Shero Commerce, a Shopify agency specializing in SEO and AI services, audited 1,851 sources cited by Google AI Mode, ChatGPT, and Perplexity across 60 product categories. Brand-owned pages accounted for only 2.8% of those citations, while a single category of third-party sites, including Good Housekeeping, Verywell Fit, and Reviewed.com, made up 59%. The gap separates two outcomes ecommerce teams have started to treat as one: being named in an AI answer and being the site that answer actually links to.
Shero also tracked cases where AI tools recommended a brand by name rather than just citing a source. Across the three platforms, it found 159 such recommendations, and the recommended brand’s own page was cited in only 31% of them. For a squat-proof leggings query, both ChatGPT and Perplexity named Gymshark, Alo, and Beyond Yoga, then backed those recommendations with links to third-party retailers rather than the brands’ own sites.
Google AI Mode showed the narrowest reach into brand-owned content. It cited or recommended a sampled brand in 9.5% of relevant store checks across the 60 categories, and roughly one-third of those categories returned no sampled brand at all. Shero did not report a comparable recommendation rate for ChatGPT or Perplexity, so the 9.5% figure describes Google’s behavior in isolation. It is not a benchmark against the other two platforms, a distinction the report is careful to flag.
Shero frames the citation gap as connected to duplicate content. In a stratified sample of Shopify stores, 20% of product descriptions matched or closely resembled text published elsewhere, mostly among retailers and marketplaces reselling the same products rather than between competing brands. That is a plausible mechanism: identical text across multiple domains gives an AI system less reason to prefer any one source. But the report stops short of proving it. It documents the overlap without testing whether removing it would shift which domain gets cited.
Shero’s own methodology adds a second caveat. Of 173 stores it could check against raw HTML, 27 carried fewer than 50 words of product-specific copy, meaning the description fields Shero first measured undercount how much original text some pages actually contain. The full dataset behind the findings covers 8,573 product descriptions from 883 stores with live product data, a sample large enough to establish the pattern but not, on its own, to explain it.
The report’s clearest limitation is the experiment it does not run. Shero has not tested whether rewriting syndicated copy, adding original description text, or restructuring a product page changes what gets cited. That test is exactly what would tell an SEO team whether content investment moves citation share, and the absence of it means the “why” behind the 2.8% figure stays open.
For ecommerce brands tracking AI Mode, ChatGPT, and Perplexity mentions as a visibility channel, brand-mention counts alone overstate the win. A 31% self-citation rate on named recommendations means most of that visibility still routes the click to a competitor’s retail partner or a review site. Teams should audit whether their own product pages carry enough unique, crawlable copy to compete with the marketplaces and review sites currently winning the citations, rather than assuming a brand mention converts into traffic.
Search Engine Journal reported the Shero Commerce findings on August 26, 2026.
