{
  "id": "G06",
  "url": "https://cartforagents.com/guides/agent-comparison-cost",
  "title": "Buried product facts: why AI assistants skip them when comparing products",
  "description": "When an AI shopping assistant has to dig for a fact, it often compares without it. What the research shows, where facts hide on Shopify listings, and a worksheet to count the steps on your best seller.",
  "answer": "When a product fact takes extra steps to reach (a variant, a size chart image, the bottom of a long description, another page), AI shopping agents often compare products without it, especially when the shopper's request is vague. In a lab study across eight models, hiding facts behind costly lookups cut correct picks from 89% to 29% under a vague request. Put the fact your shoppers ask about in the title or the first lines of the description, as a number with its unit.",
  "take": "Agents don't read your listing the way you do; they skim it under a budget. The cheapest fix in agent commerce is moving the deciding fact to where it gets read first.",
  "summary": "An AI assistant comparing four products has a budget, like any shopper. Every fact it has to dig for is a step it might not take. Put the fact a shopper asks about where it is read first, and comparing you costs nothing. Here is what the research shows, and how to check your own listings.",
  "claims": [
    {
      "text": "In the Tool-Lab study, eight commercially deployed models from Google, OpenAI and Anthropic chose between two coffees whose dollars, cents and weight were hidden behind lookups. With free lookups, they picked the better-value coffee in 94% of sessions when asked for the lowest price per ounce and 89% when asked to \"find the best deal\". At $10 per lookup, the specific request still got 83% right, while the vague one fell to 29%. Under the vague request, the models kept checking dollars but skipped cents and weight.",
      "source_title": "Wadi & Ma, Shopping by algorithm: How agentic AI deploys human heuristics as a surrogate consumer (arXiv 2609.28372v2), Table 1",
      "source_url": "https://arxiv.org/html/2609.28372v2",
      "source_date": "2026-10-13",
      "method": "controlled-simulation",
      "limitation": "A lab task with two invented coffees and an explicit cost per lookup. Not a test of ChatGPT shopping, Shopify Catalog or any live ranking."
    },
    {
      "text": "The Tool-Lab authors report that validation experiments traced most of the wrong choices to missing information rather than arithmetic mistakes.",
      "source_title": "Wadi & Ma, Shopping by algorithm (arXiv 2609.28372v2)",
      "source_url": "https://arxiv.org/html/2609.28372v2",
      "source_date": "2026-10-13",
      "method": "controlled-simulation",
      "limitation": "The authors' own validation, on their task."
    },
    {
      "text": "In the ACES marketplace experiments, the office-lamp category was the only one where an AI-rewritten description gained share consistently across all six buying models tested. The authors attribute it to the original description not stating \"Office\" early, so it was cut off on their mock platform; the rewrite moved \"Office\" to the front.",
      "source_title": "Allouah et al., What Is Your AI Agent Buying? (arXiv 2508.02630v3)",
      "source_url": "https://arxiv.org/html/2508.02630v3",
      "source_date": "2026-11-04",
      "method": "controlled-simulation",
      "limitation": "A mock marketplace with its own description cut-off. The authors offer the truncation explanation as a hypothesis."
    },
    {
      "text": "Shopify documents a two-step flow for agents using its Global Catalog: first search or look up matching products, then call get_product on a product or variant to explore options and get pricing and checkout links.",
      "source_title": "Shopify: Global Catalog MCP, How it works",
      "source_url": "https://shopify.dev/docs/agents/catalog/global-catalog",
      "source_date": "2026-11-04",
      "method": "documented",
      "limitation": "Describes the tools Shopify offers agents. It does not say how often any assistant makes the second call."
    },
    {
      "text": "Shopify marks fields it generates or enriches as Inferred, including the catalog description, attributes, technical specifications, top features and unique selling points; these might be missing or vary in accuracy and should be treated as discovery and merchandising signals, not merchant-authored facts.",
      "source_title": "Shopify: Global Catalog MCP, Inferred response fields",
      "source_url": "https://shopify.dev/docs/agents/catalog/global-catalog",
      "source_date": "2026-11-04",
      "method": "documented",
      "limitation": "Living developer documentation. It does not say how assistants weigh these fields."
    },
    {
      "text": "OpenAI says ChatGPT may generate simplified product titles and descriptions from information it receives from third-party providers.",
      "source_title": "OpenAI Help Center: Shopping with ChatGPT Search",
      "source_url": "https://help.openai.com/en/articles/11128490-shopping-with-chatgpt-search",
      "source_date": "2026-11-04",
      "method": "documented",
      "limitation": "Living help page. It does not say which facts survive the shortening."
