#BlinkTank: What Shopify is telling AI agents about your store
Two #BlinkTanks in three days, you lucky people. Following on from Monday’s Special Update, the Agentic dashboard and the native /llms.txt files only really started appearing over the weekend, so this is all still new ground. However, we’ve been digging into it on client accounts for the last couple of days and the picture is becoming clearer - if the underlying foundations are shaky, the new surface tends to expose it pretty quickly.
Also in this issue: a Shopify article showing 71% of AI-attributed orders are coming from the long tail, ChatGPT opens self-serve ads with the $50k spend floor removed, variant-level publishing finally lands in Shopify, Google drops FAQ rich results entirely, and Klaviyo expands its MCP integration with Claude. Plus a new guide on Shopify profitability from Quickfire has just gone live, a finding from our client work on where the commercial signal lives in your GSC data, and a separate piece on why GA4 is probably underreporting your organic revenue.
On to the news!
What Shopify is telling agents about your store
We've been working through the new AI infrastructure on client accounts since Monday. The pattern that keeps emerging is fairly straightforward. The Agentic Storefronts dashboard and the /llms.txt file are not telling AI agents anything new about your store - they are exposing what your store already contains through a new lens. It's not a big new strategic frontier, it's a rendering layer on top of what's already there.
Three things worth looking at on your own store.
First, /collections/all. The llms.txt file Shopify is now serving on every store points AI agents to /collections/all as the canonical "all products" route. If you maintain that collection properly - current, complete, structured - then fine. But a lot of large-catalogue stores don't. Some have it as a default catch-all that hasn't been thought about since launch. Some have it filtered or restricted in ways that don't reflect the real catalogue. Some have it but it's missing recent products because the collection rules haven't been updated. It's worth opening your own /collections/all URL and asking whether what's there is what you want AI agents to see as your store.
Second, contact details. The llms.txt file pulls a contact email and phone number from whatever Shopify has on file for your store. On stores set up with customer-facing details, that's fine. On stores set up with internal-facing details - a client portal email, a UK domestic phone format that wasn't intended for public exposure - those details are now being served to any agent or scraper that fetches the file.
Third, the agents.md is essentially a Shopify template with three fields swapped in. The store name, contact email, and phone number change between stores. Everything else - the UCP discovery URL, the MCP endpoint, the "typical agent flow," the supported versions, the capabilities advertised - is identical. Which means every Shopify store is telling agents "here's how to discover, search, cart, checkout via UCP" regardless of whether the merchant has agentic storefronts switched on, whether their catalogue is set up to support it, or whether any of the routes actually work. There's currently no merchant control over what the file says.
The wider point is that a common interpretation is that AI commerce is a new workstream needing new tooling and strategy. Our view is that it isn't. It's the same product data hygiene, URL hygiene, and content depth work that has been good practice for years - rendered through a new platform that makes existing problems much more consequential.
🔗 Shopify Help Centre: managing your agentic storefronts
AI favours the specialist, and the long tail is now the majority
Shopify published an article on Sunday from Nell Thomas, their VP of Data Science. This stated that product categories outside the top 100 by gross merchandise volume now account for nearly 55% of all sales on Shopify. The long tail is quite substantial.
The number that should get attention is this: 71% of AI-attributed orders on Shopify in 2025 came from long-tail categories. Nell’s framing is the line worth quoting: “A search engine rewards popularity. An AI agent recommends relevance.”
That distinction explains a lot. An AI agent asked to find the best kite for flying without wind, or the right pill case for someone with hand tremors, pulls from product data, descriptions, reviews, and policies rather than popularity signals. The store with the better data on the more specific product wins.
For a large-catalogue operator, this isn’t a story about niche stores winning (although any will). It’s about the long tail inside your own catalogue.
🔗 Shopify: Entrepreneurs are creating new markets and outselling the mainstream
ChatGPT opens self-serve ads with the $50k floor removed
OpenAI launched a beta self-serve Ads Manager on 5 May at ads.openai.com, alongside CPC bidding, a Conversions API, and pixel-based measurement. The $50,000 minimum spend that gated the original pilot has been dropped, with recommended starting CPC bids of $3 to $5. US advertisers only at launch; UK, Brazil, Japan, South Korea, and Mexico are in the next phase.
