Episode 14: The AEO Podcast EP 14: Product Data Is the New Product Page

HOST: Welcome back to the AEO Podcast. Jessica here with Matt. Today's episode has a very 2026 title: product data is the new product page.

MATT: Which is annoying because product pages still matter.

HOST: Great start.

MATT: They do. Humans still land on product pages. Google still indexes product pages. But AI shopping systems increasingly see your product as a bundle of data: title, description, images, price, availability, variants, reviews, attributes, policies, and feed fields. If those are thin or inconsistent, the beautiful product page does not matter as much as you think.

HOST: Why is this current now?

MATT: Because ChatGPT Shopping, Shopify Agentic Storefronts, Google AI Mode, Gemini, Copilot, and Perplexity are all moving toward product comparison experiences. The AI is not just answering "what is this store?" It's helping a buyer choose between specific products.

HOST: What does an AI need to compare products well?

MATT: It needs comparable facts. Price. Size. Material. Use case. Availability. Reviews. Shipping. Return policy. Compatibility. Constraints. If your competitor has those facts and you don't, the AI has a confidence gap.

HOST: Confidence gap. Explain that.

MATT: AI systems don't only ask "is this product relevant?" They also ask, implicitly, "do I have enough evidence to recommend it?" A product with a vague description is risky. The AI might be wrong about fit, ingredients, compatibility, or shipping. So it often chooses the product with clearer data.

HOST: This is not just SEO copywriting then.

MATT: No. This is merchandising data discipline. AEO is exposing messy catalogs.

HOST: Let's define a good product data set.

MATT: I think about it in four buckets: identity, proof, logistics, and machine-readable consistency.

HOST: Start with identity.

MATT: Identity is the basic "what is this?" layer. A descriptive product title. Not keyword-stuffed, but specific. "Organic Cotton Baby Swaddle - Sage Green - 47 by 47 inches" beats "The Willow Swaddle." Then a first paragraph that says what it is, who it is for, and why someone buys it. Then variant names that make sense outside your site. "Small, black, regular" is better than internal codes.

HOST: Proof?

MATT: Proof is the evidence layer. Attribute bullets: material, size, dimensions, fit, care, compatibility, dietary or safety details if relevant. High-quality images with descriptive alt text. Reviews that mention real use cases, not just "love it." If you use a review app like Judge.me, Yotpo, Loox, or Shopify's own review tooling, ask questions that create useful detail.

HOST: Logistics?

MATT: Current price and availability. Shipping and return expectations. Care instructions. Compatibility limits. For replacement parts, that might be model numbers, dimensions, voltage, and "does not fit the 2022 version." For books, it might be format, page count, series order, genre, tropes, and age suitability. These facts are not decoration. They are how AI shopping assistants compare options.

HOST: And machine-readable consistency.

MATT: Structured data or feed fields that match the visible page. Google Merchant Center, Shopify Catalog, product feeds, schema, and the product page should all tell the same story. If the feed says one price and the page says another, trust drops. If the structured data says there are reviews but the page doesn't show reviews, that is a problem. Then add FAQs that answer objections, but only if the answers are actually product-specific.

HOST: That is a lot.

MATT: It is. But notice how little of it is "write more." Most of it is "make the facts explicit."

HOST: Let's do a bad and good example.

MATT: Bad product: "The Everyday Tote. Our bestselling bag, made for modern life. Stylish, durable, and versatile."

HOST: I feel like I've read that page.

MATT: Everyone has. Good product: "Structured vegan-leather work tote that fits a 13-inch laptop, water bottle, notebook, and phone. Zipper top. Interior laptop sleeve. Weighs 1.8 pounds. Best for commuting and office use. Not designed for 15-inch laptops."

HOST: That last sentence is surprisingly important.

MATT: It is. AI assistants are often asked for constraints: "fits a 15-inch laptop," "good for wide feet," "safe for toddlers," "works with iPhone 17." If the answer is no, say no. Accurate exclusions build trust and reduce returns.

HOST: Does saying no hurt conversions?

MATT: It hurts the wrong conversions. It helps the right ones. Shopify's own 2026 analysis says AI-referred shoppers often arrive with higher intent, convert at higher rates, and spend more. That makes sense because the comparison happened before the click.

HOST: Reviews sit in the proof bucket, right?

MATT: Yes. Reviews are underrated AEO data. Not because AI systems blindly trust star ratings, but because reviews contain real buyer language. "Runs small." "Good for sensitive skin." "Bought it for my dad." "Held up after three washes." That language maps to prompts.

HOST: Should merchants edit reviews?

MATT: No. Never fake or rewrite reviews. But you can structure review collection. Ask post-purchase questions that invite useful detail: What did you use it for? What size did you buy? What problem did it solve? Would you recommend it for a specific use case?

HOST: Images are proof too?

MATT: Images matter twice. First, shopping interfaces are visual. ChatGPT's current shopping experience is built around visual browsing and comparison. Second, image alt text and captions are text signals. "Model is 5 foot 8 wearing size medium" is useful. "IMG_2048" is not.

HOST: Where do product feeds fit?

MATT: Feeds are the structured version of the same truth. OpenAI says product feeds help merchants control how products appear and keep information up to date. Google Merchant Center has worked this way for years. Shopify Catalog is the big Shopify-native pipe. The feed should not contradict the product page.

HOST: If I'm in Shopify Admin, what do I actually touch?

MATT: Product title and description first. Then variants. Then media alt text. Then metafields for specs that should be consistent across products. Then policies, reviews, and catalog or feed status. That is the practical admin pass. You don't need to solve every product in one sitting. Start with ten products that already matter.

HOST: What's the first SEOMelon workflow for this?

MATT: Scan for thin products. Pick the products that matter: best sellers, high-margin SKUs, products with ad spend, products that should show up in AI recommendations. Generate improved descriptions and FAQs. Then review them like a merchant, not like a content marketer. Are the facts true? Are the constraints clear? Does the copy answer real buying questions?

HOST: What should they avoid?

MATT: Avoid bloated product descriptions that sound AI-written. Avoid adding attributes you can't verify. Avoid hiding important details in images or accordions that don't render cleanly. Avoid using the same FAQ on every product. Generic content gives AI systems fewer product-specific signals, and it makes every product sound interchangeable.

HOST: Give me the practical takeaway.

MATT: Your product page is still the storefront. But your product data is the thing AI systems carry into the conversation. If the data is vague, your product gets summarized vaguely. If the data is clear, your product has a chance to be recommended clearly.

HOST: What's the first pass?

MATT: Take one best-selling product. Add ten explicit facts that a buyer would otherwise have to infer. Then make sure those facts appear on the page, in the feed where applicable, and in structured data when appropriate.

HOST: Next episode, we're checking whether the crawlers can even reach all this improved product data.