Introduction

As AI search platforms reshape how users discover information, understanding citation patterns has become critical for publishers seeking to maintain visibility in this emerging landscape. Unlike traditional search engines that present a list of links, AI search platforms synthesize information and selectively cite sources. Making those citation decisions pivotal for publisher traffic and authority.

This week's report analyzes citation patterns across ChatGPT-4, ChatGPT-5, and Perplexity using data from ALLMO.ai’s AI Search Trends research. The methodology encompasses diverse prompts spanning news topics, autocomplete suggestions, and category-specific queries across multiple languages, providing a comprehensive view of how these platforms prioritize sources.

The findings reveal stark differences in platform strategies, dramatic concentration around traditional wire services on newer models, and significant opportunities for specialized publishers. See the full data at https://allmo.ai/trends.

Top 3 Highlights: Week 42, October 2025

1. The Reuters Dominance on ChatGPT-5: Extreme Winner-Take-Most Dynamics

This week Reuters captures an extraordinary 74.2% share of citations on ChatGPT-5 responses, a 60% increase from its already-dominant 46.5% performance on ChatGPT-4. This level of concentration suggests OpenAI's latest model increasingly relies on a few authoritative wire service, potentially marginalizing mid-tier publishers who are competing for visibility. The shift signals that brand authority and comprehensive coverage may matter more than ever in AI search optimization.

2. Perplexity's Video-First Strategy: YouTube Commands Half of All Citations

YouTube accounts for 50.4% of top citations on Perplexity, demonstrating a fundamentally different approach from ChatGPT platforms where video content falls outside the top 10 entirely. This platform divergence creates a critical strategic decision point for publishers: optimizing text-based content for ChatGPT or investing in video for Perplexity requires distinct content strategies and resources.

3. Wikipedia's Dramatic Decline on ChatGPT-5: From Reference Standard to Secondary Source

Wikipedia citations plummeted from 37.8% on ChatGPT-4 to just 22.5% on ChatGPT-5, suggesting newer models increasingly prioritize primary news sources over encyclopedic references. This 40% drop indicates that ChatGPT-5's training emphasizes contemporary journalism and authoritative reporting, potentially disadvantaging general reference content in favor of timely, specialized coverage.

Overall Platform Leaders

ChatGPT-5

• Reuters dominates with a 74.2% citation rate, nearly 1.6× its ChatGPT-4 performance and demonstrating unprecedented concentration around a single wire service.

• AP News maintains a strong presence at 30.0%, showing traditional wire services collectively command over 100% of top citations (indicating frequent multiple citations per query).

• The Guardian emerges with 18.3%, representing a significant jump from minimal ChatGPT-4 visibility and suggesting the platform values international perspectives from established news brands.

• Financial Times and TechCrunch achieve 8.3% each, indicating specialty publishers can compete when offering authoritative vertical expertise.

• Wikipedia drops to 22.5%, marking a dramatic shift away from reference material toward primary news sources.

• Traditional broadcasters like ABC News and CBS News fall outside the top 10, suggesting ChatGPT-5 prioritizes print journalism and wire services over broadcast media.

Key takeaway: ChatGPT-5 demonstrates extreme concentration around traditional wire services, with Reuters alone capturing nearly three-quarters of top citations. The platform’s preference for established news brands with comprehensive global coverage creates formidable barriers for mid-tier publishers, while the reduced reliance on Wikipedia signals a strategic shift toward primary journalism over encyclopedic references.

ChatGPT-4

• Reuters leads at 46.5%, establishing wire-service dominance though at significantly lower concentration than ChatGPT-5.
  Wikipedia maintains strong presence at 37.8%, serving as a primary reference source for factual queries and background information.

• AP News captures 29.1%, reinforcing the wire-service duopoly that characterizes ChatGPT’s source preferences.
  Axios achieves 10.2%, demonstrating that concise, structured news formats resonate with the platform’s synthesis approach.

• Le Monde reaches 7.9%, indicating international sources gain visibility when covering region-specific queries.

• Democracy Now and Time appear at 7.1% and 5.5% respectively, showing diverse editorial voices maintain meaningful presence.

Key takeaway: ChatGPT-4 exhibits more balanced citation patterns than its successor, providing pathways for diverse publishers including alternative media, international sources, and specialty outlets. The stronger Wikipedia presence suggests this model relies more on reference material for contextual grounding.

Perplexity

• YouTube dominates with a 50.4% citation rate, reflecting the platform’s unique emphasis on video and multimedia responses.

• Wikipedia achieves 24.4%, maintaining relevance as a reference source though significantly lower than on ChatGPT-4.

• ABC News and CBS News both reach 13.0%, demonstrating broadcast media performs notably better on Perplexity than ChatGPT platforms.

• JPMorgan captures 11.5%, revealing institutional and corporate sources gain unusual prominence for financial topics.

• Euronews reaches 9.2%, suggesting international broadcast media finds stronger footing on this platform.

• Traditional wire services like Reuters and AP News fall outside the top 10.

Key takeaway: Perplexity operates with a fundamentally different source strategy, prioritizing video, broadcast, and institutional sources over the wire-service dominance seen on ChatGPT. Publishers with video operations benefit most here.

