AI is Moving the Sale Earlier Than Your Reporting Can See

A friend of mine, a schoolteacher, was visiting me recently and asked whether AG1 Greens was better than IM8, the supplement he takes.

I pulled up Google’s AI Mode and compared them while we were talking. It asked whether he wanted nutritional substitution or longevity, since IM8 is better for longevity. That’s not what I would have thought to ask about, but the model knew to ask it.

“How did you know that?” he said.

“I’m looking at it. I’m doing the research right now.”

It took two seconds.

A few weeks later, I was looking for shoes for somebody recovering from a knee replacement. I wanted support, cushioning, and something easy to get on and off. The model compared the options much faster than I could have done on my own, and it factored in things I hadn’t even listed, like grip on hard floors.

That answered what to buy. Then I asked where to buy it, and the model kept going.

Amazon.com won because Prime shipping was cheaper and its return process is painless. For my buddy’s supplement, the brand's own site won because the subscription price was significantly lower than any retailer's. Then artificial intelligence found a discount code, too.

That’s how I research almost anything now. I barely use search. If I’m going to make any purchase, AI works out what to buy and where to buy it.

The problem for brands is that all of it, the longevity question, the shoe comparison, the discount code, happens before I ever visit their website, so none of it shows up on a retailer's dashboard.

The Click Was Never the Whole Story

I've spent years watching companies credit the last visible touchpoint in a purchase and miss everything invisible that led up to it. This same situation is happening with AI search.

I saw the first version of this problem 20 years ago, at Overture, the search advertising company that pioneered pay-per-click ads before Google adopted the same model. Someone saw a search ad, clicked it, and converted. The last click got the credit because it led directly to the conversion.

Position in search results mattered, but winning the position never meant you had won the customer. Mobile made that obvious. A customer taps an ad for a specific product, gets sent to the App Store, downloads the app, and lands on the homepage. The product is gone, the promotion that got them to tap is gone, and nobody knows why the consumer showed up in the first place.

The company paid to create that intent, then threw it away at the exact moment it mattered most.

Branch has spent years fixing that handoff between web and app, and the same problem is resurfacing with AI, except the stakes are higher. An AI agent may already know what the customer wants, what they're willing to spend, where they're located, and whether they need shipping or pickup. That journey exists before the customer ever gets to the retailer's site.

Send that customer to a generic homepage and they have to explain their purchase twice: once to the agent that already understood it, and again to a website that has no idea an agent was involved.

Whatever click eventually closes the sale will still get full credit for it, and the work that actually won the customer will disappear the same way it did in search.

The Real Work Now Happens Before Anyone Clicks

Purchase decisions are getting made earlier now, inside conversations with large language models (LLMs).

A customer describes what they need, and the model does the comparison shopping that used to take a person 10 browser tabs to do badly. It weighs several brands or products, recommends one tailored to their needs, and often the customer’s eventual purchase looks direct, with all the context left behind.

That means a brand has to be understandable to the machine before a customer ever arrives on its site. If the product information is wrong, the inventory is stale, or the current offer is missing, that brand may get filtered out of consideration before a human ever sees it.

Paid AI ads leave more of a trail. A good mobile measurement partner (MMP) should show whether that customer reached the right product, installed the app, and converted.

However, paid placements are still only part of the picture.

LLMs are going to want credit for all the research and discovery, even when the conversion happens somewhere else entirely. Marketers who keep measuring channels as isolated point solutions will miss most of what's actually happening.

What they need is a complete view of the customer journey, or increasingly, the agent acting on the customer’s behalf.

Test the Journey Yourself Before Peak Season Finds the Gaps for You

Peak season is where all of this gets expensive fast, and Black Friday is the worst of it. When prices change hourly and inventory moves quickly, an agent might recommend one retailer in the morning and a competitor by that afternoon.

So run the journey yourself, now, before traffic peaks. Ask the major LLMs the same questions your customers are about to ask and see whether your brand shows up.

Then, when the model recommends your brand, chart your own journey all the way to purchase. Watch for the point where context was lost and reporting became a black box.

If your reporting still stops at the same place it always has, you won’t be investing wisely. And if you make your customer describe what they want without providing a direct link, you’ll keep losing sales without knowing why.

David Karnstedt is CEO of Branch, which delivers deep linking solutions that work across environments, along with AI-powered measurement to help organizations optimize advertising spend and improve ROI.