By Steve Merrill, Founder of WRKNG Digital | July 2, 2026
The metrics Shopify merchants have tracked for a decade (keyword rankings, domain authority, backlink counts) were built to measure performance in a world where humans typed queries into Google. That world is shrinking. The signals that determine whether AI shopping assistants recommend your products are different. Entirely.
Here are the five signals that actually matter now.
1. AI Citation Rate
AI citation rate measures how often AI assistants (ChatGPT Shopping, Perplexity, Google AI Overviews, Copilot) mention your brand or products when answering relevant purchase queries. This is the direct analog to organic search ranking. High citation rate means AI models are pulling your brand into conversations with buying intent. Zero citation rate means you don't exist to the fastest-growing discovery channel in ecommerce.
Measuring it requires running a consistent set of target prompts across AI platforms and logging which brands appear. Brightedge's 2026 AI Search Visibility Report found brands with complete structured data receive AI citations at 3.4x the rate of stores with incomplete schema. That gap is widening. This is what keyword rankings used to tell you.
2. Structured Data Completeness Score
AI shopping assistants are built on structured data. When a model decides whether to recommend your product, it's reading machine-readable signals: Schema.org Product markup, OpenGraph tags, JSON-LD. If those fields are missing, wrong, or outdated, the model has nothing to work with. It skips your product and moves to one that gave it what it needed.
Structured data completeness score measures the percentage of required schema fields filled correctly across your catalog. Required fields include name, description, image, offers (price, availability, currency), brand, and identifier (GTIN or MPN). According to Google's documentation on product structured data, missing just the "offers" block is enough to disqualify a product from rich results entirely, and by extension from AI-surfaced recommendations that draw on those same signals.
Most stores score below 60% on a full audit. We've run hundreds. The number is that bad.
3. Product Feed Quality Score
This one surprises merchants. The product feed (the file you submit to Google Merchant Center, Meta Commerce, or Bing Shopping) has become a primary data source for AI shopping agents. These agents don't browse your store the way a human does. They ingest feed data and use it to match products to buyer queries.
Feed quality score reflects how well each product record satisfies the fields AI systems weight most heavily: GTIN (the global trade item number that confirms product identity), description specificity, image count (products with 3+ images surface more frequently), and category accuracy. Salsify's 2025 Product Experience Report found products with complete GTINs are 2x more likely to appear in AI-generated recommendations than unidentified products. GTIN is a field many Shopify merchants skip because Google Merchant Center doesn't hard-require it for basic approval. Domain authority doesn't move this needle. Feed hygiene does.
4. Brand Entity Recognition
AI models don't just read your website. They carry prior knowledge about brands built from training data and live retrieval. Brand entity recognition measures how well AI models "know" your brand independent of a specific product query: can they answer questions about what you sell, who you're for, and what makes you different.
AI shopping assistants pre-filter results based on brand familiarity. A model asked "best eco-friendly activewear brands" draws on entity knowledge first, then confirms with current data. If your brand isn't in that graph, you're not in the initial consideration set. Wil Reynolds of Seer Interactive has written extensively about brand entity building as an AI visibility prerequisite. Google's entity graph rewards consistency: consistent brand name, descriptions, and categories across your site, feeds, Google Business Profile, and third-party mentions. Backlink count was a proxy for authority. Brand entity recognition is the direct signal.
5. Agent Crawlability
This is the newest signal and the one merchants are least prepared for. AI shopping agents (OpenAI's shopping agent, Perplexity's buy mode, Shopify's Sidekick extensions) don't crawl your site the way Googlebot does. They use a combination of headless browser rendering, structured data parsing, and API-style data requests. If your product pages block these agents, render critical content only in JavaScript that the agent can't execute, or return errors on key product endpoints, the agent marks your store as unreliable and moves on.
Agent crawlability is measured by auditing which AI user-agents your server accepts, whether product pages render fully without client-side execution, and whether your robots.txt is blocking crawlers you actually want in. Google's Search Central documentation on JavaScript SEO covers the rendering gap clearly. Most stores haven't touched their crawl configuration in years. Technical SEO always mattered. Agent crawlability is the new version of it.
How to Measure These Signals
Start with a structured data audit using Google's Rich Results Test across a sample of 20-50 product pages. Run your feed through Google Merchant Center diagnostics and flag every suppressed item. For citation rate, build a prompt set of 10-15 queries and run them weekly across ChatGPT, Perplexity, and Gemini. Brand entity recognition can be spot-checked by asking AI assistants direct questions about your brand. Agent crawlability requires a technical review of your robots.txt and server logs. None of this requires expensive software. It requires treating AI assistants as a discovery channel with its own data requirements.
FAQ
Q: Do I still need to track keyword rankings?
Yes, for Google traffic that still runs through traditional results. For AI-generated product recommendations, keyword rankings tell you almost nothing. A page can rank on page one and still receive zero AI citations if its structured data is incomplete. Track both, but don't let rankings crowd out AI-specific signals.
Q: How quickly can I improve my structured data completeness score?
For a Shopify store on standard themes, adding a complete JSON-LD product schema block takes 1-2 days and immediately changes how AI systems read your pages. GTINs and feed quality take longer because they require catalog-level data work, but partial improvements show up in feed diagnostics within 24-48 hours of a resubmit.
Q: Does brand entity recognition apply to small or newer brands?
Yes. Newer brands have an advantage here: you can build the entity record correctly from the start. Establish consistent brand name, description, and category language across your site, Google Business Profile, and any third-party directories. Consistency is the signal. Older brands with inconsistent naming across years of content often score worse than newer brands that built it clean.
Q: What's the biggest mistake merchants make with agent crawlability?
Blocking legitimate AI agents in robots.txt by accident. Many stores use blanket "disallow" rules copied from SEO guides years ago that now inadvertently block shopping agents. The second most common mistake is product pages that require JavaScript execution to render price and availability, the two fields AI agents prioritize above everything else. If your price is in a JS-rendered element and the agent can't execute it, your product appears to have no price. That's an instant disqualification from AI-generated recommendations.
Want to know where your Shopify store stands on all five signals? We run AI Commerce Readiness Audits at WRKNG Digital: a full breakdown of what AI assistants can and can't see, with a clear list of what to fix first. Start here.
