GEO, short for generative engine optimization (you will also see it called GEO SEO, AI SEO or LLM SEO), is the work of making your pages easy for ChatGPT, Perplexity, Copilot and Google’s AI answers to find, understand and cite. For a Shopify store that means two things at once: pages that answer questions, and product data a machine can read without guessing.
We are noprob agency and we sell the Shopify Launch Sprint, so we have a stake in this topic. We also have something most GEO articles lack: our own site, which you can inspect. This piece starts there, then moves to what changes on a Shopify store. GEO is young, measurement is imperfect, and wherever we are guessing we say so.
GEO vs SEO: what stays the same and what changes?
Most of the work is the same. Google says its AI Overviews and AI Mode are built on its core ranking and quality systems, and its guide to optimizing for generative AI features asks for what it always asked: indexable pages, original content with a point of view, clear structure, good page experience. If a page is not indexed, no AI answer built on Google’s index will quote it. That is why we never sell GEO as a replacement for SEO, and why a migration needs its SEO checklist before any AI work.
What stays the same
- Pages crawlable, indexed and fast.
- Original, first-hand content, not a rewrite of what the top results already say.
- Clear headings and semantic HTML.
- Structured data that matches what the page shows.
What changes
- The unit is the passage, not the page. An AI answer lifts a sentence or a short block, so every section has to make sense on its own.
- The goal is a mention, not only a click. A store can be named in an answer without a visit.
- Other sites matter more. AI answers tend to lean on what third parties say about you: reviews, listings, press, forums.
- Product data becomes a channel. Catalog data can reach AI shopping surfaces without anyone opening your product page.
- Measurement gets worse. Rankings are a number; citations are a sample of answers that change between runs.
Our view: GEO is good SEO plus better-structured answers and better product data. Anyone selling it as a separate magic discipline is selling you something. Google’s own guide tells site owners to be wary of third-party tools that promise ranking success.
What did we do on noprob.agency, and what can we show?
On our own site we did four things you can inspect right now: an llms.txt file, robots rules that name AI crawlers, JSON-LD on every page type, and sections that open with their own answer.
The llms.txt file
Our llms.txt is a plain text file at the root of the domain, following the llms.txt proposal. It is written to be quoted, in this order:
- A one-paragraph definition of who we are and what we do.
- An About block: legal entity, founding year, founder, main platform, services, languages.
- Services with the public price of each, and a link to its page.
- Key differentiators.
- Proof points, with the date of the last check and the rule that we publish percentages, never clients’ absolute revenue.
- Who the ideal client is, and who is not a fit.
- Contact details.
- Links to our guides, case studies (each one lists its source platform) and blog articles.
Does it work? Google’s guide says plainly that Google Search does not use llms.txt and that creating one neither helps nor harms visibility there. We keep it anyway: it takes an hour, it is the one place where we state prices and proof in a form we control, and other AI tools and agents may read it. We have no evidence that it moved a single citation, and we will not claim it did.
Robots rules for AI crawlers
Our `robots.txt` allows everything except `/api/` and `/admin/`, and then names six AI user agents explicitly: GPTBot, ChatGPT-User, ClaudeBot, anthropic-ai, PerplexityBot and Google-Extended. Honestly, the wildcard rule already allows them all. The explicit lines document intent and survive the day someone adds a blanket block. One detail from OpenAI’s crawler documentation: the bot that decides whether you appear in ChatGPT search is OAI-SearchBot, separate from GPTBot, which relates to training. Ours is covered by the wildcard; if you write explicit lines, include it.
JSON-LD on every page type
We emit Organization data site-wide, Service and Offer with the public prices on the service pages, BlogPosting and BreadcrumbList on articles, and FAQPage wherever a page has real questions: the Sprint landing, the migration guide and every article with a FAQ, including this one. We do not do it to win a rich result in Google. We do it because it states question and answer unambiguously, and because it forces us to write short answers that stand alone.
Sections that open with the answer
Every H2 on this page is a question, and the first sentence under it answers it before any context. That is the cheapest GEO change there is, and it costs nothing but editing.
How do you do GEO on a Shopify store?
On Shopify, GEO comes down to four layers: product data, collection and content pages, the technical files, and the AI sales channels. Take them in that order.
Product pages and product data
Shoppers ask AI for products by attribute, so the product page is where a store wins or loses. What to do:
- Write the facts as text: material, dimensions, fit, ingredients, who it is for, how to use it, what is in the box, care, shipping and returns. Not only in images or in accordions that load late.
- Use Shopify’s structure: product type, variants, standard attributes and metafields. Shopify’s own Agentic Storefronts announcement says merchants define their product schema with standard attributes and metafields, which is what lets agents describe items accurately.
- Make the markup match the page. Shopify’s Liquid structured_data filter turns a product into schema.org Product markup (ProductGroup when it has variants), so check that your theme or app actually outputs it. Then check that price, availability and brand equal what the shopper sees; Google’s Product structured data page lists the properties.
- Keep real reviews in the page itself, on the product, not in a pop-up.
On Nazzi’s store, a supplement brand from a compounding laboratory, shoppers ask about ingredients and how to take a product, so those are the facts that must sit on the page in plain text. We also published an agentic discovery sitemap on that store.
