The search engine optimization landscape is undergoing one of its most disruptive periods in history. For years, structured data—commonly referred to as schema markup—was championed as the ultimate bridge between human-readable web content and machine-readable databases. SEO professionals spent countless hours writing, testing, and deploying JSON-LD code to earn coveted rich snippets, enhance click-through rates (CTR), and ensure search engines understood their content’s context.
However, two major developments have sent shockwaves through the search marketing community, forcing a critical re-evaluation of structured data’s actual value. First, Google systematically stripped away FAQ and How-To rich results from the search engine results pages (SERPs). Second, groundbreaking research from Ahrefs has challenged a foundational belief: that schema markup is a critical driver for gaining citations in modern, AI-powered search engines.
As search engines transition from classic keyword retrieval systems into complex, AI-driven answer engines, SEOs must adapt. Here is an in-depth exploration of why these changes occurred, what the latest data reveals, and how you should pivot your optimization strategy to remain visible in both traditional and generative search landscapes.
The Evolution and Sudden Fall of Google’s FAQ Rich Results
To understand why schema’s value is being questioned, we must first look at how Google handled visual real estate on the SERPs over the past decade. Schema markup was introduced in 2011 by a collaboration between Google, Bing, Yahoo!, and Yandex under the Schema.org initiative. The goal was simple: create a shared XML/JSON-LD vocabulary to help search engines understand what web pages are actually about.
For years, implementing schema yielded highly visible rewards. The most popular of these was FAQ schema. By adding a few lines of code to a webpage, publishers could display dropdown question-and-answer accordions directly beneath their organic search listings. This provided several distinct advantages:
- Increased SERP Footprint: A single listing with FAQ schema could occupy twice the vertical space of a standard listing, pushing competitors further down the screen.
- Improved Click-Through Rates (CTR): Interactive elements naturally draw the human eye, leading to higher engagement.
- Pre-empting User Intent: Answering common queries directly on the search page established immediate authority.
However, this led to widespread manipulation. SEOs began adding irrelevant FAQ schema to almost every page to monopolize SERP real estate. In response, Google gradually rolled back the feature. The final blow came when Google officially announced it would restrict FAQ rich results to highly authoritative, well-known health and government websites, effectively rendering the markup useless for the vast majority of commercial and informational publishers.
This move signaled a broader shift in Google’s strategy. The search giant was no longer interested in giving away valuable SERP real estate for free to webmasters who simply optimized their structured code. Instead, Google began preparing its interface for a clean, streamlined look designed to accommodate its own AI-generated answers.
The Ahrefs Study: Deconstructing Schema’s Value in AI Search
With traditional rich results fading, many SEO professionals shifted their narrative. They argued that even if schema no longer generated visual snippets on standard Google search pages, it remained vital for “AI SEO.” The theory was that Large Language Models (LLMs) and conversational search engines (like Perplexity AI, Google’s AI Overviews, and ChatGPT) relied heavily on structured data to crawl, parse, and cite sources.
To test this hypothesis, the data science team at Ahrefs conducted a comprehensive study analyzing the relationship between schema markup and citations in generative search engines. The findings challenged the long-held assumption that structured data is a prerequisite for AI visibility.
Key Finding 1: AI Search Engines Skip the Schema
The research revealed that generative search engines do not rely on JSON-LD or microdata to understand content and generate citations. Instead, these advanced engines process the raw, unstructured HTML and natural language of a page. Because LLMs are trained on massive datasets of human language, they are incredibly proficient at understanding context, relationships, and entities directly from standard text without needing a structured code translation.
Key Finding 2: High Citation Rates Occur Without Schema
Ahrefs analyzed a vast pool of queries that triggered AI search summaries. They discovered that a significant portion of the web pages cited by AI search engines did not have specialized schema implemented. Pages with simple, clean semantic HTML (such as standard paragraph tags, bulleted lists, and clear table formats) were cited just as frequently—and in some cases, more frequently—than pages heavily optimized with complex schema packages.
Key Finding 3: Crawl Efficiency vs. Contextual Understanding
While search engines do use structured data to verify specific factual details (such as price, availability, or event dates), their retrieval-augmented generation (RAG) pipelines do not rely on schema to synthesize answers. The RAG systems pull directly from the visible text on the page to build their conversational responses. If your content is buried inside structured data but isn’t clear, readable, or valuable on the front end of the page, the AI crawler is highly likely to ignore it.
Why LLMs and AI Search Engines Don’t Need Schema Anymore
To comprehend why schema’s role is shrinking, we have to look at the underlying technology powering modern AI search. Early search engines were syntactic; they matched keywords on a page to keywords in a search query. Schema was a crutch that helped these basic algorithms understand that “Apple” referred to the technology company and not the fruit.
Today’s search systems are semantic. They utilize vector embeddings and natural language processing (NLP) to understand concepts, user intent, and real-world entities. Here is why modern AI search systems can bypass schema entirely:
1. Advanced Semantic Understanding
Modern LLMs process text by converting words into mathematical vectors in a multi-dimensional space. This allows them to understand synonyms, tone, context, and structural relationships naturally. An AI doesn’t need a JSON-LD tag telling it “this is an author” when it can read the sentence “Written by Jane Doe, a certified financial analyst with ten years of experience” and extract that entity relationship instantly.
