Walmart’s ChatGPT Checkout Problem: Why AI Shopping Still Converts Worse Than Websites

Walmart tested purchases completed directly inside ChatGPT and found something pretty blunt: in chat conversion was about one third of the conversion rate they got when ChatGPT sent shoppers out to Walmart.com.

That one datapoint is doing a lot of work.

Because it cuts through the hype and basically says, yes, conversational shopping is real. People will ask an AI to find the right thing. They will even consider buying.

But when you ask them to actually finish the purchase inside a third party interface, the math gets ugly.

The reporting that surfaced this came through Search Engine Land and it is worth reading if you want the clean summary and framing: Walmart says ChatGPT checkout converted worse than click-outs to Walmart.com. And then Wired added more color on what OpenAI and Walmart are adjusting behind the scenes: Wired’s breakdown of the Walmart OpenAI agentic shopping deal shakeup.

The follow on detail matters too. OpenAI is reportedly shifting away from “Instant Checkout” toward merchant-managed app checkout. Translation: let the AI assist, but let the merchant run the actual buying flow.

If you run growth, ecommerce, SEO, or you are building AI workflows inside your org, this is a trend signal. Not a death sentence for AI shopping. A signal that the “where” of conversion still matters. A lot.

Let’s break down what likely happened, why merchant checkout is winning, and what brands should do now if they are experimenting with ChatGPT shopping, Gemini shopping, Perplexity style shopping answers, or whatever comes next.

---

The core problem: conversion is not just intent, it is environment

In-chat checkout sounds frictionless on paper.

The user is already in a conversation. They already said “yes”. So you assume the next step is just payment.

Except ecommerce conversion is a pile of micro decisions that depend on trust cues, UX patterns, reassurance, and merchandising that has been tuned for years. When you move checkout into a generic interface, you remove a lot of that, and you replace it with… uncertainty.

A few examples that sound small, but are huge at scale:

• Where do I see delivery date and return policy without hunting for it?
• Am I getting the right size, color, compatibility, warranty?
• Is this seller legit, is this price final, what about substitutions?
• Can I bundle, add accessories, upgrade shipping, apply loyalty perks?
• If something goes wrong, who owns the problem?

When you click out to Walmart.com, shoppers land in a familiar, brand-owned environment that answers those questions quickly. In chat, many of those answers are either buried, missing, or just don’t feel real enough to bet your money on.

So conversion drops. Not because AI cannot recommend products. Because the last mile of commerce is not just “buy now”.

---

Why in-chat checkout likely underperformed (the practical reasons)

Walmart did not publish a full teardown, so we are inferring based on how ecommerce conversion usually works. But these are the common failure points brands should assume until proven otherwise.

1. Trust and risk perception spikes at payment time

People will take advice from a chatbot more easily than they will hand it their credit card.

Even if the underlying payment rail is secure, shoppers often do not know what is happening. Is this Walmart. Is this OpenAI. Is this a reseller. Where is the receipt. Can I cancel.

Your website and app have years of muscle memory behind them. Chat does not.

2. UX constraints make “comparison shopping” worse

On a product page, comparison is effortless. Tabs, filters, specs tables, reviews, images, Q&A, “customers also bought”.

In chat, comparison becomes conversational. And that is slower than you think, especially when the user is near purchase and wants to scan, not chat.

Chat is great for narrowing the set. It is not always great for the final visual confirmation.

3. Promotions, bundles, substitutions, and edge cases are harder

Real carts are messy.

Users want to apply a promo code, use a gift card, split payment, choose pickup vs delivery, swap variants, add a protection plan, confirm compatibility, remove an out of stock item.

A merchant checkout flow handles this gracefully because it is built for it. A generic in-chat checkout tends to be narrower, meaning more “wait, can I…?” moments. Those moments kill conversion.

4. Merchandising and upsell logic gets muted

Websites are not just storefronts, they are optimized selling machines.

They show:

• add-on accessories
• higher margin equivalents
• replenishment prompts
• warranties
• “often bought together”
• subscription options

In-chat checkout can do some of this, but it is fighting the interface. The user is there to finish. Anything that feels like friction feels like the chatbot is being salesy. So either upsells are reduced, or they are shown in a way that reduces trust.

5. Customer identity and loyalty benefits are less visible

If Walmart can recognize you on Walmart.com, it can surface member pricing, saved addresses, default store, reorder history, and personalized recommendations.

In a third party environment, identity is harder. Even if it is technically possible, it often feels opaque. If a user is not sure they are getting their usual perks, they hesitate.

6. Analytics, attribution, and testing are weaker

This part matters for operators.

