IndexWire | Week of May 14, 2026
Ahrefs tracked 1,885 pages that added JSON-LD schema markup between August 2025 and March 2026. They matched those pages against 4,000 controls with similar pre-treatment citation baselines, ran a difference-in-differences analysis, and validated it through four statistical approaches. The result: no platform showed a meaningful citation increase. Google AI Overviews moved -4.6% (statistically significant, directionally negative); Google AI Mode came in at +2.4% (within noise); ChatGPT at +2.2% (within noise). (Ahrefs, May 11, 2026)
That is a clean causal test on a question the field has mostly answered with correlation. And the correlation is real: pages cited by AI Overviews carry JSON-LD roughly three times more often than uncited pages. But Ahrefs controlled for that. Pages already in citation rotation added schema and nothing moved. The schema was not doing the citing work.
There is a scope caveat the study is transparent about: all 1,885 treatment pages already had 100 or more AI Overview citations before the test began. The study does not tell you whether schema helps previously uncited pages get noticed during crawl and indexing. That is a different and still-open question. If your team is adding schema to pages already appearing in AI Overviews to extract more citations, the data says redirect that effort. If you are adding schema to brand-new pages to aid initial discoverability, this study does not address that use case.
A related searchVIU experiment adds a useful constraint: five AI systems (ChatGPT, Claude, Perplexity, Gemini, Google AI Mode) fetched pages in real time and all extracted only visible HTML. JSON-LD, hidden Microdata, and hidden RDFa were not processed during direct retrieval. (Search Engine Journal, May 2026) Schema may still matter at crawl time via Googlebot; that is a different mechanism than citation selection at answer-generation time, and conflating the two is how answer engine optimization (AEO) programs end up chasing the wrong signal.
Google Updated Five Things About Links in AI Overviews. One of Them Is Actually Structural.
On May 6, Google announced five changes to link behavior inside AI Overviews and AI Mode: subscription highlighting (preferential display for publications users already subscribe to), inline citations (positioned adjacent to relevant text rather than grouped at the end), creator attribution for social and forum sources, hover previews on desktop, and exploration suggestions appended to responses. (Search Engine Land, May 6, 2026)
Four of the five are interface adjustments. The fifth is structural: subscription highlighting creates a two-tier link ecosystem inside AI Overviews. Publishers whose readers are logged in and subscribed get preferential display. Publishers without that relationship compete on the same surface they always have. Google reported users were "significantly more likely to click links labeled as their subscriptions" but disclosed no percentage, and no third-party CTR data has been published. Treat this as directional until Search Console data surfaces in your own programs.
The zero-click problem is unchanged. Agarwal and Sen's April 2026 SSRN study found a 38% organic click reduction from AI Overviews with no user satisfaction gain. (SSRN, April 3, 2026) Five interface updates do not move that number.
A 14% Overlap Figure Worth Tracking (Not Yet Citing)
Multiple 2026 analyses report that Google AI Mode and AI Overviews cited the same URLs approximately 13-14% of the time, with pages in AI Overviews having roughly a 15% probability of also appearing in AI Mode for the same query. (StackMatix, 2026) Flag this as directional only: the specific figure appeared across multiple third-party summaries, but a single primary-methodology source could not be verified. Treat the pattern as a hypothesis to pressure-test against your own data, not a benchmark to cite in a client report.
The underlying structural point is more solid. AI Mode runs up to 16 simultaneous sub-queries per search, returns no organic results, and selects sources independently of AI Overviews. If AI Mode scales to default search, a site is either cited or absent. There is no fallback organic position. Current AEO tooling was built for AI Overview tracking; most rank trackers do not report AI Mode citation share separately. If you resolved to run separate rank and citation tracking after reading about the rank-citation decoupling pattern, a third program may be needed sooner than the expansion timeline suggested.
AI-assisted content production: this newsletter was researched and drafted with AI assistance. All factual claims cite named primary sources; methodology limitations and data caveats are disclosed in text.