    },
    {
      "text": "For the search \"humidifier for dry winter air large room quiet top fill\", the listing at position 9 said \"low noise level\" in its description, while only its Shopify-generated spec list said \"Noise Level: 26 dB\"; no room coverage appeared anywhere in its record.",
      "source_title": "Cart for Agents: AI visibility audit, A real example",
      "source_url": "https://cartforagents.com/guides/ai-visibility-audit",
      "source_date": "2026-10-26",
      "method": "observed",
      "limitation": "One unpersonalized Shopify Catalog snapshot. Quotes what the record said, not how loud the product is."
    },
    {
      "text": "For the search \"toddler snowsuit rated for minus 30\", the listing at position 2 stated its rating in the title: \"protects to -30 °C\".",
      "source_title": "Cart for Agents: Receipts cited by our guides",
      "source_url": "https://cartforagents.com/research/agent-shopping-research#receipts-cited-by-our-guides",
      "source_date": "2026-10-26",
      "method": "observed",
      "limitation": "One saved Shopify Catalog result. Quotes the title, not the product's real warmth."
    },
    {
      "text": "For a small Canadian lip gloss brand's gift box, checked as a Canadian shopper on October 28, 2026, Shopify Catalog held a blank product-level description while the variant record carried the full text. By November 4 the product-level description had arrived in full.",
      "source_title": "Cart for Agents: AI product visibility case study",
      "source_url": "https://cartforagents.com/research/ai-product-visibility-case-study",
      "source_date": "2026-10-28",
      "method": "observed",
      "limitation": "One product, one saved catalog record. We don't know which record each assistant reads first."
    }
  ],
  "sources": [
    {
      "title": "Wadi & Ma: Shopping by algorithm (arXiv 2609.28372v2)",
      "url": "https://arxiv.org/html/2609.28372v2",
      "accessed": "2026-11-04"
    },
    {
      "title": "Allouah et al.: What Is Your AI Agent Buying? (arXiv 2508.02630v3)",
      "url": "https://arxiv.org/html/2508.02630v3",
      "accessed": "2026-11-04"
    },
    {
      "title": "Shopify: Global Catalog MCP",
      "url": "https://shopify.dev/docs/agents/catalog/global-catalog",
      "accessed": "2026-11-04"
    },
    {
      "title": "OpenAI Help Center: Shopping with ChatGPT Search",
      "url": "https://help.openai.com/en/articles/11128490-shopping-with-chatgpt-search",
      "accessed": "2026-11-04"
    },
    {
      "title": "Cart for Agents: AI visibility audit",
      "url": "https://cartforagents.com/guides/ai-visibility-audit",
      "accessed": "2026-11-04"
    },
    {
      "title": "Cart for Agents: Receipts cited by our guides",
      "url": "https://cartforagents.com/research/agent-shopping-research#receipts-cited-by-our-guides",
      "accessed": "2026-11-04"
    },
    {
      "title": "Cart for Agents: AI product visibility case study",
      "url": "https://cartforagents.com/research/ai-product-visibility-case-study",
      "accessed": "2026-11-04"
    }
  ],
  "limitations": [
    "The research here comes from lab studies, not from ChatGPT, Gemini or Copilot themselves. No assistant publishes how many steps it will take to check a fact.",
    "The \"steps\" in our worksheet are our own rule of thumb for how hard a fact is to reach, not a measurement of any assistant.",
    "The listing examples come from saved Shopify Catalog results from October 26 and 28, 2026. A live check today could look different.",
    "Store and brand names are withheld."
  ],
  "related": [
    "https://cartforagents.com/guides/agent-commerce-optimization",
    "https://cartforagents.com/guides/product-constraint-coverage",
    "https://cartforagents.com/guides/product-claim-provenance",
    "https://cartforagents.com/guides/negative-product-attributes",
    "https://cartforagents.com/guides/agent-ready-product-data",
    "https://cartforagents.com/guides/ai-product-descriptions",
    "https://cartforagents.com/guides/size-chart-accessibility",
    "https://cartforagents.com/research/vague-specific-product-facts-simulation",
    "https://cartforagents.com/guides/clipped-product-descriptions"
  ],
  "published": "2026-11-04",
  "updated": "2026-11-04",
  "checked": "2026-11-04",
  "kind": "guide"
}