A new 5W research report synthesizes nine independent datasets on AI citation behavior — and the picture it produces should change how the entire industry budgets earned media.
For decades, public relations operated on a stable hierarchy. Tier 1 media at the top, trade press in the middle, blogs at the bottom. Domain authority, editor relationships, and the rough rule that a Wall Street Journal hit beat a trade pub hit, every time. That hierarchy is no longer how influence works.
When a buyer asks ChatGPT, Claude, Perplexity, Gemini, or Google AI Overviews about a brand, a category, or an executive, those systems do not respect the PR tier structure. They pull from a fragmented, dynamic, and structurally different set of sources. Wikipedia and Reddit dominate. LinkedIn and YouTube are climbing fast. Review platforms drive recommendations. And traditional Tier 1 media — the targets PR has spent a century optimizing for — is dramatically underrepresented.
That is the picture that emerges from the Q1 2026 edition of The 5W Citation Source Audit, a new quarterly research report from 5W, the AI Communications Firm. The report synthesizes findings from nine independently published research datasets — Similarweb, SEMrush, Profound, Peec AI, SE Ranking, Goodie, Ahrefs, Evertune, and Passionfruit — covering hundreds of millions of citations and prompts. The full report is free at www.5wpr.com/research/citation-source-audit-q1-2026.
Five Findings the Industry Has Not Absorbed
The headline number: Wikipedia and Reddit together account for more than 25% of all ChatGPT citations in the U.S., per Similarweb's January–February 2026 dataset. Wikipedia alone is at 13.15%. Reddit is at 11.97%. No traditional media category comes close. Reuters, the highest-cited mainstream news outlet on ChatGPT, sits at 2.27%. Forbes — the only U.S. business publication in the top 20 — is at 1.38%. The Wall Street Journal, The New York Times, Bloomberg, and Financial Times do not appear in the top 20 at all.
The second finding follows from the first. The traditional PR tier system is misaligned with how AI engines actually source. Reuters outranks Forbes. Forbes outranks most Tier 1 media. The marquee outlets that PR teams have built decades of relationships with are doing far less work in the AI citation layer than tier rankings would predict. There are structural reasons — paywalls limit body-text extraction, licensing disputes have reduced indexing, and long narrative features extract less cleanly than tight trade or contributor pieces — but the result is that an earned-media strategy concentrated only in Tier 1 mainstream is structurally underweighted in AI.
Third: AI citations are long-tail, not winner-take-all. Outside Wikipedia and Reddit, no domain exceeds 3% of ChatGPT citations. The remaining citation share spreads across thousands of long-tail sources. This is a fundamentally different distribution than traditional SEO, where the top 10 results capture roughly two-thirds of clicks. Distribution beats concentration.
Fourth: the platforms are volatile. The biggest single shift of 2025 was the September collapse, when ChatGPT's citation share for Reddit dropped from approximately 60% to roughly 10% of prompt responses in a two-week window. Wikipedia followed a similar pattern. Forbes doubled in the same period. The platforms tune retrieval behavior aggressively. Brands measuring annually are reporting against citation patterns that no longer exist.
Fifth: the engines source meaningfully differently from each other. ChatGPT is the most Wikipedia-heavy. Google AI Mode favors Google-owned properties. Perplexity skews toward research-credible sources like NIH and G2. Gemini integrates Google search results directly. There is no single "AI SEO" strategy.
The Structural Insight
Underneath the data, the report argues that AI engines work fundamentally differently from search engines. They do not rank authority. They assemble answers. They favor extractable content over narrative, prioritize consensus across many sources over a single authoritative one, and reward repetition over prestige.
That is the single biggest shift from traditional PR. Authority is no longer controlled by editors. It is distributed across platforms. The brands that win in AI visibility are not the ones with the most prestigious clip book — they are the ones whose name appears, consistently, in the structured surfaces the models actually pull from.
What Practitioners Should Do
The report's operator playbook lays out four levers that must be executed against simultaneously: control your ground truth (Wikipedia, owned site, schema markup), build distributed authority (Reddit, LinkedIn, YouTube, review platforms), create extractable content (case studies with numbers, FAQs, deep single-topic reference pages), and increase repetition across sources. None is optional. Each one feeds the others.
5W is publishing the audit quarterly. The Q2 2026 edition, due in July, will layer the firm's own primary research on the synthesis baseline — 1,500 fixed prompts run across the four major LLMs in a single week, with the full methodology released for public replication. The Q1 report and the future quarterly editions are free at the URL above.
