By Steve Merrill, Founder of WRKNG Digital — June 28, 2026
89% of Shopify stores lack the complete product data AI assistants need to recommend them. That number comes from our own audits — across dozens of stores, in dozens of categories. And in almost every case, the store owner had no idea.
ChatGPT Shopping launched broadly in late 2024. Perplexity has been surfacing products for over a year. Google AI Mode now answers "what should I buy" questions before a single organic result loads. These aren't experiments anymore. They're where your customers are starting their purchase journey.
Your store either shows up there or it doesn't. This audit tells you which.
Why Does an AI Visibility Audit Matter Right Now?
Because you can't fix what you can't see. Most Shopify merchants are still measuring success in Google rankings and Meta ROAS. Those metrics don't tell you anything about AI shopping visibility.
I've run this exact audit on 40+ stores. The pattern is almost always identical: solid Google presence, decent SEO, and then completely invisible to every AI assistant that matters. The store was optimized for a search engine their customers are using less each quarter.
According to a 2025 BrightEdge report, AI-generated answers now appear in over 84% of product-related searches on Google. That number keeps climbing. If your products don't have the structured signals AI models need, you're invisible in the channel that's growing fastest.
The audit takes about 90 minutes. Here's exactly how to run it.
Step 1: What Does Your Product Feed Actually Look Like?
Your product feed is the foundation. Without it, AI shopping assistants have nothing to pull from.
Log into Google Merchant Center and go to Products > All Products. Sort by status. Any "Disapproved" or "Pending" products are already invisible to AI-powered shopping results. Look at the disapproval reasons — the most common ones are missing GTIN, unclear product category, and mismatched price or availability.
For each active product, verify these fields are populated: title, description, price, availability, brand, GTIN (or MPN), Google product category, and at least one product image over 800px. These aren't optional. AI product recommendation models — including the ones powering ChatGPT Shopping and Google AI Mode — treat missing fields as disqualifying.
In Shopify, your feed syncs through the Google & YouTube channel app or a third-party feed tool. Check the sync frequency. If it's updating less than daily, price and availability mismatches can get your products filtered out automatically. That alone kills more AI visibility than anything else I've seen.
Step 2: Is Your Structured Data Actually Working?
Structured data tells AI crawlers exactly what your page is about. Most Shopify themes generate some schema by default. Most of it is incomplete.
Go to Google's Rich Results Test and paste in a product URL. Look for Product schema. It should include: name, description, image, brand, offers (with price, priceCurrency, availability, and URL), and ideally aggregateRating.
Missing aggregateRating is a common gap. AI assistants weight review signals heavily when deciding which products to surface. If your schema doesn't expose your review data, you're competing without one of your best signals.
Check your homepage for Organization schema. It should include your brand name, URL, logo, and social profiles. This is how AI models build a coherent picture of who you are — not just what you sell.
Then look at your category pages and FAQ content. If you have pages answering questions like "what's the difference between X and Y" or "how do I choose the right Z," add FAQ schema. Those Q&A pairs are exactly what AI assistants extract when someone asks a recommendation question.
Step 3: Can ChatGPT and Perplexity Actually Find You?
Open ChatGPT and Perplexity. Type the prompts your customers actually use. "Best [product type] for [use case]." "What's a good [product] under $X." "Where should I buy [product] online."
Note everything. Which brands appear? What language does the AI use to describe them? Are there citations, and where do they point? Does your brand appear at all?
Do this for five to ten prompts. Write down the results. This is your baseline. You'll use it to measure progress after you fix the technical issues, and to guide the content you need to create.
If competitors are showing up and you're not, read their pages carefully. Look at how they structure their product descriptions, what questions their content answers, and whether they have review content that's clearly extractable. You're reverse-engineering what the AI already rewards.
Step 4: Do You Have an llms.txt File — and Are AI Crawlers Blocked?
Type your domain followed by /llms.txt in a browser. If you get a 404, you don't have one. That's a problem.
