We’re not talking about the future of e-commerce anymore.
We’re in the AI era now where agents are already making buying decisions, summarizing reviews, comparing products, and nudging people to checkout.
The other day, I was shopping on nununu.com, but left my phone with my 8-year-old daughter so she could add products to the cart for me to review.
Ten minutes later, I found her deep in a conversation with the site’s AI assistant, asking about colour options, her size, best fit, and what was in stock.
That’s not how I shop. But for her, it’s natural. And it’s a glimpse into where shopping behaviour is headed.
Amazon’s Rufus. Klarna’s shopping assistant. TikTok’s recommendation engines. What my daughter did on Nununu — chatting naturally with an AI about sizes and colours — is exactly what these platforms are training shoppers to expect. They’re turning product discovery into a conversation, not a search task.
And these assistants aren’t experimental features. They are live, learning, and quietly rewriting the rules of how products are found and bought, and at least for the next generation of shoppers, this will be the default.
This is the new front door to e-commerce. And most e-commerce teams are still standing at the old one, wondering why fewer people are knocking.
PS: Find the bonus AI table at the end of the blog post that contains lots of experimentation AI tools you can use today.
The Questions Have Changed
It’s no longer “How do we get people to convert on our site?” It’s:
- How do AI agents interpret our product pages?
- Do they understand our value proposition?
- Can they connect it to a shopper’s intent?
- Are we giving them the signals they need to choose us?
There are also two major ways AI is influencing shopping right now, and each one changes who you’re competing against:
LLM-based shopping agents (think ChatGPT, Amazon Rufus, Klarna AI). These tools pull in multiple sources, compare options, and surface recommendations. Here, you’re competing head-to-head with other brands, often in the same result list, and the AI will rank based on price, clarity of value, and trust signals.
AI-driven experiences inside your own ecosystem (search, recommendations, personalization on your site). Here, you’re competing against your own content and product catalogue. The question is whether your data, descriptions, and reviews make your products the easiest and most logical match for customer intent.
If you don’t optimize for both contexts, you risk losing twice:
- First, when AI compares you to everyone else,
- Second, when your own systems overlook your products.
Big moves are being made in the AI e-commerce war:
- OpenAI (with massive reach behind it) integrated with Shopify, turning it from an information app to an end-to-end shopping platform.
- Amazon is outpacing all others by releasing 6 AI agents in 2025 (Alexa+ for buying research and transactions, Nova Act for autonomous actions across the web, and Buy For Me for buying non-Amazon products).
- Perplexity launched the ‘Buy with Pro’ feature, letting you buy online through an AI interface.
- Google already has a massive search data and user base, which would make it a formidable player if it adds shopping to Gemini.
The Problem: Most E-commerce Orgs are Still Trying to “Optimize Pages.”
The future is about optimizing for machines that interpret pages.
- If your product descriptions are vague, bloated, or generic, you’re invisible to the new gatekeepers.
- If your reviews are sparse or inconsistent, you’re untrusted.
- If your PDPs can’t clearly articulate who a product is for and why it matters, you’re not recommended.
- And if your site behaviour shows users bouncing from search to category to homepage and back again…
Guess what the AI will assume? You’re confusing. That’s the signal you’re training it on.
The Solution: 5 Ways Ecom Leaders Can Advance With AI Today
Here's how you optimize your pages for LLMs without spending too much time and energy. Ok, some of these methods may need more resources, but they're worth the effort.
Be Clear Not Clever
Cut the fluff. Your PDPs should shout: Who is it for. What it does. Why it’s worth it.
If the price is high, just showing “premium” isn’t enough you need to justify it.
For higher-priced or premium products, AI agents (and people) are looking for clear signals of value, such as:
- Sustainability & environmental impact: materials, certifications, low-carbon processes.
- Longevity & craftsmanship: details on durability, construction techniques, repair services.
- Performance proof: product testing standards, professional endorsements, awards.
- Heritage & brand story: authentic origins, cultural ties, sport/community roots.
- Community validation: reviews, UGC, stories from real customers or ambassadors.
If the price is high, “premium” isn’t enough — explain why in ways that both humans and algorithms can understand, verify, and repeat.