For this audience, the meaningful detail is the category list. ChatGPT ads are open to consumer goods, local services, travel and entertainment, and digital products and education, which covers a lot of large-catalogue Shopify territory. Regulated verticals remain restricted while OpenAI works through policy.
The structural take connects to last week’s Stripe + Google story. The agentic ad layer is now contested across Google, Stripe, and OpenAI, with each playing a different position. None of them has won yet. The brands testing all three early are going to learn the most.
🔗 OpenAI: New ways to buy ChatGPT ads
Variant-level publishing finally lands in Shopify
Shopify shipped variant-level publishing on 7 May. Individual variants can be published or unpublished per channel or catalog from the product details page, the variant page, or the bulk editor. Defaults are unchanged, so existing apps that publish at the product level still work.
Anyone running a large-catalogue store has been working around the absence of this for years. Use cases include staging upcoming variants before launch, retiring discontinued options without deleting the variant or its order history, restricting region-specific variants to the right markets, and gating bulk-quantity variants to B2B catalogs only.
A timely example. We had a request from a client this week that fits the new functionality almost exactly: retire one specific variant from the online store without deleting it (because it had order history attached) and without writing custom theme code to hide it conditionally. Before last week, that would have meant a duplicate product or a Liquid snippet checking the variant title. As of last week, it’s a checkbox on the variant page.
🔗 Shopify changelog: product variant publishing
Google drops FAQ rich results entirely
As of 7 May, Google has stopped showing FAQ rich results in search. The Search Console rich result report and Rich Results Test lose FAQ support in June, with API support following in August.
We’re continuing to implement FAQ schema regardless. Just because it isn’t being surfaced in SERPs doesn’t mean it isn’t valuable. Other engines and AI surfaces still parse it, and things change - what isn’t being shown today may well be shown again in a different form in twelve months.
The more practical reason is that we template schema as a default. Once that templating infrastructure is built, including FAQ markup costs nothing extra to maintain and enforces good data discipline across the rest of the structured data on a store. If Google reverses, or another engine starts giving FAQ more weight, the stores that kept implementing it are already in position.
🔗 Search Engine Land: Google to no longer support FAQ rich results
Klaviyo’s MCP integration with Claude expands
Klaviyo announced an expanded integration with Anthropic on 7 May. The MCP connector is now available more broadly across Claude’s products, including Claude.ai and the new Claude Cowork desktop product, with a Query Metric Aggregates tool that exposes raw performance data alongside the existing connector capabilities. In practical terms: connect Klaviyo, ask Claude questions about campaign performance, customer segments, and flow data in natural language, and get summaries or brief drafts grounded in actual data rather than generic prompts.
The wider context sits alongside Monday’s Special Update. MCP is rapidly becoming the default integration standard for marketing tools that want to be useful inside an LLM workflow - Shopify already has one, Stripe joined the UCP Tech Council last month, Klaviyo expanded this week.
🔗 Klaviyo: new agentic workflows come to Claude
The Quickfire Guide to Profitability is live
Quickfire Digital have just published The Quickfire Guide to Profitability, and we are very happy to have contributed a section to it.
The guide brings together contributions from agency partners and merchants across the Shopify ecosystem, with a useful argument running underneath: revenue and profit are different problems, and the playbook for the first decade of DTC focused almost entirely on the first. The piece covers the shift from ROAS to POAS as a measurement framework, the most common margin killers and how to eliminate them, and how to grow through retention, bundles, and smarter incentives rather than deeper discounting.
This is exactly the conversation we have been having with most large-catalogue Shopify operators over the last twelve months. CACs have climbed, paid efficiency is harder, and growth-at-all-costs has stopped working as a strategy. If profitability has moved up your priority list, the guide is worth reading.
🔗 The Quickfire Guide to Profitability
Collection pages are where the commercial signal lives
We’ve been correlating Google Search Console non-brand clicks against Shopify net sales as our primary commercial signal across our client base. GSC is the level at which organic search is actually visible.
