Schema markup has long been treated as a tool for winning rich results — star ratings, images, FAQ snippets in the search results. But its real value runs deeper: schema markup for AI search is core infrastructure that helps search engines and large language models understand what a website actually is, by explicitly declaring the entities on a site and the relationships between them.
The starting point is the knowledge graph, which stores entities as nodes and their relationships as edges. That structure lets machines move beyond keyword matching into semantic understanding. Take a university business school as an example: rather than reading it as a string of text, a well-built knowledge graph recognizes it as a single entity — an organization that offers courses and employs people who teach them.
JSON-LD schema functions as an on-ramp for this understanding. Instead of waiting for a bot to infer entities and relationships from body copy, JSON-LD declares them directly. The point isn't to stuff in FAQ schema for rich-result clicks — a web page is really a vehicle carrying nodes and edges into the graph.
To operationalize this, one framework compares three layers: custom schema markup (in one case, more than 60 custom entities built on top of Schema.org's 23 standard types for a specific industry vertical), vector embeddings of site content, and an AI agent that analyzes semantic proximity between the two. Comparing these layers surfaces entity gaps — the missing pieces a content strategy should fill. The goal isn't implementing every possible entity; it's first defining what a complete "ideal entity model" looks like, then prioritizing the gaps by business value.
Research on schema's actual effect is split: one 2025 study found no impact on LLM visibility, while other research suggested LLMs don't read on-page schema at all. Countering that, Microsoft Bing has confirmed Copilot uses schema markup to understand content. The practical takeaway is to separate using JSON-LD to help machines understand a site from treating JSON-LD as a shortcut to citations, and to measure success beyond rich results — tracking prompts, brand sentiment, and AI mentions, then tying those back to actual conversions and lead quality.
For teams building out an AI-search content and structured-data strategy, Best Partner's services can help translate this into a prioritized roadmap, and our related piece on topical authority in ChatGPT category share digs further into the same shift.
It's the practice of using structured data (typically JSON-LD) to explicitly declare a site's entities and their relationships, helping both search engines and LLMs understand content beyond rich-result decoration.
The difference between an ideal, complete model of the entities a site should represent and what its current schema and content actually cover — found by comparing custom schema, content embeddings, and semantic-proximity analysis.
Evidence is mixed — some studies show no measurable effect, while Microsoft Bing has confirmed Copilot uses schema to understand content. The safer approach is to use schema for genuine machine understanding, not as a rich-result shortcut, and measure impact broadly.
Rather than implementing every possible entity, first define an ideal entity model for the business, then prioritize the resulting gaps by business value.
To apply what you just read to your own site, start with a free audit of where things are now.
A strategist replies within 24 hours on business days.
An E-E-A-T checker was never automatable because there is no API for it. Here is how an AI coding agent read Google's quality guidance, crawled a page sample, and produced a scored audit — plus where it stops.
A GEO audit is not checking whether your brand appears in an AI answer. Here is a nine-point GEO checklist split into four on-site items, four off-site, and one measurement layer.
LinkedIn, YouTube, Reddit, and TikTok are building systems to identify low-value AI content. Here are the limits of watermarking and what content still gets distributed.