The Front Door War: Who Controls the Future of Shopping?

82% of Retail Is Still Up for Grabs. That's Why They're Fighting.

Here's a number that should make every merchant sit up: only 18% of US retail purchases happen online. Eighteen percent. After three decades of ecommerce, after a global pandemic that was supposed to permanently shift buying behaviour, after billions poured into digital infrastructure — we're still barely scratching the surface.

I've spent 26 years in ecommerce, and I can tell you: every major shift in this industry has really been a fight over the front door. Who controls where the customer starts their journey? In the 2000s it was Google. In the 2010s it was Amazon's search bar. In the 2020s it was social media feeds. Each time, the company that controlled the entry point extracted the most value from the entire chain beneath it.

Right now, we're watching the next front door war unfold in real time. Except this time, three of the most powerful companies in technology — Shopify, Amazon, and OpenAI — are approaching it with completely different strategies. One is building bridges. One is building walls. And one just quietly admitted it moved too fast.

If you're a merchant and you're not paying attention to this, you're going to wake up one morning and find that someone else controls how customers find you. Again.

Shopify's Bet: Make the Agent the New Storefront

Shopify's president Harley Finkelstein stood on stage at the Upfront Summit in Los Angeles on 16 March and laid out a vision that, if you were listening carefully, was one of the most significant strategic declarations any ecommerce platform has made in years.

"We're going to begin to use these agentic applications as these kinds of personal shoppers," Finkelstein said. Simple enough on the surface. But the implications are enormous.

Finkelstein's argument is that agentic shopping is "fundamentally merit-based." He used the example of searching for trainers. On a search engine, you type "sneakers" and you see Foot Locker — because Foot Locker has the SEO budget and the ad spend to sit at the top. But an AI agent that knows you prefer On running shoes? It's going to surface On. Not because On paid for placement, but because the agent knows you.

That distinction matters more than most people realise. It cuts to the heart of how online retail has worked for the past twenty years. For two decades, discovery in ecommerce has been pay-to-play. Google Shopping ads, Amazon Sponsored Products, Meta's dynamic product ads — the entire machinery of online retail is built on the premise that brands pay intermediaries to put their products in front of customers. Agentic shopping, at least in theory, upends that model entirely.

"We're probably more excited about this particular new era of commerce than we have ever been," Finkelstein told the audience, "because we think it's just going to create so much opportunity, not just for the large merchants, but for the long tail of merchants."

That last bit is the strategic tell. Shopify's entire business is the long tail — the millions of small and mid-size merchants who can't compete with Amazon's advertising machine. If agentic shopping really is merit-based, if AI agents genuinely surface products based on quality and relevance rather than ad budgets, then Shopify's merchants suddenly have a fighting chance they haven't had in years.

Shopify isn't just talking about this. They've been building it. As PYMNTS reported, the company has developed Sidekick, an AI assistant for merchants, and is positioning agentic shopping as "another spoke on the retail machine." They announced Agentic Storefronts back in December — a system that automatically makes eligible merchant products available to AI platforms through Shopify's existing data infrastructure. No dedicated ChatGPT app required. Products get syndicated through the Shopify Catalogue so AI assistants can surface them in response to user queries.

According to Modern Retail, Shopify emailed merchants in March telling them their products would soon be "discoverable and purchasable inside ChatGPT" through Agentic Storefronts. The feature launches by default — merchants don't need to opt in. That's a bold move. Shopify is essentially telling its entire merchant base: we're putting you in front of AI agents whether you ask us to or not, because this is where commerce is going.

It's the right call. Most merchants won't proactively adopt new channels until they're forced to. Shopify making this the default is the kind of platform-level decision that separates companies with vision from companies that run surveys.

OpenAI's Quiet Retreat: When the Future Isn't Ready

But here's where it gets interesting. While Shopify was telling merchants they'd be selling inside ChatGPT, OpenAI was quietly pulling back on the very mechanism that would make that work.

OpenAI's Instant Checkout — the feature that let users complete purchases without ever leaving the ChatGPT interface — has been effectively shelved. Purchases now redirect to the merchant's own website, either through an in-app browser on mobile or a separate tab on desktop. The seamless, in-chat buying experience? Gone. At least for now.

An OpenAI spokesperson gave Modern Retail a carefully worded statement: "We're evolving how we approach commerce in ChatGPT to better meet merchants and users where they are. Instant Checkout is moving to Apps, where purchases can happen more seamlessly."

Translation: consumers weren't buying things inside ChatGPT, and the whole experiment was struggling.

Ad Age reported the blunt version: consumers "remained largely hesitant to purchase anything in an AI platform." The Information reported that OpenAI was scaling back its shopping plans more broadly.

I'm not remotely surprised. I've been building ecommerce systems since before most AI researchers had email addresses, and the one constant truth in this industry is that consumers do not change checkout behaviour easily. Trust at the point of payment is the hardest thing to earn in digital commerce. It takes years of familiarity, millions in security branding, and flawless execution. You don't just launch a new checkout surface and expect people to hand over their card numbers. Getting someone to enter their payment details in a new context — any new context — is brutally hard. Apple Pay took years to reach meaningful adoption. Google Wallet has been around for over a decade and still barely registers. The idea that consumers would casually hand their credit card details to a chatbot was always optimistic, bordering on delusional.

