AIshopperscan’tseeyourShopifystore.WefixthatinLiquid.

When a buyer asks ChatGPT or Perplexity to recommend a product, engines read server-rendered schema — not your JavaScript-injected reviews or a Dawn theme’s thin Product markup. We scan your storefront across 7 AI engines, then ship the Liquid and JSON-LD fixes that get your products named in the answer.

• 7AI engines tested
• 100+store signals audited
• 4xavg AI citation lift · 60 days
• 48haudit to scored report

What we bring to Shopify AI optimisation.

An audit is only useful once the fixes ship. We built the scanner, we know the theme, and we implement in Liquid — server-rendered so the engines that can’t run JavaScript still read your products.

• We run the scanner, not a reseller license

  The audit runs on GEO Intel, our own GEO SaaS at geo.atellius.com — the same 7-engine scoring engine we built and operate. You are not paying us to run someone else’s tool and read the output back to you. When a check needs a Shopify-specific rule, we add it to the scanner ourselves.

• We implement in Liquid, end to end

  An audit is only useful once the fixes ship. We open your theme — theme.liquid, product and collection templates, snippets — and implement on a duplicated theme you preview before publish. Everything we add is server-rendered JSON-LD, so the engines that can’t execute JavaScript still read your product data.

• Scored across 7 engines, weighted for commerce

  We test ChatGPT, Perplexity, Google AI Overviews, Gemini, Copilot, Claude, and Grok — because the shopper asking “which serum for oily skin” gets a different source list on each. A schema fix that wins Perplexity can do nothing on AI Overviews. You get a per-engine score out of 100, not one blended number that hides where you lose.

• Product, Offer and Review schema at full depth

  Dawn ships a minimal Product block, and many review apps inject ratings in JavaScript no AI crawler executes. We build the full graph server-side — Offer per variant with price and availability, AggregateRating pulled from your real Judge.me, Loox, Yotpo or Okendo data, MerchantReturnPolicy, shipping, GTIN, MPN and brand. That is the difference between a passing mention and a cited recommendation.

• We know Shopify theme internals

  Online Store 2.0 JSON templates, sections and blocks, metafields, robots.txt.liquid, Markets and hreflang — we work the actual storefront surface, not a generic CMS. We never touch checkout or your payment flow, and every change ships on a theme you approve first. Shopify Plus and Hydrogen / headless storefronts are included.

Scan, fix in Liquid, prove the delta.

Five stages from store scan to a 60-day re-scan on the same baseline. You preview every theme change before it publishes; checkout is never touched.

1. 01

   Store scan

   GEO Intel runs across the whole storefront — every product and collection template, robots.txt, sitemap, and rendered-vs-server HTML — scoring 100+ signals across all 7 engines. You get an overall score out of 100, per-engine readiness, and a findings inventory ranked by projected citation impact, not alphabetically.

   Days 1–3
2. 02

   Theme & Liquid remediation

   We duplicate your theme and fix the structural blockers in code: robots.txt.liquid bot access, server-rendered content where reviews or specs were JavaScript-injected, canonical and hreflang hygiene, sitemap gaps. You preview every change before it publishes; checkout is never touched.

   Days 4–9
3. 03

   Product & review schema build

   We build the full Product graph server-side — Offer per variant with price and availability, AggregateRating pulled from your review app’s real data, MerchantReturnPolicy, shipping, GTIN, MPN and brand — mapped from metafields so it stays accurate as your catalogue changes. Every block validates in Google’s Rich Results Test on the first pass.

   Days 6–11
4. 04

   PDP & FAQ content

   We rewrite product and collection copy answer-first — one concrete claim per passage, specific materials, dimensions and use-cases an engine can quote verbatim — and add FAQ blocks and metafields with FAQPage schema on the pages shoppers actually ask about. Marketing filler and unattributed superlatives get cut.

   Days 10–16
5. 05

   Re-scan & citation delta

   At 60 days we re-run the same 100+ checks and the same buyer prompts across all 7 engines and hand you the delta: which findings closed, how per-engine scores moved, and where your products now get cited that they weren’t before. Same baseline, measured before and after.

   Day 60

What your Shopify store should do.

Fifteen checks, ordered by citation impact. We show you what to fix and why it moves the needle — the exact directives for your store come from the scan.

• 01Low effort

  Allow AI crawlers in robots.txt.liquid

  If GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot and peers are disallowed, your catalog cannot be read or cited by AI answer engines at all — the single biggest on/off switch for AI visibility.

  Exact fix in your scan report

  →Unblocks indexing across ChatGPT, Claude, Perplexity and Google AI

• 02Low effort

  Keep the storefront on server-rendered HTML

  AI crawlers do not run JavaScript, so server-rendered Liquid (or SSR if headless) is what makes product facts machine-readable on first load.

  Exact fix in your scan report

  →Product data visible to non-JS AI crawlers

• 03Medium effort

  Enrich Product JSON-LD with brand, GTIN and MPN

  Dawn outputs name/price/availability but omits the identifiers AI shopping and Merchant Listings use to resolve a product to a real SKU.