On your own site, you can:

• run A/B tests
• inspect funnels
• attribute campaigns
• segment cohorts
• replay sessions
• run heatmaps
• fix drop-offs

In chat, instrumentation depends on what the platform exposes. Your optimization cycle slows down, and you start flying blind. Slower iteration means conversion stays low longer.

---

The shift to merchant-managed checkout is… predictable

If OpenAI is moving away from Instant Checkout toward merchant-managed app checkout, it is basically acknowledging a simple truth:

The merchant has the best conversion engine.

Merchant checkout wins because it keeps:

• established trust signals (brand, policies, support)
• full cart and payment functionality
• loyalty and identity
• full merchandising surface
• first party analytics and experimentation
• operational guardrails (fraud, inventory logic, substitutions)

And from the AI platform’s perspective, merchant checkout reduces risk too. Payments, disputes, refunds, chargebacks, compliance, all of it is cleaner when the merchant owns it.

So we are drifting toward a model that looks like:

1. AI helps discovery and decision making.
2. AI deep links the user into the merchant’s best converting flow.
3. Merchant closes the sale and owns the customer relationship.

Which is… kind of what search has done for 20 years. The twist is the interface is conversational.

---

What this means for ChatGPT shopping, and for Gemini style shopping experiences

Two implications can be true at the same time:

1. Conversational shopping will grow. People like asking questions in natural language. Especially for complex categories.
2. Owned environments still matter. Websites and apps remain the conversion endpoints for many brands, especially for higher consideration purchases.

So if you were hoping AI shopping would replace your storefront, Walmart’s test should cool that down.

But if you were hoping AI could become a top of funnel and mid funnel assistant that sends more qualified traffic, this is actually encouraging. It suggests platforms are going to optimize for click-outs and merchant checkout integrations, not trap the whole purchase inside chat at all costs.

For Google specifically, Gemini shopping integrations will probably follow the same gravity. Google wants the transaction, sure, but it also needs a healthy merchant ecosystem, and merchants need the ability to merchandise and measure. Expect a lot of “AI assisted SERP” experiences that still route into merchant checkout, especially for categories where trust and returns matter.

---

Practical lessons for brands experimenting with AI shopping

This is the part you can actually act on next week.

1. Measure AI commerce like a new channel, not like “search”

Start by treating AI referrals as their own bucket. Even if they look like referral traffic today, they behave differently.

What to track:

• Click-out rate from AI surfaces (where available)
• Landing page conversion rate for AI traffic vs search vs paid
• Assist rate (sessions that start with AI discovery but convert later via direct or email)
• AOV and attach rate for AI-referred users (are they buying simpler baskets?)
• Return rate (AI might reduce fit, or increase it, you need to know)

If you can, create dedicated landing experiences for AI traffic to reduce mismatch. Not “AI landing pages” in a gimmicky way. Just cleaner “decision pages” that answer the questions AI users tend to have.

2. Optimize the click-out landing page for continuation, not persuasion

When a user comes from a chat, they already did the “talking”. They want to confirm and complete.

So the landing page should quickly support:

• variant selection clarity
• price and delivery certainty
• return policy reassurance
• reviews and Q&A visibility
• compatibility and specs
• trust badges and support options

If your product pages are cluttered, slow, or hide key info behind modals, AI traffic will bounce. They are not browsing. They are validating.

3. Make your catalog “AI readable” without breaking SEO basics

AI shopping depends on product data quality. And most catalogs are… messy.

Concrete checklist:

• consistent titles (brand + model + key attribute)
• clean variant structure (size, color, pack size)
• normalized attributes (materials, dimensions, compatibility, age range)
• high quality images with meaningful alt text
• clear shipping and returns data
• accurate inventory signals
• rich FAQs pulled from real support questions

This is not separate from SEO. It is SEO. Product structured data, clean internal linking, indexable category pages, and consistent taxonomy all help both search engines and AI systems interpret your inventory.

If you are building content around products, you can also strengthen relevance and conversion with supporting articles and guides. Junia has a solid walkthrough mindset on this in its piece on conversion rate improvements using AI content. The point is not “write more”. It is “answer the buying questions that cause hesitation”.

4. Do not give up first-party data just to say you are “AI native”

If a sale happens inside chat, you may lose visibility into:

• user behavior leading to purchase
• email capture moments
• consent flows
• post purchase personalization
• remarketing audiences
• on-site experimentation data

Even if platforms share some data, it is rarely as rich or as timely as what you get in your own stack.

So aim for a hybrid approach:

• let AI assist discovery
• route to your domain or app for purchase
• capture email and preferences
• run post purchase retention like normal

And yes, this is also a brand protection move. You do not want your entire checkout to be subject to UI changes you cannot control.