But here's what's important: OpenAI retreating on Instant Checkout doesn't mean the agentic commerce thesis is wrong. It means the implementation was premature. The discovery part — asking an AI agent to find you a product, compare options, check reviews — that works. People are already doing it. The conversion part, the actual moment of payment, still needs to happen somewhere the customer trusts. And right now, that somewhere is the merchant's own checkout page.

This is actually good news for Shopify. Under the updated approach, Shopify merchants' products still appear inside ChatGPT conversations, but the checkout happens on the merchant's site, through Shopify's infrastructure. Shopify gets the distribution benefit of agentic discovery without ceding checkout control to OpenAI. It's arguably a better outcome for them than the original Instant Checkout model would have been.

Amazon's Nuclear Option: If You Can't Own the Agent, Block It

While Shopify is building bridges and OpenAI is recalibrating, Amazon has chosen a third path: litigation.

In what might be the most consequential legal case in agentic commerce so far, Amazon won a preliminary injunction blocking Perplexity's Comet AI shopping agent from accessing password-protected parts of Amazon's site. US District Judge Maxine Chesney found "strong evidence" that Perplexity's tool was accessing Amazon systems unlawfully.

The ruling turned on a distinction that could reshape the entire landscape: user permission is not the same as retailer authorisation. Comet may have had the user's permission to shop on their behalf, but it didn't have Amazon's permission to access Amazon's systems. The injunction bars Perplexity from reaching password-protected Amazon accounts, including Prime accounts, and requires the company to destroy Amazon data it had previously collected.

The Wall Street Journal framed the decision as an early signal that retailers can keep outside AI agents from standing between them and their customers. And that framing is exactly right — because this isn't really about Perplexity. It's about Amazon protecting the single most profitable part of its business: advertising.

Think about what an AI shopping agent actually does. It bypasses search results. It skips sponsored listings. It ignores the carefully orchestrated product placement that Amazon charges brands billions for. Every purchase an AI agent makes on behalf of a consumer is a purchase that didn't go through Amazon's ad-funded discovery pipeline. For a company that generated over $50 billion in ad revenue last year, that's an existential threat. Every AI agent that bypasses the search bar is money Amazon doesn't make. Every transaction completed outside Amazon's interface is a data point Amazon doesn't capture. This isn't about user experience or consumer choice. It's about revenue protection, pure and simple.

Amazon's position is straightforward: our platform, our rules. If you want to access Amazon's catalogue, you do it on Amazon's terms, through Amazon's interface, seeing Amazon's ads. The idea that a third-party AI agent could waltz in, scrape product data, bypass the entire monetisation layer, and complete purchases without Amazon's consent? Not a chance.

Perplexity has argued the suit is about control, saying users should be free to choose their own AI tools. They've appealed, and as Ad Age reported, a judge granted a temporary reprieve — Amazon can block the agent, but enforcement is paused pending the appeal.

But make no mistake about what Amazon is really saying here: the front door is ours, and we'll see you in court before we let anyone else control it.

The Three Strategies, Decoded

Step back and look at what's happening across all three companies, and you see three fundamentally different theories about how agentic commerce will work:

Shopify: The Platform Play. Make your infrastructure the plumbing that connects merchants to every AI agent, regardless of who built it. Win by being the invisible layer that powers transactions wherever they happen. Don't fight for control of the front door — be the checkout counter that every front door leads to.

Amazon: The Fortress Play. Keep the front door locked. Force customers to come through Amazon's interface, see Amazon's ads, and shop on Amazon's terms. Use legal muscle to block any AI agent that tries to disintermediate the relationship. Win by maintaining absolute control over the customer experience.

OpenAI: The Aggregator Play. Try to become the front door itself — the place where consumers start every shopping journey. But when that proves too ambitious too fast, retreat to being the discovery layer and let merchants handle the actual selling. Win by controlling intent, even if you don't control the transaction.

Each strategy has real weaknesses. Shopify's approach depends on AI agents actually being merit-based rather than developing their own pay-for-placement models (which they inevitably will). Amazon's fortress strategy works until regulators decide that blocking AI agents is anti-competitive. And OpenAI's aggregator play requires consumers to actually trust an AI platform enough to start their shopping there — which, given the Instant Checkout retreat, is far from guaranteed.

What This Means for Merchants (And Why Most Are Ignoring It)

Here's what frustrates me. I talk to ecommerce merchants every week. Brands doing seven, eight, nine figures in revenue. And most of them have no idea this war is even happening.

They're optimising their Google Shopping feeds. They're tweaking their Amazon PPC bids. They're running the same playbook they've been running for five years. And the entire infrastructure of how customers discover and buy products is being rewritten underneath them.

Mladen Vladic, head of product for payment networks at FIS, told PYMNTS that agentic commerce represents "a transformational inflection point in the industry. Not only in this country, but globally." He's not wrong. But an inflection point only matters if you're positioned on the right side of it.

