Furniture.com's VP of Search and Discovery on the Taste AI Can't Replace

Furniture is not a keyword. It is fabric under a hand, a rug that has to match a room nobody else has seen, a sofa a guest will judge before they say a word about it. That is the problem with folding furniture search into the SEO-versus-AEO debate: the framing assumes the hard part is getting a language model to describe a product correctly. It rarely is.

Sebastian Pflumm, VP of Search and Discovery at Furniture.com, spoke exclusively with ClickZ about what that framing misses, and why his team spends less time optimizing for algorithms than protecting the parts of shopping an algorithm cannot do for someone.

A .com name is a liability before it is an asset

Furniture.com’s own domain works against it before any search strategy comes into play. “You think of Furniture.com and it feels like an old, dated thing that has probably been around for 30 years,” Pflumm said. Unaided awareness is low: few shoppers name the site unprompted when asked where they’d go to shop for furniture. Aided awareness runs the other way. Shown a list of retailers, people recognize the name instantly, then admit they don’t actually know the business behind it.

That gap shapes the entire discovery strategy. The AEO debate treats visibility as something to win from a search engine or a model. Pflumm’s team treats it as something to build directly with the shopper, since the model’s version of Furniture.com is downstream of whether people already understand what the site is.

The .com in the name has nothing to do with AEO

Pflumm draws a clean line between the brand’s legacy naming problem and the technical work of AI-era discovery. “The .com itself doesn’t matter for AEO or SEO just because it’s Furniture.com,” he said. What matters is representation on two fronts: how general-purpose models like ChatGPT, Gemini, and Claude describe the brand, and how Furniture.com’s own shopping agent, Dottie, represents it directly to customers.

The first front runs through structured groundwork and PR, making sure there’s accurate knowledge of the company circulating for models to draw on. The second is closer to brand-building than engineering. “We spend time, and this is a collaboration between machine learning and marketing and brand building, to craft the personality of our own agent,” Pflumm said, describing what the team calls a soul.md file: a personality document that gives Dottie instructions for representing Furniture.com’s tone, not just its catalog.

Product pages were never going to be enough

The version of this conversation that treats discovery as a formatting exercise, get the product page right and let the model do the rest, misses what drives a furniture purchase. “A picture’s worth a thousand words,” Pflumm said, and for furniture that holds more literally than most categories. Shoppers need to see scale in a room, feel a texture, sometimes walk into a store, because the object is going to live somewhere other people see it.

Language models remain heavily text-based in how they reason through a shopping journey. That is precisely the gap a text-only optimization strategy leaves open, since so much of a furniture decision depends on the visual and the tactile rather than the described.

Handing the decision to an agent has a real cost

Pflumm is direct about the trade-off shoppers make when they let an AI agent choose on their behalf. Done well, an agent saves time and widens what a shopper sees without forcing them through page after page. But speed is not the only variable.

“If you leave the decision entirely to the agent, then you give up the taste, the taste of what you want your home to look like,” he said. “That’s something you should not yet cede to the machine.”

He is equally clear about which side of that trade-off separates winners from losers over the next two years. Asked to finish the sentence “the retailers who lose ground will be the ones who kept treating discovery as,” Pflumm answered without hesitation: “something that they cede to the LLMs.” Data narrows the field of options, in his view, but instinct still closes the decision once the hard constraints, the dimensions, the fit, the guardrails, are already satisfied.

Price is the last question, not the first

Pflumm pushed back on a common misread of furniture shopping behavior. US furniture buying is famously promotion-driven around specific calendar moments, President’s Day mattress sales among them, which fuels an assumption that price leads the decision. It rarely does. “The first problem you want to solve is what is the mattress that’s right for my sleep and my back and my partner,” he said, “and then towards the end of the journey is how do I get the best price on it.” Teams that start with the promotion, he argued, are solving the easier problem before the harder one that actually earns the sale.

A decade on both sides of the P&L changed how he thinks about growth

Before joining Furniture.com, Pflumm spent a decade advising retail and consumer companies at McKinsey. Sitting on both the finance and product sides of the business since, he has settled on what he calls numerator thinking. “You cannot save your way to success,” he said. Cutting cost shrinks the denominator of a fraction, but it does nothing to grow the numerator, the actual value a product delivers to a shopper. Furniture.com’s priority, in his framing, is making that numerator as large as possible while staying disciplined on cost, not treating cost-cutting as a growth strategy in itself.

That instinct, protect what makes the product genuinely valuable before optimizing anything else around it, is the same one running through his approach to AI. The retailers who win the next two years of discovery won’t be the ones who out-format their product pages. They’ll be the ones who worked out which parts of the shopping journey are worth keeping human.