#BlinkTank: OpenAI drops direct checkout, and why Google looks like it's winning the shopping race
Hi everyone,
Last week, our focus was very much on the conflict in the Middle East and the closure of the Strait of Hormuz. While that situation is still ongoing and continues to disrupt global logistics, the more technical side of the industry hasn't slowed down.
While there’s been plenty of hype about OpenAI disrupting the shopping journey, it looks like they’ve spent much of their time leaning on Google’s infrastructure instead. Building the actual plumbing for commerce - the catalogues, the payments, and real-time inventory - is a massive undertaking, and this week we got a clear look at where OpenAI is hitting a wall.
OpenAI drops its plans for direct checkout
OpenAI has reportedly abandoned its ambitions for "agentic commerce" - the goal of completing a purchase entirely within ChatGPT. After brief trials with Walmart and Etsy, they’ve hit the same wall that broke Facebook Shops and Pinterest’s "Buyable Pins" (more on which shortly): normalising product data across the entire web is an immense technical challenge.
For eCommerce to work at scale, you need absolute data stability - real-time inventory, standardised attributes, and reliable pricing. Google has spent twenty years building the Merchant Center infrastructure to handle this, and that legacy is proving to be an enormous moat.
It is a reminder that while AI is built on probability, commerce requires 100% accuracy. If an AI is only 90% sure a price is correct, the transaction fails. OpenAI is now shifting to completing transactions through third-party apps instead.
📎 The Information:OpenAI drops plan for direct checkout
The product catalogue graveyard
To understand why OpenAI is pivoting, you have to look at who else has tried - and failed - to build a universal storefront:
- Meta: Tried "Facebook Shops" and "Instagram Shopping." They hit a wall trying to sync millions of disparate feeds with real-time inventory and eventually retreated to just sending traffic back to merchant sites.📎 Source: Feedonomics
- Pinterest: Launched "Buyable Pins" in 2015 but abandoned the direct "Buy" model in 2018. They couldn't maintain the data stability required for tax, shipping, and inventory across thousands of different retailers. 📎 Source: CNBC
Why Google is winning the AI discovery race
It would be easy to look at OpenAI abandoning direct checkout and conclude that AI shopping is just a bunch of hype. But the reality is more interesting: people are actually finding AI discovery incredibly useful for the search phase - they just aren’t transacting yet.
The reason for this is simple: it hasnt been easy enough to buy. While OpenAI has mastered the conversational side of discovery, they’ve hit a wall on the technical plumbing required to actually move money.
A study by Tom Wells of Peec AI highlights exactly how far behind they are; it found that 83% of the product carousels in ChatGPT are actually sourced from Google Shopping via query fan-outs. Instead of building its own index, OpenAI is effectively scraping Google’s.
All of this points to Gemini and Google winning out in the long run. They haven't just built a chatbot - they’ve spent two decades building the Merchant Center infrastructure and now the Universal Commerce Protocol (UCP) that is likely to power the entire ecosystem. Google owns the data layer that even its competitors have to rely on to look competent.
📎 Search Engine Land: ChatGPT sources 83% of products from Google Shopping
Shopify walks back metafield limit change
Last week we noted Shopify’s move to cut JSON metafield limits from 2MB down to 16KB. Following significant pushback from the developer community, they’ve reversed course.
Senior Product Lead Darius Gai confirmed in a post on X that the new limits are now set at 128KB for JSON (up from 16KB), and existing apps have been grandfathered in at the original 2MB.
It’s a good sign that Shopify is listening to its ecosystem, but the direction remains the same: the platform is pushing for leaner, more structured data.
📎 Darius Gai on X: Shopify reverses course on JSON metafield limits
The 2026 SEOFOMO eCommerce SEO survey results
The brilliant Aleyda Solís has released the latest SEOFOMO survey focused on eCommerce SEO/AI, and the results are super interesting. 69% of specialists say achieving goals is harder than it was two years ago - largely due to a "perfect storm" of SERP volatility and internal bottlenecks.
- Shopify is the default: 33% of specialists now use Shopify, cementing its status as the dominant CMS for non-enterprise brands.
- The implementation gap: The top reason projects fail isn't bad strategy, but slow development and a lack of stakeholder buy-in. This is something we talk about extensively in our article, The 8 reasons why Shopify SEO projects fail.
- AI is mainstream: 62% of SEOs have already integrated AI into their workflows, but the focus is shifting from "content volume" to building brand authority that AI models can trust.
- A pivot to PDPs: Energy is moving from category pages (PLPs) to product pages (PDPs) as discovery becomes more multi-modal and lower-funnel.
📎 #SEOFOMO: Ecommerce SEO & AI Search Optimization Survey Results
Why related collections matter
While the survey shows a pivot toward PDPs, this doesn't mean category pages are becoming irrelevant - instead it reiterates their importance. In a multi-modal search world, the PLP provides the structural "scaffolding" that helps those product pages rank.
This is where related collections (or secondary navigations) come in. Shopify’s flat URL structure makes it difficult to build a clean site taxonomy, and standard filters are often an SEO dead end because they aren't indexable. By using related collections, you create indexable "parent-child" relationships that pass authority directly to your PDPs.
We’ve been using a custom component to bridge this gap by creating indexable "parent-child" relationships between collections. Andy Russell has put together a short video walkthrough of the logic behind the component, along with a full breakdown on the blog of how it improves site architecture.
📎 Blink blog: What are related collections and why do they matter?
📎 YouTube: How we build secondary navigations on Shopify (that Google can actually crawl)
Network shout-out: OriginalVoices
We’ve recently been lucky enough to be part of a trial programme with OriginalVoices, and it’s a product we’re finding absolutely fascinating.
As we’ve said so many times, AI is a race to averageness, and the quality of the input is everything. What Original Voices does is provide a "human intelligence layer" for your workflows.
It lets you query Digital Twins of real people - owned, trained, and managed by the individuals they represent - and feed those insights directly into your prompts. We’ve found it incredibly useful for content research, ad copy generation, and refining brand messaging. It's incredibly powerful, and the product looks like it's just going to get better and better
📎 Original Voices:The human intelligence layer for AI
As ever, thanks for reading!
Sam
This newsletter was brought to you by the team at Blink, specialists in SEO and PPC for large-catalogue Shopify stores.
