platform defines it a third. Before you can measure AI's influence across channels, you need one normalized view of what is happening across them. That's not an AI problem. It's a data infrastructure problem that AI just made urgent.

Measure Inputs, Not Just Outcomes

If AI agents are evaluating catalog signals to decide what to recommend, then catalog signal quality is a leading indicator of AI commerce performance. Attribute completeness rates, pricing consistency across channels, inventory accuracy, schema health, review architecture: these are measurable, actionable, and they predict AI visibility before the sale happens. Instrumenting these signals gives teams something to optimize against, rather than waiting on conversion data to explain what already occurred.

Measure Whether You’re in the Consideration Set

In traditional search, visibility was easier to observe. You could see where you ranked, how often you appeared, and which queries drove traffic. AI shopping agents make the feedback loop less visible. A product is either included in the recommendation set or it isn't. If it's omitted, there may be no ranking drop to diagnose and no impression decline to explain.

That means measuring presence differently. Prompt testing, AI visibility audits, catalog signal scoring, and competitive recommendation tracking all become part of the measurement system. They don't replace revenue reporting, but they help explain why revenue is moving. If your product isn't appearing when shoppers ask high-intent questions in your category, the problem may not show up as a traffic decline right away. It may show up later as missed demand your current tools never knew existed.

Treat AI Traffic as a Distinct Segment, Not a Channel

The most practical move is to identify AI-referred traffic explicitly where you can. Referrer strings from ChatGPT, Perplexity, and Google AI Mode are often readable. Segment that traffic separately and analyze its behavior independently of other traffic. The conversion rates, revenue per visit, and session patterns are genuinely different. Averaging them into your overall numbers buries the signal.

The Measurement Gap Is Widening Every Quarter

AI commerce isn't a future consideration anymore. Adobe's data shows AI-influenced traffic to retail sites grew 393% YoY in Q1 2026. That growth is already inside your revenue. It's showing up as direct traffic, last-click conversions, and channel sales with no clear origin story.

You may be running your commerce program on measurement and attribution models designed for a different era of shopping behavior. Those models aren’t wrong. They're just increasingly incomplete. And as AI-influenced commerce grows as a share of your revenue, the gap between what your reporting shows and what’s actually happening gets wider every quarter.

The good news is that closing the gap doesn't require a new analytics platform or a perfect attribution model. It requires connected data, consistent definitions, and a clearer view of where your products appear when AI systems are making recommendations before the shopper ever reaches your reporting.

Channel Key helps brands build that picture. If you’re not sure what your current measurement is missing, that’s where we’d start.

Let's Talk