AI, Personalization, & Generative Search: What eCommerce Teams Still Get Wrong

A Conversation with Liberty Safe’s Kasey Wiese

Connect with Kasey Wiese on LinkedIn

AI is now part of almost every conversation about eCommerce. Platforms promise smarter targeting, automated content, predictive merchandising, and better performance across the funnel.

But inside most marketing teams, the reality looks less polished. The tools are evolving quickly. The thinking behind them often isn’t.

Kasey Wiese, Director of Ecommerce at Liberty Safe, spends a lot of time thinking about that gap. She’s been outspoken about where AI is genuinely helping eCommerce teams and where the industry is still repeating the same mistakes, just faster.

Her view is simple: the biggest gains from AI won’t come from adopting more tools. They’ll come from stronger data foundations, better judgment, and teams willing to challenge their own assumptions.

“The old model treated every click like a confession.”

Personalization Still Gets the Story Wrong

Personalization has been a staple of eCommerce marketing for years. In theory, the idea is straightforward: Learn what customers want and show them more of it.

In practice, it rarely works that way.

Wiese recently experienced the problem firsthand. After buying a baby shower gift for a friend, nearly every platform she interacted with began treating her like a new parent.
Ads shifted.
Product recommendations changed.
Entire storefronts reorganized around a life event that didn’t apply to her.

The signals themselves weren’t wrong. One purchase had happened, but the systems interpreting that signal rarely have the full context.

They reacted to the event instead of the situation.

That pattern still shapes how many personalization systems operate today. Each interaction becomes a lasting signal of intent, even when it represents a one-time action.

Customers notice when recommendations feel disconnected from reality.

Where AI may finally start to improve personalization is in how signals are interpreted. Instead of reacting to a single action, AI systems can analyze patterns across multiple behaviors.

That allows teams to ask better questions. Was the purchase part of a broader pattern? A one-time event? A gift?

Useful personalization depends on that kind of context.

Wiese is cautious about declaring the problem solved.

“I don’t think anyone has fully cracked it yet.”

The tools available today are improving quickly, but they’re still evolving. In many cases, the real advantage isn’t choosing the perfect platform. It’s building teams that understand what good personalization should look like in the first place.

Without that judgment, better tools simply automate the same mistakes.

Instead of reacting to a single action, AI systems can analyze patterns across multiple behaviors. The real advantage isn’t choosing the perfect platform. It’s building teams that understand what good personalization should look like in the first place.

Where AI Is Actually Helping eCommerce Teams

Despite the hype surrounding AI, there are areas where it’s already producing real value.

One of the clearest examples is site merchandising.

Historically, merchandising decisions inside eCommerce companies often came down to internal priorities.

AI merchandising tools approach the problem differently. They analyze browsing behavior, purchase patterns, demand signals, and inventory levels at the same time.

That allows product placement decisions to reflect customer demand rather than internal assumptions.

When those systems are allowed to operate, the results can be meaningful.

But in many organizations, human behavior still overrides the data.

AI identifies the products most likely to convert. The system adjusts placements accordingly. Then a meeting happens.

Someone wants to push a different SKU. A promotion needs priority. A product that isn’t moving becomes the new focus.

“The moment AI makes a smart merchandising decision, someone says we need to move this SKU to slot one.”

And just like that, the system gets overridden.

It’s a reminder that the biggest barriers to AI adoption are rarely technical. They’re organizational.

Measurement is another area where teams are still trying to catch up.

Many companies are experimenting with AI tools across marketing and merchandising, but attribution models haven’t fully adapted to how those systems operate.

The result is a growing number of dashboards and pilot programs designed to measure AI’s impact without necessarily changing how teams work.

As Wiese puts it, teams sometimes end up “buying tools to measure the tools.”

Testing new technology is necessary. But experimentation only creates value when it eventually leads to decisions.

The biggest barriers to AI adoption are rarely technical. They’re organizational.

Generative Search Is Changing Discovery

One of the biggest shifts happening in eCommerce right now isn’t happening on websites. It’s happening in search results.

Generative search engines are increasingly summarizing product options, comparing brands, and answering shopping questions directly on the results page.

In many cases, consumers receive recommendations before they ever visit a brand’s site.

That change has pushed marketing teams to start thinking about generative engine optimization, or GEO.

Many brands are still figuring out what that shift means for visibility across search and AI platforms. For teams trying to understand how generative discovery works and how brands appear inside AI answers, Go Fish explored this in AI in Advertising: How Google AI Search, Paid Media, and ChatGPT Ads Are Reshaping Discovery.

But according to Wiese, many teams are approaching the problem backwards.