Category Analysis

Economics & Finance

• Reuters leads at 48.5%, though notably lower than its overall ChatGPT-5 dominance, suggesting even premier categories show some diversification

• IMF captures impressive 27.3%, demonstrating institutional sources command extraordinary authority in specialized economic topics

• AP News maintains 21.2%, providing wire service competition though significantly trailing Reuters in this vertical

• JPMorgan achieves 18.2%, showing corporate financial institutions successfully compete as primary sources for market analysis

• S&P Global reaches 15.2%, indicating ratings agencies and financial data providers gain citation authority

• Investment firms including T. Rowe Price (12.1%) and John Hancock (12.1%) appear prominently, revealing AI platforms treat asset managers as authoritative voices

• TechCrunch falls outside top 10 entirely, confirming this category prioritizes traditional financial institutions over tech-business crossover publications

Key takeaway: Economics and finance display high institutional concentration—multilateral organizations and banks rival media outlets for authority.

Start-ups & Technology

• Reuters leads at 28.4%, maintaining wire service advantages even in specialized tech coverage, though with significantly lower dominance than other categories

• TechCrunch achieves 25.4%, demonstrating vertical specialists can approach parity with wire services when offering deep domain expertise

• Wikipedia reaches 14.9%, reflecting reliance on reference material for technology company information and founder biographies

• StartupBlink and Axios both capture 14.9%, showing data-focused startups and structured news formats compete effectively in this category

• JPMorgan maintains 11.9% even in tech topics, indicating financial institution analysis extends beyond pure finance categories

• Multiple niche publications (KPMG 10.4%, WEF 9.0%, StartupSavant 9.0%, TechStartups 9.0%) achieve meaningful visibility, creating the most fragmented citation landscape of any category

Key Takeaway: Technology coverage shows the least concentrated citation patterns, with specialist publishers achieving near-parity with traditional wire services. The prominence of data aggregators, consulting firms, and niche tech publications suggests expertise and comprehensive startup databases matter more than general news authority, offering clear pathways for vertical specialists to compete.

Sports

• Wikipedia dominates at 40.0%, reflecting AI platforms' reliance on structured athlete biographies, career statistics, and historical records that encyclopedic sources provide as background information, even when asked about recent sport events

• Reuters maintains 33.3%, showing wire services remain relevant for game coverage and breaking sports news despite encyclopedia competition

• ESPN captures 31.1%, demonstrating sports specialists can achieve strong visibility when offering comprehensive coverage and statistical depth

• YouTube reaches 17.8%, indicating video highlights and analysis gain traction in sports more than other news categories

• Sky Sports achieves 15.6%, showing international sports broadcasters maintain meaningful presence alongside U.S.-focused outlets

• Sports Illustrated reaches 11.1%, while niche outlets like Bleacher Report and UEFA.com both hit 8.9%, revealing diverse specialist sources compete effectively

Key Takeaway: Sports represents the only category where Wikipedia outperforms all news sources, driven by AI platforms' need for biographical data, career statistics, and historical context that structured encyclopedic content provides. The strong showing by vertical specialists like ESPN suggests comprehensive statistical databases and expert analysis create competitive advantages against general news wire services.

General News & Politics

• Reuters achieves 46.5%, demonstrating traditional news authority reaches peak importance for general political coverage

• Wikipedia maintains 44.6%, nearly matching Reuters as platforms rely on reference material for political figures, governmental institutions, and historical context

• AP News captures 31.7%, creating a wire service duopoly that commands over 75% of top citations

• YouTube reaches 27.7%, showing video news analysis gains meaningful visibility for political topics on Perplexity-style platforms

• Axios achieves 6.9%, representing the highest-visibility specialty publisher in general news, likely due to structured political coverage

• Niche publishers fall outside top 10, suggesting general news creates the highest barriers for specialized outlets competing against established brands

Key Takeaway: General news and politics demonstrates the most traditional citation hierarchy, with wire services and Wikipedia dominating in patterns resembling legacy news consumption. The near-absence of specialist publishers outside the top tier suggests breaking political news and governmental coverage favor established institutional brands with comprehensive beat coverage, creating the steepest competitive barriers of any category.

Regional Analysis: German-Language Queries

• Government sources dominate German-language visibility, with the Bundestag (23.3%) and Bundesregierung (21.9%) leading across all platforms, showing AI models heavily favor official institutional information.

• Reuters maintains strong cross-language authority, reaching 51.9% on ChatGPT-5 despite being English-language, underscoring global trust in wire services over local outlets.

• German Wikipedia (20.5%) outperforms commercial media, filling gaps where domestic news coverage is limited or paywalled.

• Public broadcasters lead among media, with Deutschlandfunk (15.1%) and ZDF (13.7%) ranking above all commercial newspapers.

• Traditional print media underperform, as Süddeutsche Zeitung (12.3%) and Die Welt (12.3%, 29.6% on ChatGPT-5) trail government and broadcast sources, while FAZ drops out of the top 10 entirely.

