What May's Citation Intelligence Data Tells Us

AI citations are now a source map, not just a visibility score. Zeover shows which sources AI engines cite, which platforms are gaining weight, and where your brand needs stronger evidence before the next monthly report. See your citation map.

May’s Citation Intelligence data points to a simple operating rule: social citations are sorting by evidence type. Reddit carried the broadest cross-model weight, YouTube stayed strongest for explanations and demonstrations, and LinkedIn remained the clearest professional-context layer. X/Twitter mattered for recency on x-ai/grok-4, but it didn’t behave like a universal citation surface.

The dataset covers Zeover’s verified social citation slice for March, April, and May 2026 through May 18. The directional read is useful because the model-platform split is already visible without forcing a single-platform conclusion.

TL;DR

• Reddit was the broadest May social citation surface in the verified slice. It led openai/gpt-5.4, x-ai/grok-4, and x-ai/grok-4.3, and it became a major perplexity/sonar-pro source after not appearing in the visible March and April rows.
• YouTube remained the explainer engine. perplexity/sonar-pro gave YouTube 44.1% of May social citation share, down from near-total concentration in March and April but still the largest Sonar surface.
• LinkedIn didn’t lead May, but it stayed important. It held 32.3% on openai/gpt-5.4, 14.5% on x-ai/grok-4.3, and 12.7% on x-ai/grok-4.
• X/Twitter was material only where the model had a strong real-time path. x-ai/grok-4 gave it 20.3% in May, while x-ai/grok-4.3 barely used it at 0.2%.

What The May Slice Shows

The May read isn’t a generic “social is up” story. Each platform appears to carry a different kind of evidence. Reddit supplied community validation, YouTube supplied demonstrations and explainers, LinkedIn supplied professional authority, and X/Twitter supplied recency where the engine routed that way.

The parenthetical value shows percentage-point movement from April to May. “New” means the platform or model did not appear in the visible April rows for this filtered citation slice.

The most useful May finding isn’t the ranking alone. It’s the spread. No single platform owned every model. Even Reddit, the broadest winner, shared serious weight with YouTube and LinkedIn. That means GEO work has to map query families to source types rather than treating social publishing as one bucket.

OpenAI’s ChatGPT Search launch note describes web answers with links to relevant sources and says the product uses third-party search providers plus partner content. Perplexity’s Sonar documentation describes Sonar as a web-grounded API. X’s Grok help page says Grok can decide whether to search public X posts and conduct real-time web search. Those docs explain possible retrieval paths, but Zeover’s data shows which paths actually surfaced in the measured citation mix.

What Changed Against April

The strongest month-over-month change was Sonar’s move away from YouTube concentration. In March and April, perplexity/sonar-pro was almost completely YouTube-led in the filtered social citation rows. In May, YouTube still led at 44.1%, but Reddit reached 39.9% and LinkedIn reached 9.7%. That’s diversification, not a YouTube collapse.

openai/gpt-5.4 also changed shape. LinkedIn led in April at 41.1%, while Reddit sat at 32.3%. In the May slice, Reddit moved to 47.7% and LinkedIn fell to 32.3%. The practical read is that OpenAI’s social citation mix can rotate between professional authority and community validation depending on query mix.

x-ai/grok-4 was steadier. Reddit stayed near one-third, YouTube stayed in the low twenties, and X/Twitter stayed material while slipping from 22.8% in April to 20.3% in May. x-ai/grok-4.3 entered the May data with a different profile: Reddit first, YouTube second, LinkedIn third, and X/Twitter nearly absent.

Platform Reads

Reddit: May made Reddit the broadest social evidence layer. It led three of the four visible model-month profiles and emerged strongly for Sonar. Reddit should be audited first for review, comparison, support, and “what do people think” query families.

YouTube: YouTube stayed the strongest procedural source. It mattered most for Sonar, but it also held 27.7% on x-ai/grok-4.3 and 22.5% on x-ai/grok-4. Teams should treat transcripts, titles, descriptions, chapters, and source links as AI-search assets, not channel housekeeping.

LinkedIn: LinkedIn remained the B2B and professional-context surface. It didn’t lead May, but it was too large to ignore on OpenAI and both Grok profiles. Named experts, company pages, role context, and category commentary all matter here.

X/Twitter: X/Twitter should be treated as a recency surface, not the whole Grok strategy. X’s Grok help page says Grok can search public X posts, yet May’s x-ai/grok-4.3 citation mix barely used X/Twitter. Access and citation weight are different things.

Facebook, Instagram, TikTok, Quora, and Pinterest: These were secondary in the verified May slice. Facebook mattered most on x-ai/grok-4.3 at 7.7%. Instagram and TikTok stayed low, with TikTok’s largest pocket at 8.5% on openai/gpt-5.4. These platforms can support consumer proof, local context, and creator-led discovery, but they shouldn’t be the first citation bet for most B2B GEO work.

What It Means For GEO, AEO, And AIEO

For GEO, May reinforces source-type mapping. The question is no longer whether a brand “does social.” The question is whether the brand has the right evidence on the platform the engine already uses for that query family.

For AEO, the May data points back to extractability. Google Search Central’s structured data introduction frames structured data as a way to help Google understand page content. AI citations follow the same practical discipline: titles, headings, captions, transcripts, and page structure make evidence easier to retrieve and quote.

For AIEO, the lesson is measurement. Agents and AI search systems do not only need a brand mention. They need a reason to trust the answer. May’s platform split shows why presentation has to match source type: a Reddit answer, a YouTube transcript, and a LinkedIn post aren’t interchangeable evidence.

May Focus

The highest-value May action is a source audit by query family.

Start with the prompts where Reddit appeared. Check whether the cited threads describe the category accurately, whether outdated claims are still circulating, and whether the brand’s own site has a better source of record.

Then audit YouTube for explainer queries. A video without accurate captions, chapters, and a useful description is weak evidence even if the channel performs well. The video page should answer the query in text as well as in speech.

Finally, audit LinkedIn for category and people context. If AI engines are using LinkedIn as professional evidence, the brand needs consistent company language, named operator posts, and clear role-linked expertise.

Zeover helps teams connect monthly citation patterns to query-level content priorities: community validation, transcript-ready explainers, and professional authority.