Adding JSON-LD did not produce a clear increase in AI citations in Ahrefs’ 2026 study—but that is not evidence that structured data is useless or that it hurts visibility. The study followed pages that were already heavily cited, and Google continues to document structured data as useful for helping Search understand pages and enabling eligible rich results. Keep accurate markup when it serves a concrete Search or site need; treat an AI-citation boost as a hypothesis to test, not a promise.
What Ahrefs measured—and what it found
Ahrefs’ May 11, 2026 analysis, “We Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved.”, identified 1,885 pages that added JSON-LD between August 2025 and March 2026 and matched them with 4,000 control pages. It compared AI citations in the 30 days before and after each page’s markup change, using a matched difference-in-differences analysis as its preferred method.
As an Amazon Associate I earn from qualifying purchases.
| AI-search product | Reported change relative to controls | Ahrefs’ qualification |
|---|---|---|
| Google AI Overviews | −4.6% | A small decline reported as statistically significant, but not definitively attributable to schema. |
| Google AI Mode | +2.4% | Statistically indistinguishable from zero. |
| ChatGPT | +2.2% | Statistically indistinguishable from zero. |
Ahrefs’ overall interpretation was no clear positive or negative effect for the tested pages. Its content marketer Louise Linehan summarized the result: “Adding schema produced no major uplift in citations on any platform.” The figures do not support saying that JSON-LD caused AI Overviews citations to fall: Ahrefs said it could not definitively pin that decline on schema.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Why this study does not settle the question for every site
The sample and method narrow what the result can tell publishers. In particular, every page already had at least 100 AI Overview citations in February 2025, before adding schema. The study therefore does not directly answer whether structured data helps pages that have not yet earned AI citations.
#1 Best Overall
- Short observation window: The comparison covered 30 days before and after the markup change; longer-term effects were not tested.
- Schema types were pooled: The analysis included types such as Article, FAQ, Product, HowTo, and Organization, but does not show whether any individual type behaved differently.
- Other changes may have occurred: Ahrefs noted that pages could change in other ways alongside JSON-LD, which the analysis could not fully separate.
- Implementation scope was specific: The report examined JSON-LD in page HTML and did not test JavaScript-injected schema in the same way.
So the finding is best read as a narrow test of adding JSON-LD to already-cited pages—not a universal verdict on structured data, all schema types, or every implementation.
JSON-LD still has a documented Search role
Google’s structured-data guidance, last updated December 10, 2025, says structured data can provide explicit clues about a page’s meaning and can make pages eligible for rich results. Google recommends using the format that is easiest to implement and maintain, “in most cases” JSON-LD; it also supports Microdata and RDFa when validly implemented.
Rank #2
That is a separate rationale from expecting more citations in AI-generated answers. Google’s documentation does not promise that markup will produce a rich result or an AI citation. It also says the markup should describe information visible to users and warns against creating empty pages just to host structured data.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minuteGoogle presents case studies reporting a 25% higher click-through rate for Rotten Tomatoes pages enhanced with structured data and an 82% higher click-through rate for Nestlé pages showing rich results. These are Google-reported examples about structured-data and rich-result outcomes, not controlled estimates of AI-citation lift or guarantees for other sites.
Rank #3
What a separate retrieval experiment adds
A 2026 arXiv preprint by Andrea Volpini, Elie Raad, Beatrice Gamba, and David Riccitelli tested page representations in a purpose-built retrieval-augmented generation (RAG) setup. The experiment compared plain HTML, HTML with JSON-LD, enhanced entity pages, and agentic retrieval across four domains. JSON-LD alone produced only a marginal accuracy improvement; enhanced entity pages, which also included natural-language summaries and navigable links between entities, showed reported accuracy gains of 29.6% in standard RAG and 29.8% in an agentic pipeline.
The preprint describes a specific Vertex AI and Google ADK pipeline, not public AI-search citations. Its small four-domain dataset and a possible circularity—ground-truth answers were derived from the same knowledge-graph data used to build some page variants—also limit how broadly to apply its results. It suggests that readable information and exposed relationships can matter in that experimental retrieval design; it does not establish a universal recipe or prove that a particular file or markup format improves citations.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Should you keep or remove your site’s JSON-LD?
Make the decision page by page and type by type. A useful starting point is to weigh whether the markup serves an eligible Search feature, accurately represents visible content, and is worth the effort of maintaining without errors.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
- Keep it when it accurately describes visible page content and supports a concrete Search feature or other defined site requirement.
- Correct or remove it when it is inaccurate, duplicated, difficult to maintain, or has no clear use for the page.
- Do not remove all markup solely because of this study. Its results concern already-cited pages and do not answer every site’s situation.
- Do not add it solely on a promise of AI citations. Ahrefs found no clear uplift in its tested population.
Google’s guidance says structured data should be valid and match visible content. If you keep it, review the markup as pages change rather than treating installation as a one-time task.
How to test an AI-citation claim on your own site
A small controlled comparison is more informative than changing markup across the whole site and guessing what caused a later change. Ahrefs suggests 5–10 test pages and 5–10 control pages for a small-site experiment. Google recommends monitoring structured-data results over a few months and checking that it found valid markup; Ahrefs’ study used 30-day before-and-after windows. These are different measurement suggestions, not a single guaranteed test duration.
- Choose comparable pages. Select test and control pages with similar content and existing visibility. Record the AI-citation baseline for the products you intend to track.
- Change only the test group. Add the relevant JSON-LD to test pages, leaving controls unchanged. Avoid simultaneous content, template, and technical changes where practical.
- Validate the markup. Check that the structured data is valid and reflects visible page content, and monitor whether Google detects it.
- Measure over time. Compare citation changes for test pages with changes for controls, rather than treating a raw increase or decrease as proof of impact.
- Record the limits. Note the pages, markup types, AI products, dates, and other changes. The result applies to that setup and period; it may not transfer to other pages or platforms.
If the test does not show an improvement, that is evidence about your tested pages and measurement window—not proof that structured data can never help. If citations move, the control group and limited concurrent changes help you judge whether the markup is a plausible explanation.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
