A GEO strategy for your Shopify store comes down to three things: make your product facts readable as plain text in the HTML, get mentioned on the third-party pages that language models already read, and check monthly whether you actually show up when someone asks ChatGPT or Perplexity for a recommendation in your category. Everything else is commentary. Generative engine optimisation is not a separate discipline from SEO so much as a shift in what gets rewarded — engines that write an answer instead of listing ten links need extractable claims, and most Shopify product pages don't have any.
What changes, and what carries over
The crawl, the index, the sitemap, the canonical tags: all still matter. If Google can't index your collection pages, no answer engine is going to cite them either. So the boring technical foundation carries over unchanged.
What changes is the unit of value. Classic SEO optimises a page to rank. GEO optimises a sentence to be quoted. When someone asks an assistant "which Indian skincare brands ship 30ml serums under ₹1,500 without added fragrance", the model is assembling an answer from fragments it can lift and trust. A page that says "Our formulations are crafted with care using premium botanicals" gives it nothing. A page with a spec block listing volume, price in rupees, full INCI list, pH and shelf life gives it four quotable facts.
The second change: you are competing against pages you don't control. Reddit threads, YouTube reviews, comparison blogs, marketplace listings. In our experience the citation set for a commercial query skews heavily toward third-party pages, and the brand's own site shows up as one source among several. That reorders where the effort should go.
Write product pages an answer engine can quote
Open your best-selling product page and try to answer, from the visible text alone, the eight questions a buyer would actually ask. Material. Dimensions. Weight. Care. Warranty. What's in the box. Who it isn't for. Delivery time to a tier-2 city.
Most Shopify stores fail four or five of those, and the usual reason is that the answers exist — they're just baked into a lifestyle image or a size chart JPEG. Image alt text is not a substitute. Convert every spec graphic into an HTML table or a definition list. It takes an afternoon per collection and it is the single highest-return GEO task we do.
Two more things worth the effort:
- A short "who this is not for" line. Models love a disqualifier because it makes the recommendation feel considered, and buyers who self-select out were never going to keep the product anyway.
- Explicit comparisons against your own SKUs. "The 40W model charges a 5,000mAh phone in about 55 minutes; the 20W takes roughly 95." That sentence gets quoted verbatim.
The Shopify-specific things that quietly block you
A few patterns show up on almost every audit.
Review apps that inject content client-side. Your 340 reviews render in the browser through JavaScript. Some crawlers render, many don't. If the review text isn't in the initial HTML, treat it as invisible for citation purposes. Check the app's settings for a server-side or SEO-friendly rendering option; several offer one and it's off by default.
Metafields that never reach the template. Stores migrate in with rich attribute data sitting in metafields and a theme that displays none of it. If you've imported ingredient lists, certifications or fabric composition, render them. We see this constantly after a platform migration where the data came across but the template didn't.
Infinite scroll on collections with no paginated fallback. Products past the first 24 exist only after a scroll event fires. Add paginated URLs, even if the visible UX stays infinite.
Tabbed content in accordions. Fine if the text is in the DOM on load. Not fine if the tab fetches on click.
Slow pages hurt here too, though less than people claim. A crawler with a short timeout on a 6-second product page will take what it has and move on. If your LCP is above three seconds you have a conversion problem before you have a GEO problem, so fix it for the revenue reason and take the crawl benefit as change.
Structured data: the parts that pull weight
Product schema with a real Offer block — price, priceCurrency INR, availability, shipping details. Organization schema with your legal entity, GSTIN if you're comfortable publishing it, a support number, a physical address. FAQPage on pages that genuinely have questions and answers written out.
Do not add aggregateRating unless the ratings are real and visible on the page. Do not mark up an FAQ that doesn't exist in the visible text. Both are trivially detectable and both cost you trust with the crawler you were trying to impress.
Shopify's default themes ship some Product markup. It is usually incomplete — missing GTIN, missing brand, missing shipping. Validate a product URL, a collection URL and the homepage. Fix what's flagged. Then stop, because there is no additional schema type that unlocks citations, whatever the plugin marketing says.
Get mentioned where the models already read
This is the uncomfortable half of any GEO strategy, because it isn't work you can do inside your Shopify admin.