Collections, guides and policy pages
Collection pages and guides win the broad questions; policy pages win the trust questions. Give each collection an intro that says what is inside and how to choose. Write buying guides with an opinion, not a summary of competitors. Publish shipping, returns and sizing as real, crawlable pages with plain answers, because AI assistants get asked about them constantly. Shopify’s Knowledge Base app is the place to store policies, FAQs and brand voice for agents.
llms.txt and robots.txt on Shopify
The Shopify stores we checked already serve an llms.txt, and you can edit your robots.txt. On 27 October 2026 we requested `/llms.txt` on two well-known Shopify brands, kith.com and allbirds.com. Both returned a platform-generated file for agents, pointing to Shopify’s Universal Commerce Protocol endpoints and the Shop skill. We did not find this documented, and we do not know whether merchants can change it. Open `yourstore.com/llms.txt` before you do anything, and spend your effort on content and data instead.
On robots, Shopify’s default file keeps crawlers out of admin, cart, checkout, account and sorted or filtered collection URLs, and you can customize it with the robots.txt.liquid template. The check that matters: make sure no customization, app or bot-protection layer in front of your store blocks the AI crawlers you want. Allowing GPTBot, which relates to training, is a business decision. Allowing OAI-SearchBot is what keeps you eligible for ChatGPT search answers. We allow both.
Feeds and Shopify’s AI channels
Shopify’s Agentic Storefronts publish your catalog to ChatGPT, Perplexity and Microsoft Copilot from one setup in the admin. According to Shopify’s announcement, you can switch each platform on or off, Shopify Catalog syndicates your product data (inferring categories, extracting attributes, merging variants), shoppers can buy in the chat or on your store, you stay the merchant of record, and orders are credited to the AI channel in the admin. Here the catalog is the page, so feed quality is GEO. For Google, the same logic applies to Merchant Center data, which Google’s guide names as a way for products to appear in AI responses.
The scale is rising: AI-driven traffic and orders to Shopify stores tripled year over year in the second quarter of 2026, according to Digital Commerce 360. That is a platform-wide figure, not a result of ours. The mechanics are in our guides to agentic commerce for Shopify and ChatGPT shopping for Shopify.
What does the Custom AI System do for GEO?
In the Shopify Launch Sprint, the Custom AI System builds GEO into the store instead of adding it later. Two of its six superpowers cover this: catalog and structured data ready for ChatGPT, Copilot and Google AI Mode, and an automatic GEO layer, so every new product is born with schema and data generated on its own. These are qualitative claims: we do not publish a speed or citation number for them, because we have no before-and-after measure.
A migration is the cheapest moment to fix product data, since every product is touched anyway. It is also the riskiest moment for SEO, so the redirects come first: our guide to migrating to Shopify covers that. The Sprint takes 14 days, with go-live between day 15 and 30 and 90 days of support. If your store is already on Shopify, the AI Accelerator is the lighter route: 30 days one-to-one to find the growth opportunities and switch on the first AI solutions with your team.
How do you measure GEO?
You can measure GEO only partly, and anyone showing you one GEO score is overselling. Use four imperfect sources together:
- Shopify admin: with Agentic Storefronts, orders are attributed to the AI channel.
- Analytics: look for chatgpt.com, perplexity.ai and Copilot as referral sources. Some AI visits arrive with no referrer, so you will undercount.
- Search Console: Google points to its reporting for generative AI features.
- A fixed list of prompts you run by hand every month: are we named, which page is cited, who else appears. Answers vary between runs, so look at the trend.
Measurement is why we fix tracking before comparing channels. On Cumini, the tracking audit score went from 31 to 92 out of 100 around the migration, and only then did channel comparisons mean anything.
What is still unproven about GEO?
Most GEO tactics have no public evidence behind them, including some of ours. We do not know that llms.txt changes citations. We do not know how much explicit crawler rules or schema change what a model says. Google’s guide says no special file or markup is needed for its AI features and warns against creating pages just to manipulate AI responses. What we can say: clear answers, accurate product data and open crawling cost little and help classic search too, so we do them whether or not AI cites us. We have no AI-referral number for noprob.agency that we would publish, and no promise of citations to make you.
Where should a Shopify store start this week?
Start with the checks that cost nothing and the pages that sell. In this order:
- Confirm your key pages are indexed and that no app or firewall blocks AI crawlers.
- Open `yourstore.com/llms.txt` and see what Shopify serves.
- Rewrite your five best-selling product pages with the facts as text and check the Product JSON-LD matches.
- Add an answer-first intro to your main collections and a real shipping and returns page.
- Turn on Agentic Storefronts for the platforms you want and review the catalog data they use.
- Write down 10 prompts a customer would ask and run them monthly.
The Shopify Launch Sprint moves your store to Shopify in 14 days with our Custom AI System: a fixed calendar, a public price and 4 written guarantees.
Antonio Manitta
Founder & eCommerce Manager at NoProb Agency
For over 7 years I have helped fashion, supplements, and DTC brands scale their eCommerce. I work in long-term partnership with three brands at a time, building stable systems that compound.
About NoProb →
