No published measurement shows that adding schema.org markup raises citations in AI answers. Google says structured data is not required for its generative AI search and that no special markup is needed. Ahrefs tracked 1,885 pages that added JSON-LD and found no meaningful uplift in AI Mode or ChatGPT, and a small decline in AI Overviews. Microsoft recommends schema in general terms but gives no data. Keep accurate markup for the features Google documents, and do not expect it to buy AI visibility.
The evidence, by source and strength
We ranked each item by what kind of evidence it is. Official statements tell you how a company says its system works. Measured studies tell you what happened in a sample. A vendor claim without a sample size and method is a weak signal either way.
| Source | Type | What it says | Limits |
|---|---|---|---|
| Google, AI optimization guide | Official statement | Structured data "isn't required for generative AI search" and no special schema.org markup is needed. The guide lists overfocusing on structured data among things to ignore, while still recommending it for rich results eligibility. | Covers Google Search only. Says nothing about ChatGPT or Copilot. |
| Microsoft Advertising blog, Oct 8, 2025 (Bing product manager) | Official statement | Describes schema as code that helps search engines and AI systems understand content, and says it can label content as a product, review, FAQ or event. | No sample, no measurement, no description of how Copilot uses it. |
| Bing Webmaster Blog, Feb 2026 | Official statement | Names clear headings, tables and FAQ sections as helping surface key information. The text we read does not mention schema. | Announcement of a report, not a ranking guide. |
| Ahrefs, May 11, 2026 | Measured study | 1,885 pages that added JSON-LD between August 2025 and March 2026, each matched to controls that never added it. Citations counted 30 days before and after. | Only pages with 100+ AI Overview citations; schema types pooled; 30-day window; schema edits often came with other edits. |
| Ahrefs, same post, earlier analysis | Correlation | Across 6 million URLs, pages cited in AI Overviews were almost three times as likely to carry JSON-LD. | The authors call it correlational. Sites that invest in schema often invest in everything else. |
| searchVIU, Oct 30, 2025 | Small experiment | Eight fictitious products, price shown in one place each. No AI system named a price that appeared only in JSON-LD. | One page, one point in time, not repeated. Did not test AI Overviews or Copilot. |
| Vendor guides with lift figures | Vendor claim | One guide we read gives several different lift numbers for the same idea, such as 3x and 36%. | No sample size, method or link to the underlying data. |
What Microsoft's guidance does and does not support
Microsoft is the one company that points toward schema, and it does so in plain terms: schema helps systems understand a page. That is a statement of intent, not a measured result. The Bing guidance we read lists schema next to headings, modular layouts and concise answers, so it is one habit among several. For the structures Bing says help most, see our Bing AI Performance guide.
Reading the evidence honestly
- No effect found is not proof of no effect. Ahrefs tested pages that were already heavily cited. A new page with nothing else going for it was not in the sample, and a slower effect than 30 days could be missed.
- Markup alone does not carry facts. The searchVIU result points the same way as Google's statement: AI systems read the visible page. A price or opening time that exists only in JSON-LD may never be used.
- Different systems, different answers. Gemini, ChatGPT and Claude behaved differently in the same test, so one system's result does not transfer to another.
What to do with this
| Your situation | Reasonable action |
|---|---|
| Hoping markup will win AI citations | Spend the time on the visible answer, crawl access and indexing first. Use the quotability checker to see what a page offers an answer engine. |
| Important facts live only in markup | Show them in the visible HTML as well, and keep markup identical to it. |
| A Google feature exists for the page | Add markup for that feature. See which rich results still show and the schema type picker. |
| Existing FAQ markup | Read what changed for FAQ markup. |
Check what a page currently outputs with the schema checker. For the terms themselves, see structured data versus schema markup.
Questions people ask
Should I remove schema markup if it does not help AI search?
No. Google still recommends structured data for rich result eligibility on supported features. Removing accurate markup for AI reasons has no evidence behind it.
Why do some SEO sites claim a big citation lift from schema?
Many compare sites that already have schema with sites that do not. Sites that add markup tend to be better maintained overall, so the comparison mixes cause and coincidence. Ahrefs' before-and-after design was built to separate the two.
Does Microsoft Copilot use schema?
Microsoft says schema helps search engines and AI systems understand content, but publishes no measurement of its effect on Copilot citations. Treat it as a stated intent.
What would change this verdict?
A platform documenting that it reads markup for answers, or a controlled test covering pages that start with no AI visibility, a longer window and each schema type separately.
Related
Sources
- Google Search Central: AI optimization guide, accessed 2026-11-12.
- Ahrefs: We tracked 1,885 pages adding schema (May 11, 2026), accessed 2026-11-12.
- searchVIU: Schema markup and AI put to the test (Oct 30, 2025), accessed 2026-11-12.
- Microsoft Advertising: Optimizing your content for inclusion in AI search answers (Oct 8, 2025), accessed 2026-11-12.
- Bing Webmaster Blog: AI Performance in Bing Webmaster Tools (Feb 10, 2026), accessed 2026-11-12.
- Search Atlas: Schema for AEO (vendor guide, claims without method), accessed 2026-11-12.