Speero’s Heuristic blueprint can help you here. It helps you quickly analyze your website or app via 5 heuristic themes:
- Value: does the content communicate the value to the user?
- Relevance: does the page meet user expectations in terms of content and design?
- Clarity: is the content/offer on this page as clear as possible?
- Friction: what is causing doubts, hesitations, uncertainties, and difficulties?
- Motivation: does the content encourage and motivate users to take action towards the goal?
Make The Search Teach Users
Internal search is a goldmine of real intent data, but most brands treat it like a side feature. For premium brands, search queries reveal category gaps and value perception issues:
- Are people searching “discount” or “sale” alongside your products?
- Are they looking for specific materials (“organic cotton hoodie”) or performance attributes (“waterproof hiking boots”)?
- Are they searching for comparisons (“vs competitor”)?
Fix the basics:
- Relevance,
- Filtering,
- Auto-suggest
But also feed this data into product, merchandising, and marketing decisions. That bounce after search? It’s not just a UX problem. It’s a trust problem.
Related resource: Speero ResearchXL methodology — our comprehensive research framework for identifying value propositions that resonate with both humans and AI.
Develop Behavioral Signals That Train AI
Product views, add-to-carts, scroll depth, and checkout abandons. These are the signals AI uses to judge whether your products deliver on the promise.
Premium brands can send stronger positive signals by:
- Pairing high-priced items with clear usage content (videos, 360° views, real-life scenarios).
- Using guided selling tools or quizzes to keep users engaged longer.
- Surfacing relevant cross-sells that match the same premium positioning.
If the data is messy, e.g., people clicking but bouncing instantly, you’re feeding noise, not proof.
Prioritize Insights, Not Just Shipping
Every test, every iteration is a training input for your team and for the AI agents interpreting your store.
For premium brands, this means running experiments that go beyond “will this button colour convert better?” to:
- Testing messaging that reframes price as value.
- Measuring whether highlighting sustainability or exclusivity lifts engagement.
- Comparing the conversion impact of lifestyle imagery vs. technical detail shots.
Solution Spectrum blueprint can help you here as it lets you group tests into three categories:
- Iterative tests are those small, frequent changes focused on optimization, like a button color test. They're good for quick wins but don't teach you much.
- Substantial tests are bigger changes that might involve new messaging or design.
- Disruptive tests are the "big swings," like experimenting with new products, pricing, or channels.
If you’re just shipping backlog items without measuring impact, you’re moving but not learning, Experimentation Decision Matrix Blueprint can help you with this by letting you clarify not just the outcome of a test but the action you plan to take afterward. It reminds you and your stakeholders that the goal is not just to "win" a test but to make good decisions that lead to change.
Build Experimentation Operating System (XOS)
You need an internal “brain” that connects product, marketing, and leadership teams to break silos and ensure that what you learn in one area shapes decisions everywhere else.
For premium brands, that OS should:
- Track what value messages resonate across different markets.
- Monitor which benefits (durability, design, sustainability) correlate most with higher AOV.
- Feed findings into both human-facing experiences and machine-readable data so AI agents surface your products more often.
Speero XOS Consulting — how we embed a high-velocity, insight-driven experimentation operating system across teams.
Conclusion
The brands that will survive this shift aren’t just creative — they’re systematic. They can learn fast, ship with confidence, and make decisions grounded in signal, not noise.
AI will change e-commerce whether you’re ready or not. But you can shape how it changes your brand. That starts with the fundamentals:
- Clear product stories
- Clean, structured data
- Consistent insight loops
- Bonus: Operational efficiency
Because here’s the thing: what happened on nununu with my 8-year-old isn’t a one-off. She didn’t think twice about asking an AI for size and colour recommendations; it was her first instinct.
In a few years, that won’t just be how kids shop. It will be how everyone shops.
AI will recommend what it can understand and trust what it can verify.
The question is: are you training it intentionally, or leaving it to guess?
This new reality starts with dynamic, conversational experiences and one of the easiest ways to start legeraging is by using some of the AI tools from chatbots to AI-powered shopping assistants. The key is to start with the right tools that help you not only manage these interactions but intentionally start training the AI.
Below, we've outlined some key players and their applications, with a focus on how they enable you to engage with customers in this new AI-first world.
.svg)