Answer engines lean on sources with a lot of independent corroboration. Practically, that means: subreddit threads in your category, YouTube reviews with accurate transcripts, "best X in India" roundups on publications that actually get read, Quora answers, and your listings on marketplaces where the specs are complete. A brand with fifty mentions across those surfaces gets recommended more than a brand with a beautiful site and none.
What we've found works, in rough order of effort-to-return: seeding accurate product data to comparison sites and category roundups; getting real reviewers a unit rather than paying for a scripted post; answering questions in the communities where your buyers already argue about products; and publishing genuinely comparative content on your own blog, including comparisons where you lose on price. A page that admits a competitor is cheaper reads as credible to a model in the same way it does to a human.
Paid placement in listicles is a grey area. It works, until the publication loses trust, at which point you have spent money to be cited by a source nobody weights.
Rupees, COD, GST: the details Indian shoppers ask about
Queries from Indian buyers carry constraints that global brands don't bother stating. Is COD available? What's the return window? Is the price inclusive of GST? Does it ship to Guwahati, and how long? Is the food product FSSAI-licensed, and what's the licence number?
Put those answers in text on the page, not only in a policy PDF or a chat widget. A shipping section that says "Delivered in 2–4 days to metros, 4–7 days elsewhere. COD available up to ₹5,000. All prices include GST." is three sentences and it answers the three questions that decide the purchase.
For jewellery, electronics and anything where price moves, state whether the displayed price is live or indicative. Models will happily quote a stale figure and the customer will hold you to it.
robots.txt.liquid: allow or block?
Shopify lets you edit the robots file through a robots.txt.liquid template in your theme, so you can add or remove crawler rules yourself. The question is whether you want to.
Our default for D2C brands is allow. You want to be in the answer. Blocking GPTBot or PerplexityBot to protect product copy that is already public is a defensive move with no upside for a store whose problem is discovery.
The exception is a brand whose moat is editorial: original photography, long-form guides, proprietary research that competitors scrape and republish. There, blocking training crawlers while allowing retrieval crawlers is a reasonable position, and worth a conversation with whoever owns your IP. Note that the two are different bots with different names, so read what each one does before you write the rule. Getting it backwards blocks your citations and permits the training.
Measuring it without kidding yourself
You can't read raw server logs on Shopify, so crawler-hit analysis is off the table unless you're running a headless front end. Two methods that work:
Referral traffic. In GA4, filter session source for chatgpt.com, perplexity.ai, copilot.microsoft.com, gemini.google.com and claude.ai. Volume will be small. Quality is usually high, because the visitor arrived pre-qualified by an answer. Take a store at 40,000 organic sessions a month converting at 1.4%. If 260 sessions come from AI referrals at 3.6%, that's 9 orders — under 2% of organic volume, and worth roughly what a mid-size paid keyword returns. Small, real, growing.
Prompt panel testing. Write 30 buying prompts a real customer would type. Run them across three engines on the first Monday of each month. Log whether you appear, where, and what source was cited. 30 × 3 = 90 checks; if you show up in 7, that's 7.8% and it's your baseline. Two hours a month, and it is the only metric that maps directly to the work.
Both numbers wobble. Answers vary between sessions for the same prompt, so treat a single result as noise and the trend across three months as signal.
What we'd skip
An llms.txt file. It's cheap to add and adoption by the major engines is unconfirmed, so add it if it makes you feel better, but don't count it as work done.
Publishing forty AI-written blog posts to "feed the models". Thin content has never worked and generative engines are, if anything, harsher about it than search.
Any tool that promises to track your "AI visibility score" as a single number. The underlying data is a prompt panel like the one above, and you can run that yourself for free.
Where to start
Pick your top ten products by revenue. Get every spec out of images and into text, add a short disqualifier line, and write the shipping and COD terms into the page body. That's a week of work and it fixes the layer everything else sits on. Then run your first 30-prompt panel so you have a number to measure against.
If you'd rather have someone else find the gaps first, our free store audit covers crawlability, schema and product-page structure, and we're happy to include a prompt panel for your category. The SEO and content work is where the rest of it lives.