Last updated July 2026
The claim travels fast: “Add FAQ schema and you will get into the AI answer box.” It sounds clean. It is the kind of advice that spreads because it is actionable and easy to measure. The problem is that the two most credible data points on this question land on opposite sides, and the side that appears to support schema actually argues against it on closer reading.
This piece lays out both sides cleanly, tells you what the evidence actually supports, and gives you the real lever to pull.
The evidence for schema: a 1,500% lift that nobody can replicate
In late 2025, Otterly.AI rolled out five schema types across their own website. Over the following three months, they tracked Google AI Overviews appearances across 319 prompts and seven AI platforms.
The number they reported was striking: a +1,500% increase in Google AI Overviews appearances during the period following the rollout.
That sounds like a clean proof case. Schema goes up, AI visibility goes up by 15x. Done.
Except Otterly.AI’s own researchers reached the opposite conclusion. When they looked at what happened to competitors who made no schema changes during the same window, those competitors showed equivalent movement. A broader platform-level algorithmic shift was already underway. Schema did not cause the spike.
Two more details matter here:
- The 1,500% figure covers Google AI Overviews appearances specifically. On ChatGPT, Gemini, and Copilot, citations dropped over the same period.
- The experiment was run on Otterly.AI’s own site. Self-experiments by vendors have obvious limits: the baseline, the content quality, and the domain authority are all specific to one SaaS brand.
Otterly.AI did the right thing by publishing the caveat. That caveat is the finding. The lift was real. The cause was not schema.
The evidence against schema: 1,885 pages, no movement
Ahrefs ran a controlled study published in May 2026. They tracked 1,885 pages that added JSON-LD schema between August 2025 and March 2026, matched against 4,000 control pages, and measured AI citations across Google AI Overviews, Google AI Mode, and ChatGPT.
The results by platform:
| Platform | Change after schema addition | Statistically meaningful? |
|---|---|---|
| Google AI Overviews | -4.6% | No (small decline) |
| Google AI Mode | +2.4% | No (noise-level) |
| ChatGPT | +2.2% | No (noise-level) |
Every number is within the noise band. Adding schema produced no meaningful uplift in AI citations on any of the three platforms studied.
One important scope note: every page in the Ahrefs dataset already had 100 or more AI Overview citations before the treatment. These were not invisible pages getting their first schema. They were already performing pages testing whether more schema would lift them further. That design choice makes the finding conservative in one direction. It is harder to lift pages that are already being cited than to lift pages with zero visibility. Whether schema would move the needle for completely uncited pages remains an open question.
What the study does establish clearly: for pages already receiving AI citations, adding schema does not increase how often you get cited.
What actually drives AI citation selection
If schema alone does not move the needle, what does? The best available evidence from independent and partially-verified sources points to four factors:
1. Organic search rank position. The Fischman SSRN study (“Does Schema Markup Predict AI Citation?”) found that Google organic rank position was the dominant predictor of AI citation likelihood, not schema presence. Pages that rank well organically are selected more often as AI sources. This finding is consistent with Ahrefs’ separate research showing that AI Overviews draw heavily from pages already performing in traditional search.
2. Content placement in the first third of the page. According to Kevin Indig’s 2026 analysis of 1.2 million ChatGPT responses, 44.2% of citations were drawn from the first 30% of a page’s content. Putting your direct answer early matters more than any markup layer.
3. Third-party citation and authority signals. Studies consistently find that the large majority of AI citations come from third-party sources rather than brand-owned websites. Earning mentions and links from high-authority domains builds the authority signal engines use when selecting sources.
4. Content structure and directness. According to an AirOps study of 12,000 URLs, 68.7% of ChatGPT-cited pages used a sequential heading structure (H1 to H2 to H3), compared to 23.9% of Google’s top-ranked pages. Nearly four out of five cited pages included at least one structured list. These are formatting signals that help engines extract your answer, and they work independently of JSON-LD schema in the page source.
Schema still belongs in your stack. It helps engines parse content correctly, it supports rich results in traditional SERPs, and it is a prerequisite for some AEO tool workflows. The mistake is treating it as the primary lever when the heavier work is content quality, authority, and structure.
How to audit your position without guessing
The practical question is not “should I add schema?” but “how do I know if any of this is working?” That requires tracking actual AI citation rates before and after any change, across multiple engines, for a stable set of prompts.
A handful of tools support this kind of controlled measurement:
Otterly.AI (/tools/otterly-ai) is worth noting here specifically because their own schema experiment is the dataset that complicates the “schema works” claim. Their GEO Audit Engine scores URLs across more than 20 citation-readiness factors and tracks prompt-level citation rates across six platforms. Starting at $29/mo on the Lite plan, it is the most affordable dedicated AEO monitoring option with a public track record on schema measurement.
Surfer integrates content scoring and AI visibility tracking in a single editor. Its Content Score system includes structured data checks alongside NLP-level content signals, so you can see schema status in context rather than as a standalone metric. Useful for teams that want to write and audit in one place.
Writesonic (/tools/writesonic) includes schema generation as part of its content creation workflow, which is useful for teams that need to produce FAQ markup at scale without manual JSON-LD authoring. It does not provide deep citation monitoring, so pair it with a tracking tool for measurement.
HubSpot’s AEO grader gives a quick page-level audit that flags schema gaps alongside other AEO readiness signals. It is free and useful as a spot-check before publishing a page.
Temso (/tools/temso) covers the full AEO loop (monitoring citation rates across eight engines, generating FAQ/schema content, and tracking changes over time) at a flat $89/mo. For teams that want schema generation and citation measurement in one tool rather than stitching together separate tools, it is one credible option at the affordable end of the category. It is not a schema specialist; it is a full-loop AEO platform where schema is one part of the workflow.
The full list of tools is at /rankings/aeo-tools.
Whatever tool you use, the measurement discipline matters more than the schema itself. Run your target prompts before a schema change, track them for 60 to 90 days after, and compare to pages where you changed nothing. That is the only way to know if schema moved your specific numbers.
The verdict on schema and AI answer boxes
Schema markup is necessary but not sufficient for AI answer box inclusion. That phrase does real work here:
- Necessary: engines need to parse your content to cite it. Schema helps with that. A page without any structured markup is harder for an engine to classify, and some AEO workflows require it.
- Not sufficient: adding schema to a page that lacks organic authority, direct answers, and third-party mentions will not get you into AI Overviews or ChatGPT results. The Ahrefs study of 1,885 pages makes that clear. The Otterly.AI experiment confirms it from the other direction.
The myth worth busting is not “schema matters” but “schema alone is the unlock.” It is not. It is one item in a longer checklist, and most of the items above it on that checklist are harder to do: write better answers, earn more third-party mentions, build organic authority, and put your direct response in the first paragraph.
Fix those first. Then add the schema.
Want to know which AI answer boxes you are currently losing? The AEO tools ranking covers the tools built to answer that question, with pricing, engine coverage, and what each one actually does.