Schema markup is still worth doing as technical SEO hygiene, but treating it as the main AI citation lever is a poor use of incremental marketing effort. The better move is to keep schema clean, then shift the next hour of work toward clearer answers, stronger authorship, external mentions, and sources AI systems already trust. Ahrefs recently tracked 1,885 web pages that added JSON-LD schema between August 2025 and March 2026. They matched those pages against 4,000 control pages and measured citation changes across Google AI Overviews, Google AI Mode, and ChatGPT. The result was not a meaningful citation lift. Google AI Overviews showed a 4.6% decline relative to matched controls, while AI Mode showed a 2.4% increase and ChatGPT showed a 2.2% increase. Ahrefs described the AI Mode and ChatGPT effects as statistically indistinguishable from zero. Their overall conclusion was careful, which is the right tone for this kind of data: schema could have done a tiny bit of good or nothing at all. That is very different from the way schema is being sold in some AI visibility conversations. Schema has become attractive because it feels controllable. The CMO playbook for the citation era makes the underlying budget question concrete: if AI Overviews absorb informational queries, the marginal hour of work has to move toward substance, not markup. A team can audit it, add it, validate it, and show a before-and-after implementation report. For marketing leaders under pressure to respond to AI search, that kind of work is tempting because it looks concrete. The danger is confusing a clean technical task with a strategy for being cited, mentioned, and trusted.
Schema helps machines read. It does not make the underlying source worth citing
Structured data can help search engines understand page entities, article metadata, authorship signals, product information, FAQ structure, reviews, and other elements. That is useful. It reduces ambiguity. It can support eligibility for rich results and improve the technical clarity of the site. AI citation is a different question. A model deciding what to cite, summarize, or mention is not only asking whether the page is machine-readable. It is also evaluating whether the page is relevant, clear, authoritative, fresh, consistent with other sources, and useful for the query. A weak page does not become worth citing because the schema validates. A generic article with Article schema is still generic, and a service page with clean Organization schema still needs a strong point of view, proof, and external validation. The markup may make the package easier to parse, but it does not create the substance. That is why the Ahrefs study matters. It gives leaders permission to stop overinvesting in schema as the answer to AI visibility. Keep it in the technical backlog, but do not let it absorb the budget that should be going into content quality, source building, and distribution.
The reallocation map
| If the plan says… | Keep as hygiene | Reallocate incremental effort to… | Business reason |
|---|---|---|---|
| Add Article schema to every blog post for AI citations | Yes, if missing or broken | Rewrite priority pages around one specific buyer question | AI systems need a clean, quotable answer, not just metadata |
| Add FAQ schema across generic guides | Use only where FAQs are real and useful | Build single-question explainers with direct answers and source-backed claims | Long, vague guides are weaker than precise answer assets |
| Add Person and Organization schema | Yes, maintain accurate authorship and entity data | Strengthen author bios, bylines, credentials, and topic ownership on the page | Trust is partly technical, but it is also visible to humans and models |
| Add Product or Service schema | Yes, where appropriate | Improve service pages with scope, fit, proof, process, and qualification details | Commercial pages still need buyer confidence and conversion |
| Add Review schema | Only when compliant and real | Build third-party review presence and customer proof on credible external sources | AI answers often lean on external validation |
| Add schema to old content at scale | Prioritize broken or important pages | Consolidate, split, or retire outdated articles | Freshness and clarity matter more than markup on weak pages |
| Treat schema implementation as the AI visibility project | No | Create a monthly AI Visibility Review with owners and decisions | Visibility work needs operating ownership, not a one-time technical ticket |
Technical SEO still matters. It just cannot carry the marketing judgment that belongs in the content, proof, and distribution plan.
What actually deserves the next hour
Start with clearer single-question content. AI Overviews appear heavily on informational queries, especially questions, definitions, and explanations. A B2B company does not need another 5,000-word guide that tries to cover a whole category. It needs answer assets that respond to the exact questions buyers ask when they are trying to understand a problem. For Structured Rebellion’s world, those questions might look like:
- Why does AI adoption increase marketing activity without improving pipeline clarity?
- What should a monthly marketing operating review actually decide?
- How should a B2B CMO separate content that needs clicks from content that needs citations?
- What makes Service-as-Software different from an agency or SaaS product?
Each page should answer the question quickly, explain the business implication, and support the claim with a concrete example or source. That kind of content is easier for a buyer to trust and easier for an AI system to summarize accurately. Then strengthen authorship. Thin authorship is one of the easiest ways for serious content to look interchangeable. A page about marketing operating models should not read like it came from a generic content desk. It should show the operator behind it: experience, domain, examples, and judgment. Schema can label the author. It cannot replace a visible point of view. Next, build earned mentions. Ahrefs’ ChatGPT visibility analysis points to the importance of mentions across the web, including mentions without links. The logic is straightforward: models learn from repeated signals. If authoritative sources keep associating a brand with a category, use case, or solution, that repetition can influence how the brand appears in AI answers. YouTube also deserves attention. Ahrefs found that YouTube mentions had the strongest correlation with ChatGPT visibility in their study, with Tim Soulo’s summary citing a 0.737 correlation. That should be used directionally, not as a universal law. Still, the business implication is clear enough: spoken mentions, transcripts, reviews, interviews, explainers, and category discussions on YouTube are part of the AI visibility surface now. Finally, work backward from cited domains. If ChatGPT, AI Overviews, or AI Mode repeatedly cite certain review sites, industry publications, comparison pages, communities, or educational resources in your category, those sources become part of the marketing operating model. They are not “PR” in a separate room. They are part of how buyers and AI systems validate the market.
A practical sequence
The work can start with a simple sequence. First, fix broken technical basics: schema validity, indexability, canonicals, author pages, page titles, internal links, and crawl problems. A messy site makes everything harder. The funnel-signal review is a related move on the measurement side: clean inputs let the team trust what the data is showing. Second, choose the 20 to 30 pages where AI citation or mention would actually matter. This usually includes definitions, category POVs, research pages, service explainers, comparison pages, and high-intent educational assets. Third, review those pages for answer clarity. Can a paragraph be cited without losing meaning? Is the claim specific? Is the page current? Does the author have visible credibility? Is there a named framework, example, or artifact? Fourth, map external sources. Which domains are being cited for the category? Which ones mention competitors? Which ones publish lists, reviews, videos, or educational content that buyers and models already trust? Fifth, assign owners. Technical SEO owns the hygiene. Content owns the answer quality. Leadership owns the point of view. Demand gen or partnerships may own external mentions. Sales should contribute the questions buyers actually ask. Schema still belongs in the stack. It just should not be allowed to become the strategy because it is the easiest part to ticket.
Next read: The AI Visibility Review: The Monthly Operating Rhythm B2B Teams Need — the operating cadence that prevents schema from absorbing the AI visibility budget. To pressure-test where the substance actually needs work, see our methodology.
— Fernando González Aguirre, Founder, Structured Rebellion
