I Am Going To Say It Out Loud

Every AI shopping startup has the same beautiful problem

“It’s all a fugazi. You know what a fugazi is? It’s fake. It’s not fucking real.” — Mark Hanna, The Wolf of Wall Street (2013)

The average AI shopping platform burns $3.40 to earn $1 in affiliate revenue. The smartest investors in tech are lining up to fund more of them.

Bonus: Eight Questions That Kill Every AI Commerce Pitch¹

Let me say what everyone in venture is thinking but won’t admit: every AI shopping agent is building the exact same doomed business. Beautiful technology, impressive demos, fundamentally broken economics. And we all know it.

I know this story because I’ve lived it. As co-founder of POPSUGAR and former CEO of ShopStyle, I watched a generation of discovery platforms die. The technology was never the problem. The problem was pretending technology could override arithmetic.

AI doesn’t fix broken unit economics. It makes them more expensive.

Every conversational commerce startup shares the same fatal architecture: They aggregate inventory they don’t own, from retailers they don’t control, through channels they have to rent, to customers they’ll never really possess. They’re building sophisticated bridges to other companies’ checkout pages. That’s not a business model. That’s a feature desperately seeking someone else’s product.

The math hasn’t changed since 2010. Customer acquisition: $30-100. Lifetime value through affiliate commissions: $15-25. You don’t need GPT-4 to see why this doesn’t work. You need third-grade division.

But the AI wrapper makes everyone forget fundamentals. I warned about this exact dynamic in July. Investors see a chatbot decode “something chic for a gallery opening” and lose their minds. So dazzled by technology solving the wrong problem, they ignore the right problem screaming at them.

Distribution is the right problem. And distribution has only gotten worse.

When we built ShopStyle, Google gave us millions of free visitors. Today? Google keeps over half its searches on-platform. Meta throttles organic reach to zero. TikTok is pay-to-play. The platforms these startups depend on for traffic are the same platforms building competing shopping features. You’re not competing with other startups. You’re competing with the toll road operator.

And it’s about to get worse. Google has systematically deemphasized “thin aggregators” for years, sites that aggregate without adding value. Now imagine AEO (AI Engine Optimization). The AI engines won’t even surface aggregators. Why would ChatGPT recommend your shopping middleman when it can send users directly to Everlane? You’re not just losing SEO traffic. You’re about to be completely invisible.

The conversational commerce founders think AI changes this equation. An AI that perfectly understands style preferences still pays Meta $50 to acquire each customer. A chatbot that decodes outfit context still only earns 12% when that customer buys. The intelligence of the algorithm doesn’t alter the stupidity of the business model.

Watch what happens next. These platforms will burn funding trying to growth-hack distribution. Social features. Loyalty programs. Exclusive drops. When that fails, the subscription pivot: $10 monthly for “unlimited AI styling.” But shopping isn’t streaming. You subscribe to access content, not to spend more money. The moment you charge for discovery, you’re competing with Amazon Prime, which includes streaming, shipping, and everything else for the same price as your chatbot.

Even the affiliate model itself is collapsing. Browser extensions hijack attribution. Cross-device shopping breaks tracking. Cookie deprecation eliminates visibility. A customer discovers a dress through your AI, clicks through, browses, leaves, returns three days later through Google, and Honey’s extension fires at checkout. You get nothing. That’s not a business. That’s charity for retailers.

“But Honey works!” Yes, Honey works because it’s parasitic. It sits at the bottom of the funnel, stealing attribution from everyone who did the actual work. PayPal bought it for $4 billion not because it creates value, but because it captures value others created. These aren’t business models to emulate, they’re the parasites killing your business model.

There’s one exception: luxury resale. The RealReals work because they own inventory, control pricing, capture 30-40% commissions. Notice what makes them work, they’re marketplaces with operational complexity and actual defensibility. The moment you mention this, every AI commerce founder pivots their pitch to include “luxury.” But adding the word luxury to your pitch deck doesn’t add margin to your business model.

Last week, a founder pitched me their “revolutionary” AI shopping assistant. Halfway through, I asked one question: “What’s your true attribution rate on completed purchases?”

The room went quiet.

The founder started explaining their “proprietary attribution model.” Translation: they don’t know. They never know. Because knowing would mean admitting that 60-70% of the sales they “drive” never pay them a cent.

Sometimes I wonder if these founders know exactly what they’re doing. Raise $50 million. Hire the best ML engineers from Meta and Google. Build impressive demos. Burn through cash in 18 months. Get acqui-hired before the Series B truth moment arrives. The VCs mark up the investment. Founders get signing bonuses and director titles at Google. Engineers vest at a real company. Everyone wins except the LP who thought they were funding the future of commerce.

Maybe that’s not a bug. Maybe that’s the whole business model.

At Sugar Capital, we’re always open to being wrong in interesting ways. If you’re building something that genuinely solves these structural problems, not just adding AI polish to the same broken foundation, I want to hear about it.

But if you’re pitching another conversational platform that aggregates other people’s inventory to collect affiliate fees while renting distribution from monopolists, save us both the time.

The winners won’t be the companies with the best algorithms. They’ll be the ones who understand that AI is a capability, not a business model. They’ll own something real: inventory, transaction processing, fulfillment, or distribution. They’ll use AI to enhance a structurally sound business, not paper over a structurally broken one.

Build for the constraints, not the demo. Own something real, not just the conversation.

Because in the end, the market doesn’t care how smart your chatbot is if your business model is dumb.

That’s not innovation. That’s just the most expensive job interview in history.

1. The Questions That Kill Every AI Commerce Pitch (In no particular order)

“How exactly are you acquiring customers at scale?” Show me the channel, the cost, and explain why this won’t get priced out of reach.

“What’s your true attribution rate?” The real percentage of sales you facilitate versus sales you get paid for. The delta is always 60-70%.

“How do you plan to circumvent Amazon or Shopify’s checkout?” Wait, you can’t. Shopify’s Section 1.8 explicitly prohibits processing orders outside their checkout. Amazon has similar restrictions. So your entire model depends on violating the Terms of Service of the platforms you’re trying to monetize. What’s your actual plan here?

“What’s your defense against Honey and other cookie stealers?” They’re sitting at checkout, stealing every last-click attribution. You do all the work of discovery, they fire at checkout and take the commission. How do you fight a parasite that PayPal valued at $4 billion?

“What happens when a customer has a problem?” Three items from three retailers. One damaged, one wrong size, one lost. Who handles it?

“When Google adds this feature next quarter, why do you exist?” They have the users, the data, and can run at a loss forever.

“Why won’t anyone pay $10/month for your magical experience?” If customers won’t pay directly, you’re not a product. You’re someone else’s marketing expense.

“How do you survive Google’s war on thin aggregators?” Every SEO update deemphasizes sites that aggregate without adding value. That’s literally your business model. And when AEO (AI Engine Optimization) becomes the game, why would any AI recommend your middleman platform over going direct to source? You’re optimizing for a distribution channel that’s actively trying to kill you.