Four minutes, three options, one purchase. Inside the new path to purchase, and why observed behavior is the only way to see it.
Last month I went looking for running shoes for my son. He overpronates, which means the wrong pair is not just uncomfortable. It is a knee problem waiting to happen.
The old version of this task went something like this. Open Google. Sift through nine listicles that are all affiliate content in a trench coat. Read a forum thread from 2019. Get retargeted by three shoe brands for the next two weeks. Give up and buy whatever Amazon put first.
Here is what happened instead.
I described the problem to ChatGPT. It asked a couple of clarifying questions. Then it gave me three specific models with links, and an explanation of why each one worked for overpronation.
I picked one of the three. I did not check a comparison site. I did not read a review.
The shoes were right. My son is happy. Total elapsed time, roughly four minutes.
Then the researcher in me sat up. Because I had just walked a path to purchase that almost nothing in our industry’s measurement toolkit can see.
What actually happened in those four minutes
Look at what got compressed.
Problem definition, needs analysis, product research, shortlisting, comparison, and decision. Six stages of a classic purchase funnel, collapsed into one conversation, with no tab switching and no backtracking.
And notice what was different about the input. I did not search for a product. I described a problem. My son overpronates and needs a running shoe. That is not a keyword. It is a brief.
This is the shift that should have every insights team paying attention. The shopper is no longer translating their need into the language of a search engine and hoping for the best. They are stating the need directly and receiving a shortlist built around it.
OpenAI has leaned into this deliberately, with a shopping experience designed to ask clarifying questions and assemble a personalized buyer’s guide rather than a list of links. The category is moving the same direction across the board.
For shoppers, this is a straight upgrade. Less time, less noise, less second-guessing. For brands, it is the most significant change to the consideration stage in twenty years.
The third turning point
Search was the first one. It moved the shopper from asking a salesperson to asking a query box, and it handed marketers a beautiful gift in return: keywords, referrers, clickstream. Intent became visible.
Social was the second. The feed collapsed discovery and impulse into the same three seconds, and the influencer became a functioning retail channel. That shift is still going strong. It is not being replaced by AI, and anyone writing it off is getting ahead of the data. We see that in observed behavior every day.
AI is the third. And it differs from both predecessors in one specific way that should worry every measurement team.
It leaves almost no trail.
No impression to count. No referrer worth the name. No comparison set you can reconstruct afterward. The consideration phase, the part of the journey brands have spent decades trying to influence, now happens inside a conversation that nobody outside it can observe.
Three brands were named in my sneaker search. Two of them lost. Neither of those two will ever know they were in the running.
That is the whole problem in one sentence. The most decisive moment in the journey has become the least visible one.
You cannot survey your way to this answer.
Ask me in six weeks how I found those shoes and I will give you a confident, useless answer. Something like “I did some research online.” Or worse, “I already knew the brand.”
Both would be wrong. Neither would be a lie. That is the say-do gap doing exactly what it always does.
Self-reported AI usage carries an extra layer of distortion on top of ordinary recall failure. Some people inflate it, because using AI signals that you are current. Others play it down, because it feels a bit like cheating. And most people cannot reconstruct what they typed into a chat window three weeks ago, in what order, or what came back.
So we have an industry filling up with studies about AI adoption, built almost entirely on the least reliable input available. Memory.
Meanwhile, the behavior is sitting right there. Observable. Timestamped. Unambiguous.
What we are watching
The questions we think matter most over the next year:
- How often does an AI conversation replace the comparison stage entirely, and in which categories does it not?
- When AI names three brands, which one wins, and does position matter the way it did in search?
- Does the AI path shorten the journey, or does it just relocate the research?
- What does an AI-influenced basket look like compared to a search-influenced one? Bigger, faster, more or less brand loyal?
- Where do AI and social overlap? Is the influencer recommendation getting verified in a chat window before purchase, or the other way round?
- And the one everyone wants: is AI-driven discovery pulling volume away from search and social, or adding a layer on top?
Every one of those is answerable through observed behavior. Not one of them is answerable through recall.
Coming to Qriousity
This is why we built the AI Measurement module, launching soon inside Qriousity.
It observes consumer behavior inside AI. LLM usage across ChatGPT, Gemini, Perplexity and more. AI-driven discovery. In-chat shopping. Passively collected, first-party, continuously updated.
Alongside it, Shopper Journeys covers the full path to purchase inside Amazon, Walmart and Target. Search, product views, basket activity, checkout. Put the two together, and you can follow a shopper from the question they asked an AI to the thing they actually bought.
Three turning points into this, we finally get to watch one happen live instead of reconstructing it afterward from what people remember.
The truth is in the behavior.
Want to see what your category looks like in observed behavior? Request a Demo: https://clr0vty6cdm.typeform.com/to/prdbQcyu?typeform-source=www.linkedin.com