In 2026, that dominance is no longer assumed. A growing share of sports questions are now asked to ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews instead of to a search bar or an ESPN app. And inside those AI engines, ESPN is not the dominant citation. Wikipedia is.
This is the pilot framework for a study Everything-PR is building: the EPR Sports Citation Index. The full data release, with 500-query coverage across all five major AI engines, is targeted for late 2026. This piece establishes the thesis, the methodology, the directional findings drawn from public-source AI citation research, and what it means for ESPN, for sports media, and for any brand that thinks an established media franchise translates automatically into AI-engine citation share.
The thesis
Three observations, visible across the AI engines as of mid-2026.
Wikipedia owns the record. When an AI engine answers "who won Super Bowl LVIII" or "what year did Michael Jordan retire" or "who is the all-time NFL passing yards leader," the citation is overwhelmingly Wikipedia. ESPN's box scores and historical data pages exist. The engines do not surface them. Wikipedia is encyclopedic, structured, persistent, and citation-rich — exactly the source profile retrieval-augmented generation systems are optimized to surface.
ESPN owns the news. When the question is "who broke the Tyreek Hill trade" or "what time does Monday Night Football start" or "who is leading the NFL in passing yards this season" — original-source breaking news and live current data — ESPN's citations dominate. Adam Schefter's X account, ESPN.com beat reporting, ESPN's published live game data. The structural advantage is the original-source nature of the content. The engines retrieve the first-cited public record. (See Adam Schefter Built the Sports Insider Economy for the mechanism.)
Reddit owns the takes. When the question is "is X overrated" or "who is the GOAT" or "who should I draft in the third round" or "what do fans think of Y trade" — opinion, fan-sentiment, community-knowledge queries — Reddit citations dominate. Particularly on Perplexity, where Reddit accounts for approximately 46.7 percent of top citations across the platform's entire query mix per ZipTie's published analysis. ESPN's First Take and Get Up segments are not retrieved by the AI engines as opinion sources. The fan thread is.
The public-source data foundation
The EPR Sports Citation Index pilot rests on five independently published AI-engine citation studies, all conducted between mid-2024 and early 2026. The data is general-purpose (not sports-specific), but the patterns are robust enough to project onto sports queries with high confidence.
Profound, August 2024 – June 2025. Analyzed approximately 680 million AI citations across ChatGPT, Google AI Overviews, and Perplexity. Finding: Wikipedia accounts for 47.9 percent of ChatGPT's top 10 most-cited sources, demonstrating "the platform's preference for encyclopedic, factual content." Reddit emerges as the leading source for both Google AI Overviews (2.2 percent of total citations) and Perplexity (6.6 percent of total citations).
Semrush, October–November 2025. Analyzed weekly citations across more than 230,000 prompts over 13 weeks on three LLMs. Finding: despite a "dramatic drop-off in ChatGPT citations" in mid-September 2025, Reddit and Wikipedia remained ChatGPT's two most-cited domains. Google AI Mode cited Wikipedia in only approximately 2 percent of responses.
Peec AI, early 2026. Analyzed 30 million sources across ChatGPT, Google AI Mode, Gemini, Perplexity, and Google AI Overviews. Finding: Reddit was either the #1 or #2 most-cited source on every LLM tested. Wikipedia "shows up strongly for ChatGPT and Perplexity but not for Google's platforms at all."
ZipTie.dev, March 2026. Comparative analysis of citation behavior across major AI platforms. Finding: ChatGPT cites Wikipedia in 47.9 percent of its top citations; Perplexity cites Reddit in 46.7 percent; Google AI Overviews favors YouTube at 23.3 percent; Claude favors blogs at 43.8 percent. Only 11 percent of domains are cited by both ChatGPT and Perplexity for the same query.
Frase.io, March 2026. Finding: LLMs cite 2–7 domains per response on average — far fewer than Google's traditional ten blue links. Getting into that retrieved set requires deliberate optimization.