5. Build “merchant-managed checkout” into your AI workflow strategy

If you are an AI workflow builder, start designing for the likely end state:

• AI agent gathers requirements, budget, preferences
• AI generates a shortlist with links
• user clicks into merchant checkout
• merchant site confirms cart, shipping, payment
• AI optionally supports post purchase tasks (tracking, returns, reorders)

The companies that win will be the ones that make that handoff smooth.

That means deep links that preserve state. Pre-filled carts. Store pickup defaults. Clean mobile performance. And fewer surprises when the user lands.

6. Treat trust as a product feature, not a brand slogan

If your conversion drops in AI flows, it is often trust debt showing up.

Ways to pay it down:

• clearer shipping dates and fees earlier
• transparent return policy summaries near the buy button
• authentic reviews with helpful sorting
• visible customer support options
• better product imagery and sizing guidance
• clear seller identity if you are a marketplace

And if you are using AI to generate product descriptions, do not let them turn into glossy nonsense. Shoppers can feel it.

Junia has a useful angle on making AI copy feel less robotic in AI content humanization tools and also how to add a human touch to AI generated content. For ecommerce, the “human touch” usually means specifics. Real constraints. Real answers.

7. Update your SEO plan for “AI answers” without chasing ghosts

Do not overreact and start writing content “for ChatGPT”.

But do assume that AI search and AI shopping answers will increasingly pull from:

• product detail pages with strong structured data
• authoritative buying guides
• clear category hubs
• policies pages (shipping, returns, warranties)
• brand and merchant info pages

If you are a startup or mid-market brand, investing in this foundation is still the move. Junia’s guide on AI SEO for startups makes the case in a grounded way, and it maps well to this moment. Build assets that are both rankable and extractable.

---

A quick framework: where AI helps vs where your site should win

Think about the funnel like this:

AI is strong at:

• translating vague intent into specific options
• answering “what should I buy” questions
• comparing features in plain language
• suggesting alternatives when out of stock
• bundling recommendations in a conversational way

Your website/app is strong at:

• trust reinforcement at the moment of payment
• full cart management and edge cases
• loyalty and identity based personalization
• merchandising and attach strategies
• analytics, testing, retention capture

So your job is not to pick one. It is to design the handoff.

---

How to adapt your content and UX for AI-referred shoppers (without bloating the site)

Here is the simplest approach that tends to work:

1. Identify the top 20 questions people ask before buying in your category.
2. Ensure your PDPs answer the top 5 instantly.
3. Create supporting pages for the remaining 15, linked contextually from PDPs and category pages.
4. Keep those pages genuinely useful. Not 2,000 word fluff posts.

If you need help scaling this without creating junk content, that is where a tool like Junia.ai is actually practical. You can use it to generate search-optimized buying guides, comparison pages, and conversion-oriented product copy while keeping brand voice consistent. Start from the product data and customer questions, not from generic prompts.

If you are deciding between writing in ChatGPT vs building a repeatable pipeline, Junia’s comparison page is a helpful gut check: Junia vs ChatGPT. ChatGPT is great at one-off ideation. Junia is built for teams that need consistent outputs, internal linking, SEO scoring, and publishing workflows.

And if your team struggles with getting the model to sound like your brand across categories, this little utility is surprisingly handy: ChatGPT persona instructions generator. Even if you do not use Junia for everything, having better persona instructions improves the entire content process.

---

What to do right now (a non-hyped action list)

If you run ecommerce or growth:

1. Segment AI traffic in analytics and compare conversion, AOV, and return rate.
2. Audit your PDPs for continuation: shipping, returns, variants, reviews, specs.
3. Improve catalog hygiene so your products are easier for AI systems to interpret.
4. Strengthen first-party capture during checkout and post purchase.
5. Create a few high-intent guides that answer “which one should I buy” questions and link to products.

If you build AI workflows:

1. Design for AI assist plus merchant-managed checkout as the default.
2. Make the handoff stateful: prefilled cart, saved variants, correct store.
3. Instrument everything you can, and accept that some platform data will be limited.

---

The bigger takeaway

Walmart basically ran an experiment that a lot of brands were going to run anyway, just with less volume and less clarity.

And the result is comforting in a way.

It says: conversational commerce is not replacing ecommerce fundamentals. It is exposing them. If your owned experience is strong, AI can become a powerful discovery layer that sends you better qualified shoppers. But the close still happens where trust, UX, and control live.

Which, for now, is still your site. Your app. Your checkout.

If you want to lean into that and build content that actually helps people buy, not just rank, you can use Junia.ai to produce conversion-oriented ecommerce content at scale, with SEO structure and internal linking baked in. That is the unglamorous part that tends to win.