So what should merchants actually do? A few things:

First, get your product data in order. Agentic shopping is built on structured product data. If your catalogue is a mess — inconsistent descriptions, missing attributes, no proper taxonomy — AI agents won't surface your products. Full stop. The merchants who win in the agentic era will be the ones whose product data is clean, rich, and machine-readable. This isn't glamorous work. It never has been. But it's about to become the single most important thing you do.

Second, diversify your entry points. If you're 100% dependent on Amazon for discovery, or 100% dependent on Google, you are betting your business on one company's willingness to let AI agents access their platform. As we've just seen with Amazon vs Perplexity, that's not a safe bet. Shopify's Agentic Storefronts, whatever you think of the execution, represent the right idea: make your products available across as many AI surfaces as possible.

Third, protect your checkout. OpenAI's retreat on Instant Checkout is a lesson. Consumers trust your checkout page more than they trust a chatbot. Keep your checkout experience excellent, because for the foreseeable future, that's where the actual buying will happen — even if discovery starts somewhere else entirely.

Fourth, watch the legal landscape like your business depends on it. Because it does. The Amazon vs Perplexity case will set precedent for whether retailers can block AI agents. If Amazon wins on appeal, every retailer gets the legal toolkit to wall off their platforms. If Perplexity wins, the floodgates open. Either outcome changes your strategy.

The Walmart and Target Wildcard

There's a fourth approach emerging that deserves mention. While Shopify builds bridges, Amazon builds walls, and OpenAI retreats — Walmart and Target are quietly experimenting.

Walmart's partnership with Google's Gemini app, as PYMNTS noted, "signalled more than experimentation" — it "suggested that the world's largest retailer is coming to view agent-mediated checkout as something inevitable." Target's collaboration with OpenAI follows a similar logic.

These aren't ideological positions. They're pragmatic ones. Walmart and Target are hedging — working with AI platforms on controlled terms rather than trying to block them entirely or hand over the keys. They're saying: we don't know exactly how this plays out, but we'd rather be inside the tent experimenting than outside it reacting. That's the right instinct. The worst position in any platform shift is the one where you're caught flat-footed because you were too busy defending the old model to understand the new one.

The Real Stakes: Who Owns the Customer Relationship?

Strip away the technology and the legal manoeuvring, and this war is about something fundamental: who owns the customer relationship?

In the pre-internet era, the merchant owned it. You walked into a shop, you talked to the owner, you bought something. The relationship was direct. Then aggregators arrived — first eBay, then Amazon, then Google Shopping, then Instagram — and each one inserted themselves between the merchant and the customer. Each extraction layer took margin and data. Each one made the merchant a little more dependent and a little less informed about who their customers actually were.

Agentic commerce could go either way. If agents are truly merit-based — surfacing the best products for each individual consumer regardless of who pays for placement — then merchants get a shot at rebuilding direct relationships. A customer's AI agent might learn that they love your brand, prefer your sizing, trust your returns policy. That agent becomes an advocate for your business in a way that a Google search result never could.

But if agents become the new intermediaries — if OpenAI starts charging for placement, if Perplexity's business model depends on affiliate cuts, if the "merit-based" promise gives way to the same ad-funded model we already have — then we've just added another extraction layer. Another middleman between you and your customer. Another entity taking a slice of every transaction.

I've been through enough platform shifts to know which way this usually goes. The idealistic phase — where the new technology promises democratisation and meritocracy — always gives way to monetisation. Google's organic search results were genuinely meritocratic once, before ads consumed the top of every page. Amazon's marketplace was a level playing field once, before Sponsored Products became mandatory for visibility. Social media was organic reach once, before the algorithm demanded you pay to be seen.

Agentic commerce will follow the same arc. The early days will feel democratic and exciting. Small brands will celebrate being surfaced by AI agents on merit alone. And then, slowly, the monetisation machinery will crank up. Preferred placement. Affiliate fees. Commission structures. Sponsored recommendations. The question isn't whether this happens — it's how fast, and who captures the economics when it does.

The Clock Is Ticking

Three years ago, the idea of an AI agent buying your groceries or choosing your running shoes felt like science fiction. Today, Shopify is building the infrastructure for it. Amazon is suing to stop it. OpenAI tried to own it and had to back off. A federal judge is setting legal precedent around it. Walmart and Target are quietly testing it.

This isn't theoretical any more. The front door war is happening now, and the winners will be decided by a combination of technology, legal rulings, consumer behaviour, and — most critically — which merchants actually prepared for it.

Finkelstein's 18% number keeps rattling around in my head. Only 18% of US retail is online. That means 82% of retail transactions are still happening in physical stores, untouched by any of this. The prize isn't just reshuffling existing online sales between platforms. It's capturing the other 82%. And whoever controls the front door when that happens — the AI agent, the search engine, the marketplace, the merchant's own site — will be the most valuable company in commerce.

That's what this war is really about. Not today's online sales. Tomorrow's everything sales.

If you're a merchant reading this, stop optimising last year's playbook. The rules are being rewritten. The front door is moving. And the companies doing the rewriting aren't waiting for your permission.

The question isn't whether AI agents will change how people shop. They already are. The question is whether you'll have any say in how your products show up when they do.