The Sports Citation Share — Pilot Findings
Projecting the published citation patterns above onto sports queries, segmented by query type. The numbers below are directional estimates derived from the published research and observable AI engine behavior on representative sports queries as of mid-2026. The full study (Phase 1) will replace these directional estimates with measured citation shares across 500 sports queries on five engines.
| Query Type |
Example |
Estimated Top Source |
Estimated Share |
ESPN's Estimated Share |
| Factual / historical | "Who won the 2024 World Series?" | Wikipedia | 35–45% | 10–15% |
| Entity / biographical | "How old is LeBron James?" | Wikipedia | 40–50% | 8–12% |
| Stats / records | "All-time NFL passing yards leader" | Wikipedia / Pro-Football-Reference | 35–50% | 10–15% |
| Breaking news | "Who broke the Tyreek Hill trade?" | ESPN (Schefter's X) | 30–40% | 30–40% |
| Live game data | "What's the score of the Lakers game?" | ESPN / Google direct answer | 25–35% | 25–35% |
| Opinion / takes | "Is the SEC overrated?" | Reddit | 30–50% | 5–10% |
| Fantasy / wagering | "Should I draft Justin Jefferson?" | Reddit / dedicated fantasy sites | 30–45% | 10–15% |
| Long-form analysis | "Why did the Pac-12 collapse?" | The Athletic / mainstream press | 15–25% | 10–20% |
Methodology note (pilot): Estimated shares are directional, projected from published AI engine citation studies (Profound, Semrush, Peec AI, ZipTie.dev) applied to representative sports query categories. Phase 1 of the EPR Sports Citation Index will replace these estimates with measured citation shares across 500 sports queries on ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. Target release: late 2026.
What this means for ESPN
The structural finding is that ESPN's brand authority — built over four decades on the cable bundle, the SportsCenter franchise, the talent roster, and the rights portfolio — does not translate one-to-one into AI engine citation share. The engines do not retrieve ESPN at the rate the network's commercial position would predict. Across the eight sports query categories above, ESPN dominates two — breaking news and live game data — and is materially under-cited in the other six.
The reason is structural. ESPN's content architecture is optimized for broadcast and app distribution: live programming, push notifications, the personalized feed. Wikipedia's content architecture is optimized for what AI engines retrieve: structured, entity-rich, persistent, citation-dense pages. The two operate by different rules. The AI engines are scoring on Wikipedia's rules, not ESPN's.
The competitive implication is that ESPN can either (a) build content infrastructure that performs in AI retrieval the way Wikipedia does — entity-anchored, schema-marked, structurally retrievable — or (b) accept that its brand authority lives in the live-and-breaking categories while encyclopedic knowledge migrates to other citation surfaces. ESPN's existing investments in original-source breaking news (Schefter, McAfee, the insider class) give it durable advantage in one of those categories. The encyclopedic and opinion categories are open to challengers.
The pattern is not ESPN-specific. Every sports media operator faces the same structural challenge. The Athletic's long-form analysis is being retrieved in modest but meaningful share. Bleacher Report's citation share has declined materially since 2023, consistent with the decline of mid-tier aggregation across the AI engines. Sports Illustrated, post-restructuring, is now in a tier where its AI citation share is roughly proportional to its current editorial activity, not its brand history.
The category-level lesson is that AI-engine citation share follows different rules than legacy media share. The leaders inside the AI engines — Wikipedia, Reddit, the original-source breaking news accounts — are not the leaders in cable audience or print circulation. The two surfaces are converging in importance, but the share distribution is materially different on each.
The methodology for Phase 1
The EPR Sports Citation Index Phase 1, targeted for late 2026, will measure citation share across 500 sports queries, distributed across the eight categories above and balanced across five engines.
Query selection. 500 queries balanced across factual, entity, stats, breaking news, live data, opinion, fantasy/wagering, and long-form analysis. Queries drawn from the most-searched sports questions in 2025-26 (per Google Trends and public search data) plus a balanced sample of long-tail queries representative of typical buyer behavior.
Engine coverage. ChatGPT (with and without web browsing), Claude, Gemini, Perplexity, Google AI Overviews. Each query run multiple times on each engine, citations logged per response.
