How does AI Search work and which factors influence recommendations?
AI systems like ChatGPT, Perplexity and Gemini use complex algorithms to recommend products based on training data and real-time web search. They analyze structured information to detect patterns and generate relevant recommendations. The factors that influence these recommendations include:
- Structured product data: Schema.org markup with name, price, brand, availability and properties.
- Crawlability: Static, publicly reachable URLs without dependence on JavaScript rendering.
- Content precision: Exact, factual descriptions, not marketing phrases.
- Context and authority: Internal linking, mentions on trustworthy sources and consistent product data.
AI systems read, they do not rank. Whoever provides structured, machine-readable product data holds the biggest advantage.
Differences between AI ranking and classic Google ranking
AI ranking is fundamentally different from classic Google ranking. While Google mainly judges the relevance and authority of web pages through backlinks and keywords, AI systems extract facts directly from the content:
- Personalization: AI answers are tailored to the specific user question, not to generic keywords.
- Direct fact extraction: AI systems read Schema.org, JSON-LD and factual descriptions, not keyword density.
- Interactive context processing: The user question determines which product gets recommended, not backlink strength.
More on the difference: AI SEO vs. classic SEO — the full comparison →
The role of structured data, entities and context
Structured data is the single most important lever for AI visibility. It helps AI systems identify products clearly and name them in relevant answers. The three most important aspects:
- Entities: Clear definition of products, brands and categories via Schema.org Product markup.
- Contextualization: Metadata on purpose, target audience and product category helps AI systems with classification.
- Schema markup: JSON-LD with
@type: Product, name, brand, offers, description — complete, not partial.
Why many products do not appear in AI recommendations
There are typical patterns behind why products are missing from AI answers despite high quality:
- Missing structured data: Without Schema.org markup, the AI system cannot identify the product.
- JavaScript rendering: Products that load only via JavaScript are invisible to AI crawlers.
- Marketing language instead of facts: Phrases like "innovative quality product" are ignored, exact specifications are preferred.
- No permanent URL: Session parameters, temporary links or marketplace URLs (Amazon, eBay) provide no stable reference.
For a detailed analysis: Why your products do not appear in ChatGPT →
Concrete measures to increase visibility
To increase the visibility of your products in AI systems in a targeted way, the following measures are effective:
- Implement Schema.org markup: JSON-LD with a complete Product schema — name, brand, offers, description, image.
- Optimize product descriptions: Exact specifications (dimensions, weight, voltage, compatibility) instead of advertising copy.
- Create dedicated product pages: Public, static URLs in addition to your shop or marketplace.
- Use Feed-AI: Automatic generation of AI-optimized product pages with complete Schema.org markup and manual AI visibility tracking (Pro/Business).
- Keep content current: Regular updates preserve indexing relevance.
Dedicated, public product pages with complete Schema.org markup are the fastest route to AI visibility, independent of your shop system or marketplace.
Frequently asked questions
How do AI systems recommend products and services? +
AI systems like ChatGPT and Perplexity recommend products and services based on training data and real-time web search. They extract facts from structured data (Schema.org), crawlable URLs and factually precise descriptions. Whoever provides this data in full gets recommended more often.
What are the most important factors for AI recommendations of products and services? +
The three main factors are: complete Schema.org markup (JSON-LD), a permanent crawlable URL without JavaScript rendering, and clear factual descriptions with exact specifications or performance features instead of marketing phrases.
How quickly do AI systems pick up changes? +
Depending on the platform, it takes 2 to 8 weeks for AI systems to process new or updated content. Perplexity, with its real-time search, reacts faster; ChatGPT (without web search) relies on training cycles (according to the OpenAI GPTBot documentation and Perplexity AI Docs, as of 2025). With Feed-AI you can check visibility across all three systems from the Pro plan onward.