# How to make your product content readable by AI shopping agents AI agents shortlist products by reading facts and evidence, not design. Fix three surfaces: structured data, plain-text claims, and a reachable review corpus. By Rohin Aggarwal · 2026-06-08 The PDP read perfectly to a human shopper and almost nothing to an LLM scraping it for an answer. The gap between those two reads is not aesthetic. It is structural: paragraphs vs facts, prose vs schema, brand voice vs canonical phrasing. What follows is the audit we run on every onboarding. **Quick answer** - AI shopping agents shortlist products by reading facts and evidence, not by appreciating your design. - Three surfaces decide whether you are readable: structured data, plain-text description, and the review/UGC corpus. - Structured data is the floor; customer evidence is what makes an agent confident enough to recommend you. - Most brands fail not on content quality but on accessibility, the proof exists, but it is trapped where an agent cannot reach it. Search-engine optimisation taught a generation of brands to write for a crawler. Agent readability is the next version of that discipline, sometimes called answer-engine or agent-engine optimisation, and it is stricter. A crawler indexed your page. An agent has to understand it well enough to stake a recommendation on it. The full discipline is laid out in [the answer-engine optimisation playbook](/blog/answer-engine-optimisation-2026-playbook). The good news: you do not need new content. You almost certainly already hold the facts and the customer proof. The work is making them legible to a literal reader, across the three surfaces below. ## What does an agent actually read? - Structured data: the machine-readable markup (Product, Offer, AggregateRating, Review schema) that states price, availability, attributes and rating as unambiguous facts. - The plain-text description, the words an agent extracts when it strips your page to text. Vague copy here is invisible copy. - The evidence corpus, reviews and customer photos/videos tied to the product, which the agent uses to corroborate every claim before it trusts it. ## Why is structured data the floor? If your product page does not emit clean structured data, an agent has to guess, and a cautious agent that has to guess will quietly prefer a competitor it does not have to guess about. Product, Offer and AggregateRating schema are not optional SEO garnish any more; they are the contract that says "here are my facts, unambiguously". Get them complete, accurate and consistent with what is on the visible page. The schemas worth prioritising are ranked in [the twelve JSON-LD shapes agents quote](/blog/twelve-json-ld-shapes-agents-quote). > **The schema floor:** At minimum, emit valid Product schema with price, availability and key attributes, plus AggregateRating and a sample of Review schema drawn from real, attributable reviews. Never synthesise reviews to inflate a rating count, agents and search engines both penalise it, and it is the fastest way to lose trust. ## Reviews and UGC are the evidence layer Structured data tells an agent what you claim. Reviews and customer media tell it whether the claim holds up. An agent comparing two similar products leans toward the one whose claims are corroborated by a deep, recent, specific body of customer evidence, because that is the recommendation it can defend. A page with rich specs and no evidence reads, to an agent, like a confident stranger. > Structured data gets you into the agent’s consideration set. Customer evidence is what gets you out of it with a recommendation. > — Rohin Aggarwal · Co-founder, Idukki ## Agent-friendly vs agent-invisible ### Two product pages, one agent _The difference is rarely content quality. It is whether the content is reachable._ **Loses the shortlist: Agent-invisible page** Beautiful to a human, opaque to a machine. - ✓ Strong photography and brand feel - ✗ Specs live only inside images - ✗ Reviews load in a slow third-party iframe - ✗ Claims are adjectives with no corroboration - ✗ Structured data missing or inconsistent **Wins the shortlist: Agent-readable page** Works for the human and the machine at once. - ✓ Complete, accurate Product + Review schema - ✓ Specific, plain-text claims - ✓ Customer photos/videos tied to the SKU, inline and fast - ✓ Every claim has evidence behind it - ✗ Takes a deliberate audit to get right the first time ## A checklist for this quarter 1. Validate Product, Offer and AggregateRating schema on your top product pages. 2. Rewrite vague claims as specific, plain-text facts that survive being stripped to text. 3. Make sure Review schema is drawn only from real, attributable reviews with real dates. 4. Pull customer photos and videos onto the PDP, tied to the product, rendered fast and crawlable. 5. Tag that customer media so its subject matter is itself machine-readable. 6. Test: paste your product URL into an AI assistant and ask it to summarise and compare the product. Read what it gets wrong. 7. Fix the gaps it revealed, and repeat the test each month. > **How Idukki helps:** Idukki surfaces customer photos, videos and reviews on the product page as fast, structured, crawlable evidence, and its [AI tagging](/blog/ai-content-tagging-for-ugc) makes the subject matter of each piece machine-readable. That is the evidence layer an agent needs, built from content you already have. **In one line:** AI assistants now shortlist products before a shopper ever lands on your page. If an agent cannot parse your product truth, it cannot recommend you. Here is the practical checklist for being agent-readable. - **41%** — of 18-34s use GenAI to search products (Bazaarvoice 2025 SEI) - **14x** — AI weight on verified-buyer reviews (Industry consolidated) - **$2.6T** — Global social commerce 2025 (eMarketer) - **+108%** — TikTok Shop US growth YoY 2025 (eMarketer) _AEO + AI shopping benchmarks._ ### Sources & notes - [Schema.org: Product, Offer, AggregateRating, Review types](https://schema.org/Product) — The structured-data vocabulary agents and search engines consume. - [Google Search Central, structured data for products](https://developers.google.com/search/docs/appearance/structured-data/product) — Implementation guidance and validation. - [Baymard Institute, product-page UX research](https://baymard.com/research) — What content shoppers and agents rely on. --- Canonical: https://idukki.io/blog/make-product-content-readable-by-ai-shopping-agents Tags: ai-search, agentic-commerce, structured-data, seo