2. The RAG (Retrieval-Augmented Generation) Workflow
When you ask an AI search engine a question, it executes a multi-step process:
- It converts your prompt into a search query.
- It retrieves a set of relevant documents from its index.
- It parses the actual text of those documents to extract answers.
- It synthesizes those answers into a coherent paragraph and links to the source.
During this pipeline, the system extracts information from the visible text blocks on the page. Because schema is hidden in the site’s head section or code base, it often gets stripped out or bypassed during the initial text extraction phase of the RAG pipeline.
3. HTML Parsing Has Improved
Web parsers used by companies like OpenAI, Perplexity, and Anthropic are incredibly sophisticated. They can easily clean up messy HTML, identify main content areas, discard navigation bars and footers, and construct a clean text profile of a page. As a result, the “translation” service that schema once provided is no longer necessary for information retrieval.
Is Schema Markup Dead? Where Structured Data Still Matters
While the value of schema for general informational search and AI citations has declined, it is an oversimplification to say that schema is dead. Instead, its utility has become highly specialized. SEOs must stop treating schema as a universal ranking booster and instead focus on the specific areas where search engines still require structured code to perform transactional functions.
1. Merchant Center and E-commerce Search
Product schema remains absolutely critical. For e-commerce sites, schema provides real-time transactional data that search engines use to populate Google Shopping, product search grids, and price comparison features. If you want Google to show your product’s price, availability (in stock/out of stock), shipping costs, and customer review stars directly in the search results, you must use Product schema.
2. Local SEO and Organization Data
For local businesses, LocalBusiness and Organization schema are vital for establishing entity authority. This data feeds directly into Google’s Knowledge Graph and Google Maps. It confirms your business’s physical address, phone number, operating hours, and official social media profiles, reducing the risk of your business information being conflated with a competitor’s.
3. Event and Job Postings
Google maintains dedicated, highly structured interfaces for job boards and local events. These interfaces operate almost entirely on structured data. If you run a job board or event platform, omitting JobPosting or Event schema will completely exclude your pages from these specialized search widgets.
4. Video and Recipe Search
Recipes and videos rely heavily on visual grid systems in Google Search. VideoObject and Recipe schema allow Google to display cooking times, calorie counts, video runtimes, and thumbnail previews. Without this structured data, your rich media content is highly unlikely to surface in these interactive search features.
How to Pivot Your SEO Strategy for the AI Era
If over-engineering your schema markup is no longer the key to winning search visibility and AI citations, where should you direct your time, budget, and resources? The answer lies in optimizing for human clarity, entity authority, and high information gain.
1. Prioritize Clean Semantic HTML
Stop relying on code-heavy solutions to explain your content to search engines. Instead, design your pages with pristine semantic HTML. Use clear, nested heading structures (H1, H2, H3) that logically break down your topic. Utilize native HTML lists (<ul>, <ol>, <li>) for step-by-step guides and clean data tables (<table>) for comparative data. AI search crawlers can parse structured HTML tables and lists much faster and more reliably than complex custom JSON-LD.
2. Write with “Information Gain” in Mind
AI search engines summarize existing web content. If your article simply repeats the same information found on ten other websites, an AI has no reason to cite your page. Focus on “information gain”—providing unique insights, original research, expert quotes, first-party data, or case studies that cannot be found elsewhere. AI engines value unique data points and are highly likely to cite the original source of a novel stat or expert quote.
3. Build Strong Entity Connections
Instead of relying solely on Organization schema to tell Google who you are, build your brand’s footprint across the web. Seek coverage in reputable publications, maintain active and authoritative social media profiles, and ensure your brand name is consistently associated with your specific niche. When search engines crawl the web and find your brand consistently mentioned alongside industry topics, they construct strong entity associations in their knowledge bases, which naturally leads to more AI search citations.
4. Keep Schema Simple and Automate Where Possible
Do not waste valuable developer resources manually writing custom schema for every blog post or article. Use standard CMS plugins (like Yoast, RankMath, or built-in Shopify frameworks) to automatically generate basic Article, Breadcrumb, and Organization schema. Ensure it is error-free, but do not spend hours writing intricate nested schemas that offer diminishing returns.
Conclusion: Quality Content and Natural Architecture Win
The removal of FAQ snippets and the fresh insights from Ahrefs’ research mark a healthy evolution in the search industry. The era of hacking search visibility through code manipulation and schema stuffing is coming to an end. Search engines have grown smart enough to read our websites the way humans do.
Moving forward, the websites that win in both traditional organic search and the generative AI ecosystem will be those that prioritize user experience, natural language clarity, and unique, high-value information. Schema remains a helpful tool for specific transactional and local contexts, but the true driver of visibility in the age of AI search is—and will continue to be—unquestionable content quality and clear semantic site architecture.