What the research says against the hype
In May 2026, Ahrefs tracked nearly 1,900 pages that had added structured data between August 2025 and March 2026, comparing them against control pages. The result was sobering:
“Adding schema did not produce any notable increase in citations on any platform.”
Louise Linehan and Xibeijia Guan, Ahrefs, May 2026
A second finding points in the same direction. When a team in October 2025 tested what ChatGPT, Claude, Perplexity, and Gemini actually process when fetching a page directly, all four systems ignored the structured data entirely. They read only the visible text. Anyone who believes a piece of JSON-LD in the source code will be picked up by an AI service and dutifully rewarded with a citation is mistaken, based on current evidence.
Why it still matters
The conceptual error is assuming that schema acts at the moment of the query. It does not. It acts one step earlier, during indexing and in determining whether a search engine understands your business as a clearly defined entity or as a vague pile of data. Structured data tells a search engine unambiguously who you are, what you offer, and where. That clean understanding is the prerequisite for being considered a reliable source in the first place.
How significant this has become was illustrated in June 2025. In a single week, Google removed more than three billion entries from its Knowledge Graph, the network of entities that underlies search. Vague, ambiguous entries were hit hardest. Businesses with clean definitions and stable links to recognized sources such as Wikidata or the commercial register survived. Well-placed Organization schema is exactly what achieves this.
Microsoft has also confirmed that Bing uses structured data for its AI systems. The correlation is measurable: pages cited by AI carry structured data significantly more often than pages that are not cited. But that is correlation. Schema lives on well-maintained pages; it does not turn a poor page into a good one.
The most common mistake: duplicate and contradictory schema
In practice, schema rarely fails because it is missing. It fails because it is duplicated. On many WordPress sites, multiple sources output structured data simultaneously: the SEO plugin, a second plugin, and on top of that the theme. The result is contradictory entries that confuse a search engine rather than orient it. Markup that contradicts itself is worse than no markup at all.
The rule is simple: one source for structured data per page. Which source is secondary; what matters is that the others stay silent. A quick look at the source code for JSON-LD blocks, or a run through a validation tool, surfaces duplicates within minutes.
What is genuinely worth doing in 2026
- Mark up Organization cleanly and connect it via
sameAsto reliable references, such as the commercial register, Wikidata, or a professional network profile. This anchors the entity. - Keep LocalBusiness complete: industry, region, opening hours, coordinates. This layer is massively underused across the web, which makes it an easy advantage to claim.
- Mark up each service page individually rather than applying a single catch-all markup across the whole site. A search engine should understand each service as its own entity.
- Validate before publishing, using Google’s Rich Results Test and the Schema.org validator. Both are free and surface errors immediately.
- Eliminate duplicates: one source for structured data, everything else switched off.
None of these steps will produce citations overnight. Together they ensure that a search engine can place your business without ambiguity, and that is the entry ticket without which everything else becomes irrelevant. Anyone promising triple-digit percentage jumps from a single piece of source code is selling hope, not results.
Whether your markup is clean shows up quickly
A look at the structured data on your site reveals fast whether the foundation is solid or whether duplicates and gaps have crept in. That is the unspectacular but necessary first step of any serious AI visibility work. Let’s talk.
Frequently asked questions
Does schema markup produce more citations in ChatGPT or Perplexity?
Not directly. Research from 2025 and 2026 shows that AI services ignore structured data when fetching a page directly, and that no additional citation effect is measurable on pages that already appear in results. Schema acts one step earlier, in how your business is classified, not at the moment the answer is generated.
What is schema for, then?
So that a search engine understands clearly who you are, what you offer, and where. That clean classification is the prerequisite for being considered a reliable source. Without it, a business remains a vague data cluster for search; with it, it becomes a clearly defined entity.
Which schema types matter for a small business?
Primarily three: Organization with links to reliable sources, LocalBusiness with complete location details, and a dedicated Service schema for each offering. LocalBusiness and Service in particular are rarely maintained properly across the web, which makes them an easy advantage to pick up.
Why does duplicate schema cause problems?
When multiple sources on a page output structured data simultaneously, say two plugins and the theme, the result is contradictory entries. A search engine is more confused than helped by this. Markup that contradicts itself is worse than no markup at all.
How do I check my schema?
With Google’s free Rich Results Test and the Schema.org validator. Both surface errors and missing required fields immediately. Scanning the source code for JSON-LD blocks also reveals whether a page is accidentally outputting the same markup twice.
Does schema improve my Google ranking?
Not directly. Structured data is not a ranking factor in the strict sense, something Google has clarified repeatedly. It can influence how results are displayed, and it is the foundation for being correctly recognized as an entity. The value lies in classification, not in rank.