An llms.txt file is a plain-text document at your domain root that tells AI models what your site is, what content matters, and how to make sense of your pages. The llmstxt.org standard is gaining adoption fast. It's the AI equivalent of a well-structured robots.txt — a direct signal to AI crawlers about what to prioritize.
While you're at it, check your actual robots.txt. Search for GPTBot, ClaudeBot, PerplexityBot, anthropic-ai, and cohere-ai. If any of them appear under a Disallow rule, you're actively blocking the crawlers that feed AI recommendation engines. Remove those blocks unless you have a specific legal reason for them.
Also confirm your sitemap is current and submitted to Google Search Console. AI crawlers follow the same discovery paths as search crawlers. An outdated or missing sitemap means new products and pages take much longer to get picked up.
Step 5: Can AI Assistants Extract Clean Answers from Your Pages?
Paste a product URL or category page URL directly into ChatGPT and ask: "Summarize this page and tell me what products are featured and who they're best for."
If the response is vague, refuses to engage, or gives you a generic non-answer — that's the AI equivalent of your page being unreadable. The content isn't structured for extraction.
The fix is usually one of three things. First: heading hierarchy. If your H2s and H3s don't clearly signal what each section is about, AI models don't know how to parse the page. Second: no direct answers. AI assistants extract answers best when the content includes a direct statement close to each question. "This product is best for X" beats a vague paragraph that implies it. Third: buried specs. Key product details hidden in tabs, modals, or JavaScript-rendered content often aren't seen by crawlers at all. Put the data in the HTML.
What Do You Do With the Audit Results?
Prioritize ruthlessly. The feed issues come first — they're binary. A disapproved product is invisible. Fix it and it's eligible. Everything else is a gradient.
After the feed, go to structured data. Product schema gaps are the highest-leverage technical fix. Then llms.txt and robots.txt — these take under an hour and have disproportionate impact on crawlability.
Content extractability is last but not least. It's the slowest to fix because it means rewriting pages. But it's also where most of the long-term AI visibility gains compound. The stores I've seen make the biggest gains in AI shopping exposure aren't the ones with the best SEO. They're the ones where a real human question gets a direct, specific answer on the page.
Run the audit again in 30 days. The ChatGPT and Perplexity tests are your clearest feedback signal. If you're showing up where you weren't before, the fixes worked.
Frequently Asked Questions
How long does this audit take?
About 90 minutes for your first pass. The feed check and structured data tests take 30-45 minutes combined. The live AI tests take another 20-30 minutes if you're thorough. Budget extra time if you find feed disapprovals — diagnosing the root cause takes longer than the check itself.
Do I need to run this audit if I already rank well on Google?
Yes. Google rankings and AI visibility are increasingly separate signals. A store can rank on page one for its category and still be completely absent from ChatGPT Shopping results. The data structures and content signals that AI recommendation engines reward are not identical to what drives traditional organic rankings.
What is an llms.txt file and do I actually need one?
It's a plain-text file at your domain root that tells AI models what your site contains and what they should prioritize. Think of it as a structured introduction to your brand for AI crawlers. You don't need one to get indexed, but stores that have them tend to get more coherent, accurate AI representations. It takes less than an hour to create and the upside is real.
How often should I run this audit?
Quarterly at minimum. The AI shopping landscape is moving fast enough that what worked six months ago may not be sufficient today. Run it anytime you add a new product category, update your feed setup, or notice a change in your organic or referral traffic patterns.
What if my competitors are showing up in ChatGPT but I'm not?
Start with the feed and structured data steps. In most cases, the gap comes down to product data completeness and schema accuracy — not content quality. Once your technical foundation matches or beats your competitors', the content work makes the difference. Read what those competitors are actually saying about their products. AI assistants cite the most specific, extractable content. Match that bar and exceed it.
Ready to See Where You Stand?
Running the audit yourself tells you where the gaps are. Fixing them tells you whether the changes move the needle. But you don't have to start from scratch.
We've built tools specifically for Shopify stores to run this analysis at scale — across your entire product catalog, not just a page at a time. If you want to see exactly where your store stands against the AI visibility benchmarks we've established across dozens of audits, start here.