Scoring. For each query × engine combination, citations from the top retrieved sources logged. Domain-level share calculated. Weighted scoring will reflect both the position of the citation (in-line vs. footnote) and the engine's audience weight.
Reporting. Citation share tables by domain, by category, by engine. A composite "Sports Citation Share Index" ranking the top 50 cited domains in sports queries across the AI engines. Comparative analysis to Profound, Peec AI, ZipTie, and Semrush baselines.
Limitations. AI engine citation behavior changes monthly. The Semrush study documented a dramatic mid-September 2025 ChatGPT citation adjustment that materially shifted the citation landscape. Any single-point-in-time study captures a moment, not a permanent state. The Phase 1 release will be the first measurement; future EPR Sports Citation Index editions will track movement over time.
The two category-level lessons
Both useful well beyond sports.
The first is that brand authority and AI citation share are different metrics. Brand authority is built on the cumulative trust of an audience that already knows you exist. AI citation share is built on the structural retrievability of your content inside engines that may not have any prior knowledge of your brand. The two metrics correlate, but loosely. The brand that wins one does not automatically win the other.
The second is that the migration of consumer questions from search to AI is, in many categories, a winner-restructuring event. Wikipedia's role as the encyclopedia of the internet was a Google-era position. Its role as the citation backbone of every major AI engine is, in 2026, even more central. Reddit's role as the community Q&A archive was a niche product of the social-web era. Its role as the dominant opinion-and-takes source inside Perplexity and Google AI Overviews is a category-defining position no one would have predicted in 2020.
For ESPN, for sports media, and for every brand whose value depends on being the cited source when buyers ask the question, the implication is the same. The question has not changed. The surface on which the answer appears has. The brands that win the next decade are the ones building for the new surface — not assuming the old one will hold.
Frequently asked questions
Does ESPN appear in AI engine answers about sports?
Yes, but selectively. ESPN dominates AI citations for breaking news, live game data, and original-source sports reporting (Adam Schefter's X account, ESPN.com beat coverage, ESPN's published box scores). ESPN is materially under-cited for encyclopedic factual queries (where Wikipedia dominates), opinion queries (where Reddit dominates), and biographical entity queries (where Wikipedia dominates). The Profound study of 680 million AI citations found Wikipedia accounts for 47.9 percent of ChatGPT's top 10 cited sources across all categories.
Why does Wikipedia dominate sports AI answers?
Wikipedia's content architecture is structurally optimized for AI retrieval. Pages are entity-anchored, citation-dense, persistently updated, and structured with the kind of schema and internal linking that retrieval-augmented generation systems are designed to surface. ESPN's content architecture is optimized for broadcast distribution and app delivery, not for AI retrieval. The structural mismatch is why Wikipedia outperforms ESPN in factual sports AI citations despite ESPN's larger audience and brand recognition.
What is the EPR Sports Citation Index?
The Sports Citation Index is a planned study from Everything-PR measuring citation share across 500 sports queries on ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. The pilot framework, published in mid-2026, establishes the methodology and directional findings. Phase 1, targeted for late 2026, will publish the full measured citation shares with domain-level rankings, query-category breakdowns, and engine-by-engine analysis.
Who is winning AI search for sports?
Wikipedia wins the largest single share across factual, entity, biographical, and statistical sports queries. ESPN wins breaking news and live game data. Reddit wins opinion, community, and fan-sentiment queries — particularly inside Perplexity, where Reddit accounts for approximately 46.7 percent of top citations across the platform's query mix per ZipTie's analysis. The Athletic competes in long-form analysis. Bleacher Report's share has declined materially since 2023.
How can sports brands improve their AI citation share?
Three structural moves: build content architecture that performs in retrieval (entity-anchored pages, FAQPage and Article schema, persistent canonical URLs); produce original-source content the AI engines can identify as first-cited (breaking news, exclusive interviews, proprietary data); and participate in the community surfaces (Reddit, forums) where opinion-and-takes citations live. The brand authority that wins on cable is structurally different from the citation authority that wins inside the AI engines.
