I Think AI Is Creating a New Layer Between Discovery and Purchase
I think most people are looking at AI commerce through the wrong lens. The discussion usually starts with visibility. Can AI find my website? Can AI read my products? Can AI understand my content?
Those questions matter.
But they all happen before the interesting part. The interesting part is the recommendation. When someone asks an AI system: best hydration brand, best running shoes for flat feet, best protein powder. The system has to make a choice.
It doesn't recommend every store.
It recommends a few.
I keep thinking there is a new layer emerging between discovery and purchase.
A recommendation layer.
The businesses that understand how this layer works may have a significant advantage over the next few years.
Still early.
Still learning.
Anyone else building in e-commerce is thinking about this.
The discovery layer is shifting, but exact language still matters. For Kinetic Override, the same app can be searched as “Android auto clicker”, “macro recorder”, “record taps”, or “no-root gesture replay”. If AI mediates discovery, those job-to-be-done phrases become even more important than a polished category label.
Agreed. Those job-to-be-done phrases are probably becoming even more important as AI mediates discovery. What's interesting to me though is that understanding the query is only half of the equation. An AI might understand that "Android auto clicker", "macro recorder", and "gesture replay" all describe the same intent. The next question is why it recommends one product over five other products that also satisfy that intent. Understanding seems to determine eligibility. Recommendation determines winners. That's the layer I'm most curious about right now.
What gets me is you can't really optimize for it directly. You can't buy your way in, you can't tweak a meta tag.
The model has read everything written about you — forums, reviews, random Reddit threads from 2022. Your own website might be the least useful input.
Which is kind of terrifying and kind of fair, honestly.
That's uncomfortable for marketers used to controlling the message. And genuinely interesting for smaller products that punch above their weight in communities.
That's exactly what makes it interesting. In the search era, most optimization happened on the website itself. In the recommendation era, the website becomes just one signal among many. What AI knows about a brand may come from reviews, communities, mentions, customer experiences, articles, videos, and thousands of small signals spread across the web. Maybe the shift is not from search to AI. Maybe the shift is from optimizing pages to understanding how entities are evaluated and recommended.
"Optimizing entities" is probably the right frame. And most founders have no idea what signals they're actually sending..
I think that layer is true, and it changes the game because distribution isn't just about getting seen anymore. It's about giving the model enough clear signals to feel confident picking you over the other ten decent options
My guess is the winners won't just be the biggest brands. It'll be the companies that make their positioning painfully easy to understand across reviews, comparisons, communities, and use cases, because that's the kind of evidence an AI can actually lean on when it has to recommend something
Confidence is a really interesting way to frame it. Discovery gets you into the consideration set. Confidence is what gets you recommended. I'm starting to think that's where a lot of the real opportunity sits.