GEO (generative engine optimization) is rapidly becoming the new SEO. If you’re eager to maximize your chances to be included in ChatGPT’s product recommendations (and other LLMs), here’s your up-to-date guide to self-audit your site and be ready for what’s next.

✅ LLM Discovery Content Audit Checklist

🛍 Product Page Audit

• [ ] Product title includes brand, type, key attributes, and purpose Helps LLMs understand and surface your product when users search for specific needs (e.g. “lightweight waterproof hiking jacket”)
• [ ] Meta description and OG tags use natural language These are often paraphrased by LLMs into product summaries, so keep them conversational and benefit-led
• [ ] Product description uses plain language and semantic keywords Use natural, descriptive language with benefits and use cases that match how users talk (e.g. “perfect for sensitive skin”)
• [ ] Key specs and attributes are presented in structured format Use bullet points or tables (<ul> list, JSON-LD schema) to clearly display facts like material, size range, or dimensions. These formats are easy for LLMs to extract
• [ ] FAQ section answers common natural-language queries Include questions like “Is this machine washable?” or “Does it run true to size?” to match real prompt formats
• [ ] Comparison language is included Phrases like “more breathable than Brand X” or “budget alternative to Y” help LLMs make side-by-side recommendations
• [ ] Alt-text for images describes the product visually and contextually Write alt text that clarifies what's shown, e.g. “female model wearing navy trench coat outdoors in the rain”
• [ ] Review content includes fit, quality, and real-world use cases Detailed reviews help LLMs generate natural product endorsements or recommendations based on user experience
• [ ] Schema.org markup is implemented and validated Use ****Product, AggregateRating and Offer schemas to improve parsing by AI systems
• [ ] Page speed and mobile optimization are excellent Fast-loading, mobile-first pages improve crawlability and LLM-generated rankings, especially via Google AI and ChatGPT plugins

🖼 Image Optimization Audit

<aside> 💡

IMPORTANT: LLMs will not use your full PDP, rather they extract product data including images and render them inside their own UI containers with specific aspect ratios (carousels, visual product cards and answer boxes with embedded images) that have their own needs and compression limits.

If your images don’t meet those standards, your product may either look broken, be down-ranked, or be omitted entirely.

</aside>

• [ ] Primary product images are square and high-resolution (minimum 1024x1024 px)

  Required for clean rendering in ChatGPT carousels, Perplexity product cards, and Klarna’s AI listings.

• [ ] Consistent aspect ratio (1:1) across PDPs and collection pages

  Prevents unpredictable cropping or visual distortion in LLM-generated layouts.

• [ ] File sizes are optimized (under 300KB) without loss of quality

  Ensures fast load times, which improves visibility in AI ranking algorithms.

• [ ] ALT text clearly and naturally describes the image

  Helps LLMs understand what’s depicted — e.g., “close-up of organic cotton T-shirt in sage green on hanger.”

• [ ] Filenames are semantic and descriptive

  Use names like linen-wide-leg-pants-tan.jpg instead of IMG_3849.jpg.