  Exact fix in your scan report

  →Unique product identity and eligibility for richer AI shopping surfaces

• 04Low effort

  Emit AggregateRating via a review app

  Ratings and review counts are strong trust signals AI assistants surface, and they require valid Product plus AggregateRating markup with real reviews.

  Exact fix in your scan report

  →Star ratings and social proof cited in AI and search results

• 05Low effort

  Bind live price and availability into Offer schema

  AI assistants and Merchant Center trust markup over the feed, so a stale InStock or wrong price yields wrong AI answers and disapproved SKUs.

  Exact fix in your scan report

  →Correct price and stock in AI answers, fewer disapprovals

• 06Medium effort

  Publish structured specs via metafields and additionalProperty

  Materials, dimensions, certifications and use-cases answer the “is this right for me” questions AI is asked, but only if they are structured, not buried in prose.

  Exact fix in your scan report

  →AI can match products to specific buyer intent

• 07Low effort

  Validate structured data after every change

  Broken or duplicate schema is silently ignored, and Dawn releases have shipped product-ID and name mismatches in JSON-LD.

  Exact fix in your scan report

  →Guarantees AI and search actually parse your markup

• 08Medium effort

  Put product facts in readable body text, not images

  AI extracts answers from on-page text; specs trapped in graphics or PDFs are invisible and cannot be cited.

  Exact fix in your scan report

  →More extractable facts available for AI citation

• 09Medium effort

  Build informational content with Shopify Blogs

  Answer engines cite content that answers questions and buying-guide queries, not just product listing pages.

  Exact fix in your scan report

  →Citations on top-of-funnel and informational AI queries

• 10Low effort

  Enrich collection pages with descriptive text

  Collections are the category pages AI leans on for “best X” queries, and an empty collection gives the model nothing to summarize.

  Exact fix in your scan report

  →Visibility on category and “best” AI queries

• 11Low effort

  Add Organization and Breadcrumb schema

  A defined brand entity and site hierarchy help AI attribute products to a trustworthy, recognized brand.

  Exact fix in your scan report

  →Stronger brand-entity recognition by AI

• 12Medium effort

  Optimize Core Web Vitals and page speed

  Fast, stable pages are crawled more completely, and lean Liquid already tends to pass Core Web Vitals more reliably than heavy SSR frameworks.

  Exact fix in your scan report

  →More complete crawls plus better UX and conversion

• 13Low effort

  Sync a Google product feed with full identifiers

  Consistent GTIN, brand and MPN across feed and on-page schema reinforces the identity AI shopping surfaces depend on.

  Exact fix in your scan report

  →Consistent product identity across Google AI shopping

• 14Low effort

  Maintain sitemap and clean internal linking

  AI and search crawlers follow the sitemap and internal links to discover the full catalog; orphaned products go uncited.

  Exact fix in your scan report

  →Entire catalog discoverable by crawlers

• 15Low effort

  Treat llms.txt as optional, not a priority

  Major AI answer-engine crawlers currently ignore llms.txt and Google has said it will not support it, so effort spent there does not drive citations today.

  Exact fix in your scan report

  →Minor future-proofing; not a current citation driver

The same query. A different shortlist.

A shopper asks the assistant which product to buy. Before, the engine reads two competitors’ schema and yours is invisible. After, your product carries a full server-rendered graph — price, rating and stock — and gets named first.

The after answer is only possible because Offer price, AggregateRating and availability are in the server-rendered JSON-LD — not injected by a review script the crawler skips.

Before and after, on the same prompts.

Each re-scan runs the same buyer queries on the same engines as the initial audit. The delta is measured, not estimated.

• Fjörd SkincareSkincare DTC
  19/100→72/100

  Named by ChatGPT in 5 of 7 test prompts for “best vitamin C serum for sensitive skin” within 60 days — zero mentions before the audit.

  Fix appliedServer-rendered Product + Offer + AggregateRating JSON-LD (pulled from Okendo review data) and FAQPage metafields on every product page.
• Batch No.9 CoffeeSpecialty Coffee DTC
  24/100→68/100

  Perplexity now cites the roaster on 6 of 9 “single-origin coffee subscription” queries; it previously cited two marketplaces in its place.

  Fix appliedrobots.txt.liquid opened to GPTBot and PerplexityBot, ItemList schema on collection pages, and roast-level plus origin metafields surfaced server-side in Liquid.
• Halden HomeHome & Furniture DTC
  15/100→64/100

  Google AI Overviews references product pages for 3 “solid oak dining table” queries, and Gemini names the brand in comparison answers where it named none before.

  Fix appliedMerchantReturnPolicy and shipping details added to Offer schema, GTIN/MPN mapped from metafields, and answer-first PDP copy replacing marketing filler.

Store names anonymised; scores and prompts are the client’s own before / after audit runs.