I Think Most E-commerce Brands Are Optimizing The Wrong Layer
I've been spending a lot of time thinking about how AI is changing e-commerce.
Most conversations focus on visibility.
Can ChatGPT find my brand? Can Gemini mention my products? Can Perplexity recommend my store?
Those questions matter.
But I'm starting to think they're downstream effects.
The more interesting question is - what happens before a recommendation is made?
When someone asks AI for for example the best running shoes for flat feet, the best protein powder, and the best coffee grinder under $200???
The AI has to make choices.
It doesn't recommend every brand. It doesn't show every store. It narrows the market.
That's the part I think most people are underestimating.
For the last 20 years, e-commerce has largely been about influencing humans.
SEO... Ads...Landing pages...Conversion optimization...
Now there seems to be a new layer emerging between discovery and purchase.
A recommendation layer.
A layer where AI systems evaluate businesses before customers ever see them.
What's interesting is that AI evaluates stores very differently from humans.
Humans see: design, branding, emotion, aesthetics
AI sees: structured data, product information, semantic clarity, trust signals,
extractable answers
Which means a store can look amazing to humans and still perform poorly in AI-driven recommendation environments.
I suspect the next generation of e-commerce winners won't simply be the brands with the most traffic.
They'll be the brands that are easiest for AI systems to understand, trust, compare, and recommend.
Still early.
Still learning.
Anyone else building in e-commerce or AI is seeing the same shift.
What stands out to me is that a lot of founders are still optimizing for attention while AI is optimizing for clarity. A flashy landing page might impress a visitor, but it won't help much if an AI can't clearly understand what problem you solve and who you solve it for.
I think clarity is a big part of it. What I'm trying to understand is whether clarity alone is enough. An AI can clearly understand ten different stores selling the same product. It still has to decide which three to recommend and which seven to ignore. That decision process feels like the next layer to me. Visibility is about being found.
Clarity is about being understood. Recommendation is about being chosen. I suspect most of the interesting work over the next few years will happen in that last category.
The layer split applies to mobile utilities too. With Kinetic Override, the feature layer is Android tap/swipe macro replay, but the adoption layer is trust: users first want to know why Accessibility permission is needed, what stays local, and whether there are ads/accounts.
That's a good example. The interesting part is that the macro replay feature gets the product into consideration, but trust determines whether someone actually adopts it. I keep wondering if AI recommendations work in a similar way. Capabilities may determine eligibility, but trust signals may determine selection. Two products can solve the same problem equally well. The one that appears more trustworthy, explainable, and verifiable may be the one that gets recommended.
There's an actual paper that named this, "GEO: Generative Engine Optimization", worth a look. Still stuck on measuring it properly though, so the "where does the pain show up first" comment really resonated.
this is the answer for your first reply... :) ....That's actually an interesting parallel. Early user behavior helped reveal the product. I suspect recommendation behavior will reveal a similar thing for commerce. We'll probably discover that many brands optimize for what humans notice while AI systems optimize for what they can understand.
Agreed. GEO gave the problem a name. The next challenge is measurement. Brands won't care about recommendation visibility until they see competitors being recommended instead of them.
I think you're onto something. SEO optimized for search engines, CRO optimized for humans, and now we're entering a phase where brands may need to optimize for AI evaluators.
What's interesting is that many of the signals AI prefers clear product data, transparent policies, authentic reviews, detailed comparisons, structured content are also signals of a genuinely good business. In that sense, AI recommendation engines could reward substance over marketing polish.
The question I'm curious about is whether this becomes a winner-take-most dynamic. If AI narrows thousands of options down to 3-5 recommendations, visibility becomes far more concentrated than traditional search ever was.
Feels like we're watching the emergence of "AI Optimization" as a discipline distinct from both SEO and conversion optimization.
Agreed. Search expanded choice. Recommendations compress choice. That's why recommendation share may become as important as market share in the AI era.
One thing I'd be careful with:
A lot of people will agree that a recommendation layer exists.
The harder question is where the commercial pain actually shows up first.
The risk is building a framework that feels obviously true, but where the buyer, urgency, and expensive consequence are still too abstract.
I wouldn't make that call casually in-thread because the answer changes who this is really for and what problem they're paying to solve.
That's a fair point. My current hypothesis is that the pain shows up when businesses realize they're being excluded from consideration before customers ever reach their website. In search, you could often see the problem through rankings and traffic.
In AI-driven recommendation environments, the challenge may be that the loss of visibility is much harder to detect. Still early, but that's the part I'm trying to understand. :)
Possibly.
I think the risk is making the wrong commercial interpretation too early.
The useful part isn't the observation itself. It's the decision that follows from it.
I wouldn't make that call casually in-thread because it changes who the buyer is, what problem they're actually buying, and how the product gets evaluated.
If you're open to it, drop your email and I'll put the tighter version together properly.
Agreed. The real signal won't be visibility. It will be when businesses start asking why competitors are being recommended and they aren't. That's usually when new categories emerge.
This comment was deleted 16 hours ago.