On a recent client review for a large-catalogue store, the aggregate correlation came back at r = 0.21 across all URLs. Statistically weak. When we filtered to collection pages only and re-ran the analysis, the correlation jumped to r = 0.84, with p = 0.0007 across 11 months. A strong, reliable commercial signal hiding inside a weak aggregate one.
The structural reason: 60% of product URLs only appear in GSC for a single month, so PDP-level click data is too noisy to track against monthly revenue. Collection pages are the level at which the data sits still long enough to be measurable.
The practical implication is reasonably direct. Collection pages aren’t one of several things that matter, they’re the level at which the commercial relationship is visible. PDPs still do their job at the bottom of the funnel. But if you’re trying to make the commercial case for organic, or work out where the marginal hour of effort goes, collection pages are where the evidence is.
🔗 Correlating GSC clicks and revenue
GA4 is probably underreporting your organic revenue
I wrote a piece for Addingwell this week on a related issue: GA4 chronically underreports organic revenue on Shopify stores. A mature large-catalogue Shopify store should be seeing 20-25% of revenue from organic search. If your GA4 dashboard shows 10-12%, the most likely explanation is that the revenue is sitting in Direct or Unassigned, not that your SEO is broken.
Rough sense-check before doing any infrastructure work: take your reported organic revenue, add 30% of your Direct/Unassigned figure, and see how close it gets to 20-25%.
If your data tells you organic is 10% of your revenue and you take that at face value, you’ll cut investment in your highest-margin channel based on a tracking problem you don’t have.
🔗 Why your organic revenue is probably half what it should be
Jobs
Hiring:
Emily Corson is hiring an Ecommerce Merchandise Manager at Charlotte Tilbury Beauty. Senior London-based hybrid role at one of the biggest DTC beauty brands going.
Another mention for Quickfire Digital, who are hiring in their dev team. A wonderful place for a Shopify developer who wants to work on serious builds.
Sean Riley is hiring a Digital Designer for the Sanrio ecommerce team, covering Sanrio.com, social channels, the Sanrio Store app, and internal marketing. Iconic characters, lots of variety, and a strong team.
Ethan Blayne is sharing a roundup of Perth, Australia ecommerce roles. If you are in Western Australia or considering a move, worth a look.
Looking for work:
Stephanie Coffey is looking for senior content or strategy roles after seven years at Shopify. Open to tech, creative industries, and purpose-driven brands, and looking for roles where she shapes the vision rather than just delivers it.
If you are hiring and the role fits someone with large-catalogue Shopify experience, or you are looking and have that background, send me a message.
Partner shout out: Swap Commerce
Returns and exchanges are something most large-catalogue stores undercook. Either it’s a manual support process that doesn’t scale, or it’s a generic returns app that creates as much accounting mess as it removes. Swap Commerce is built for the second problem. Shopify-native, with a branded self-serve returns portal, instant exchanges, store credit incentives, custom return rules, and a working capital product that lets you pull cash forward against returned inventory. Integrates with Shopify’s native Returns and Exchanges APIs, which keeps exchange data attached to the original order rather than ending up in a separate ledger nobody reconciles.
🔗 Swap Commerce
As ever, thanks for reading.
If you found this useful, I would love it if you could share it with someone you think would find it interesting - whether that is someone managing a large-catalogue Shopify store or anyone else who follows this kind of thing.
See you next week.
Sam
Who am I and what does Blink do?
I’m Sam Wright, Managing Director and co-founder of Blink SEO. We’re a search marketing agency that works exclusively with large-catalogue Shopify stores - the kind where the challenge isn’t just visibility, but helping customers find the right product in the first place.
We specialise in SEO and PPC for merchants managing hundreds or thousands of products. Most of our work starts with taxonomy: how a store is structured and categorised, and what that means for organic performance, paid efficiency, and on-site discovery.
If you found this useful and aren’t already subscribed, you can sign up at blinkseo.co.uk/blinktank.
If you run a large-catalogue Shopify store and want to talk about organic or paid performance, reply to this email or find me at blinkseo.co.uk.