Before worrying about how content appears inside AI-generated answers, companies need to focus on the foundation those answers rely on.

Clean product data. Structured content. Consistent messaging across channels.

Those signals influence how AI systems interpret a brand’s credibility.

Without them, optimization tactics alone won’t make much difference.

Another misconception is the continued focus on traditional keyword rankings.

For years, the goal of SEO was simple: rank first.

But the rise of AI summaries is pushing traditional results further down the page.

“If ranking number one is still the entire strategy, you’re optimizing for a world that doesn’t exist anymore.”

The visibility that increasingly matters is whether a brand appears inside the AI-generated answer itself.

Those citations tend to reflect broader authority signals rather than a single ranking position.

For teams looking to understand how those signals influence AI search results, Go Fish explored the topic further in the whitepaper Winning with GEO: How Leading Brands Drive Growth on Google’s Changing SERPs.

Consumers now receive comparisons, recommendations, and research summaries without visiting multiple websites.

That shift is already visible in how shoppers research purchases across platforms. In How Consumers Use ChatGPT + Google AI to Discover, Compare, and Buy Retail Products, Go Fish breaks down how AI is compressing discovery and comparison earlier in the buying process.

For brands, that means the moment of trust often happens before someone lands on their site.

Consumers now receive comparisons, recommendations, and research summaries without visiting multiple websites.

AI Content Still Needs Human Judgment

AI has made it easier than ever to produce marketing content quickly. Product descriptions, FAQs, metadata, and blog drafts can now be generated in minutes.

But speed isn’t the main challenge eCommerce brands face.

Trust is.

“In high-consideration eCommerce, trust is the product as much as the product itself.”

That becomes obvious when AI-generated content creates inaccurate expectations.

Product imagery is one example.

If a customer orders a couch in evergreen and the product arrives looking noticeably different from the listing image, the issue goes far beyond marketing.

It becomes a returns problem. A reviews problem. A brand problem.

That’s why accuracy matters as much as efficiency when companies experiment with AI-generated visuals.

On the written side, AI can be more useful. FAQ sections, product descriptions, and structured metadata benefit from speed and consistency.

But producing more content doesn’t automatically improve visibility. As search platforms shift toward AI-generated summaries, many eCommerce teams are rethinking how AI should fit into the way marketing teams actually work. Go Fish explores that in AI Adoption Isn’t a Tool Decision. It’s an Operating Model Decision.

For now, Wiese prefers a newsroom mindset.

“A paper doesn’t go to print without an editor.”

AI can speed up the workflow, but the final judgment still needs to come from a person.

“In high-consideration eCommerce, trust is the product as much as the product itself.”

The Teams That Improve Won’t Be the Ones With the Most Tools

Despite the explosion of AI tools across marketing, Wiese believes most organizations are still trying to find their footing.

“Most brands are on their heels right now.”

The pace of change makes that inevitable.

The teams that succeed won’t necessarily be the ones with the most sophisticated technology stacks. In many cases, they’ll be the teams that learn faster.

Testing new approaches. Talking honestly about what didn’t work. Adjusting quickly.

That kind of transparency isn’t always easy in marketing environments where decisions are tied to forecasts, budgets, and executive expectations.

But leadership sets the tone.

“If a leader isn’t willing to say ‘I got that wrong,’ no one else on the team will either.”

In an environment where AI systems sometimes produce strong insights and sometimes produce confident nonsense, that willingness to reevaluate becomes a competitive advantage.

Because the goal isn’t perfection.

It’s progress.

What This Means for eCommerce Teams

A few patterns emerge when you step back from the individual tactics.

Personalization still needs context. Treating every click like a permanent signal of intent is what created the problem in the first place.

AI works best when it informs decisions rather than replacing them. Tools can surface strong signals in areas like merchandising and content strategy.

Generative search is shifting where discovery happens. Visibility inside AI-generated results is becoming just as important as traditional rankings.

Trust still determines whether AI content works. Speed helps, but accuracy matters more.

And perhaps most importantly, the teams that improve fastest will have the advantage. The organizations that test ideas, learn from mistakes, and adjust quickly will move ahead of the ones waiting for certainty.

Want to Understand How AI Is Changing Search Visibility?

Generative search is reshaping how customers discover brands. AI answers, summaries, and recommendations are now influencing buying decisions before someone ever visits a website.

If you’re trying to understand how your brand appears across Google AI Overviews, ChatGPT, and other generative search platforms, start with AI in Advertising: How Google AI Search, Paid Media, and ChatGPT Ads Are Reshaping Discovery.

Or get a clear read on your own visibility.

Our AI Visibility Audit shows how your brand appears in AI search, where competitors are winning, and what to fix first.