Platform-Specific Patterns

• Perplexity strongly favors German public media, with Deutschlandfunk (39.1%), ZDF (30.4%), and Süddeutsche Zeitung (30.4%) achieving dramatically higher visibility than on ChatGPT platforms

• ChatGPT-5 maintains wire service preference even for German queries, with Reuters dominating at 51.9% despite language mismatch, suggesting English-language authority trumps linguistic relevance

• YouTube reaches 34.8% on Perplexity but falls outside top 10 on ChatGPT, reinforcing the platform's video-first approach extends to non-English markets

Key Takeaway: German-language queries reveal government and institutional sources command authority that commercial media rarely achieves in non-English markets, with parliament, ministries, and public broadcasters collectively dominating citations. The surprising strength of English-language Reuters even for German topics suggests AI platforms prioritize source authority over linguistic match, creating challenges for regional commercial publishers competing against both governmental institutions and international wire services.

Strategic Implications

Platform-Specific Optimization Requirements

• ChatGPT platforms demand wire service authority or specialized vertical expertise, with Reuters' 74.2% ChatGPT-5 dominance indicating mid-tier general news publishers face increasingly steep barriers to visibility

• Perplexity requires video content strategy, as YouTube's 50.4% citation rate demonstrates text-only publishers essentially concede half of potential visibility without multimedia capabilities

• Multi-platform presence necessitates contradictory content strategies, forcing publishers to choose between optimizing text-based journalism for ChatGPT or investing in video production for Perplexity

• Wikipedia's differential performance (37.8% on ChatGPT-4 vs 22.5% on ChatGPT-5) suggests reference-style evergreen content matters more on older models while breaking news drives newer platform citations

• Regional platforms show distinct preferences, with Perplexity strongly favoring local broadcasters (Deutschlandfunk 39.1% in German queries) while ChatGPT maintains English-language wire service dominance globally

Key Takeaway: Publishers can no longer pursue unified AI search strategies, as platform architectures create fundamentally incompatible optimization requirements. Success demands either specialization around single platforms or significant resource investments to maintain competitive positioning across text-driven ChatGPT and multimedia-focused Perplexity simultaneously.

Vertical Specialization as Competitive Advantage

• Category specialists achieve near-parity with wire services in their domains, with TechCrunch's 25.4% technology citation rate approaching Reuters' 28.4% in the same category

• Institutional sources dominate specialized topics, as IMF's 27.3% economics citation rate and JPMorgan's 18.2% demonstrate primary institutions outcompete journalism in vertical expertise

• Fragmented categories offer publisher opportunities, with technology showing six sources above 9% while general news concentrates around Reuters-AP-Wikipedia exclusively

• Sports vertical demonstrates encyclopedia advantages, where Wikipedia's 40.0% dominance suggests structured biographical and statistical data trumps breaking news coverage

• Niche publications achieve visibility in specialized contexts, with StartupBlink (14.9%) and StartupSavant (9.0%) gaining meaningful technology citations despite minimal general news presence

Key Takeaway: General news publishers face insurmountable concentration dynamics favoring wire services, but vertical specialists offering deep domain expertise, proprietary data, or institutional authority can achieve competitive visibility. The technology category's fragmentation demonstrates specialized knowledge creates defensible positioning against wire service commodification of general news.

Concentration and Long-Tail Challenges

• Winner-take-most dynamics intensify with newer models, as Reuters' citation rate jumped from 46.5% (ChatGPT-4) to 74.2% (ChatGPT-5), suggesting mid-tier publishers face accelerating marginalization

• Top three sources command 75%+ of citations in most categories, with Economics (Reuters 48.5%, IMF 27.3%, AP 21.2%) showing typical concentration patterns that leave minimal space for additional voices

• Platform diversity matters for portfolio risk, since sources outside ChatGPT-5's top tier (like broadcast media) maintain meaningful Perplexity presence, making multi-platform visibility increasingly essential

• Algorithmic monoculture risks emerge, where single-source dominance (Reuters 74.2% on ChatGPT-5) could reduce information diversity compared to multi-source traditional search results

• Smaller publishers appear increasingly dependent on category specialization, as general news concentration leaves few pathways for non-wire service visibility outside vertical expertise

Key takeaway: Rising concentration means mid-tier publishers must specialize or risk invisibility. Platform diversity and vertical depth are now survival strategies.

Conclusion

AI search platforms are creating bifurcated publisher landscapes: wire services and specialized vertical experts gain concentrated visibility, while general news publishers without category dominance face declining citation rates. The dramatic differences between ChatGPT's text focus and Perplexity's video emphasis mean publishers can no longer optimize for "AI search" generically but must make strategic platform choices.

As these patterns continue evolving, ongoing monitoring through platforms like ALLMO.ai becomes essential for publishers tracking their visibility and adjusting content strategies. The rapid shift from ChatGPT-4 to ChatGPT-5 citation patterns suggests the competitive landscape will continue changing quickly, making sustained attention to AI search trends a strategic imperative for any publisher seeking to maintain relevance in this emerging discovery channel.

This report is part of ALLMO.ai’s ongoing weekly analysis of AI search visibility in the media & publishing industry. An updated edition will follow next week with the latest data and platform shifts.