Platform-Specific Citation Dominance: Observational Evidence of Directory Market Share in AI Recommendations

1. Introduction

As AI systems increasingly mediate professional service discovery, the sources these systems cite when generating recommendations acquire outsized influence. Unlike traditional search results where users see ten blue links and can diversify their information intake, AI-generated answers typically present a synthesised narrative drawing from a smaller number of sources. This compression amplifies the power of whichever sources achieve citation status.

The traditional web ecosystem distributes influence across directories, review platforms, editorial publications, and individual service provider websites. The question this paper investigates is whether this distribution persists in AI-generated recommendations, or whether citation concentration emerges -- where a small number of sources, or even a single source, dominates across AI platforms.

Prior work in the SIGI research programme established that AI systems apply distinct trust hierarchies when evaluating sources (see SIGI-2026-064) and that search rank does not directly predict citation probability (see SIGI-2026-062). Building on these foundations, this paper examines the empirical distribution of citations across source types in a 500-search study spanning four major AI platforms.

2. Methodology

2.1 Study Design

The study comprised 500 searches across four AI platforms (anonymised as System A through System D) using service provider recommendation queries in a single industry vertical. Each query was designed to elicit agency recommendations, and the resulting citations were classified by source type: directory listing, agency self-published content, editorial content, review platform, or other.

2.2 Citation Classification

A citation was recorded when an AI system explicitly named a source, linked to a source, or attributed a factual claim to a specific platform. Implicit references (e.g., describing information clearly derived from a directory without naming it) were not counted, making these figures conservative estimates of actual directory influence.

2.3 Platform Anonymisation

All directory platforms are anonymised. The dominant platform is referred to as "Platform Alpha." AI systems are referred to as Systems A through D. This anonymisation follows SIGI's standard protocol for observational research.

2.4 Evidence Level

This study is classified as Evidence Level 2-3 (observational). Under the Logic-First methodology, this permits language describing observed associations and patterns but prohibits causal claims about why Platform Alpha dominates.

3. Results

3.1 Citation Market Share by AI Platform

Across all four AI platforms, Platform Alpha captured the majority of directory citations. The concentration ratio (Platform Alpha share divided by next-largest directory share) ranged from 4.5:1 to 11.7:1, indicating that Platform Alpha's dominance is not a marginal lead but a structural monopoly position.

3.2 Cross-Platform Consistency

Platform Alpha held the leading citation position across all four AI systems tested. No other directory platform achieved first position on any system. This cross-platform consistency is notable because each AI system uses different retrieval mechanisms, different training data compositions, and different answer synthesis architectures. Despite these architectural differences, the same platform dominates.

3.3 Paid Placement Paradox

Platform Alpha operates a widely documented paid placement model where service providers can pay for premium positioning within the directory. At least one AI system (System A) was confirmed to have training-data knowledge of this commercial practice (see SIGI-2026-063 for the training-data bias mechanism). Despite this knowledge, Platform Alpha's citation share in System A was the highest of all four platforms at 84.5%.

This suggests that whatever trust discount is applied for commercial practices is insufficient to overcome Platform Alpha's advantages in review volume, content structure, and entity density.

3.4 Source Type Distribution

The complete absence of independent editorial citations (0% across all four systems) is consistent with the editorial vacuum finding reported in SIGI-2026-062. Directories collectively account for the majority of citations, and Platform Alpha alone accounts for more citations than all other source types combined.

4. Discussion

The citation concentration observed in this study exceeds levels that would be considered healthy in any information ecosystem. When a single platform controls 66-84.5% of all AI-generated citations in an industry vertical, the recommendations users receive are effectively filtered through that platform's listing criteria, commercial policies, and content structure.

Several hypothesised mechanisms may explain Platform Alpha's dominance, though none can be confirmed at this evidence level. First, Platform Alpha has the largest review corpus, and AI systems are known to prefer sources with higher entity density and more verifiable claims (see SIGI-2026-064). Second, Platform Alpha's structured data format aligns with how AI systems extract information, potentially creating a format advantage independent of content quality. Third, Platform Alpha's web presence across multiple queries creates repeated retrieval, and AI systems may apply a frequency heuristic where repeatedly retrieved sources receive higher citation priority.

The paid placement paradox is particularly significant. Industry commentary has assumed that AI systems would naturally discount commercially influenced directories. Our observations suggest the opposite: Platform Alpha's commercial model appears to be a minor factor relative to its structural advantages in volume, format, and ubiquity.

For service providers, these findings identify a concentration risk. Dependence on a single directory platform for AI-mediated visibility creates vulnerability to that platform's policy changes, pricing decisions, and algorithmic adjustments.

5. Limitations

• Single industry vertical: All searches were conducted in one service industry category. Citation concentration patterns may differ across industries.
• Single time point: The 500-search study was conducted at a single time point. Citation distributions may shift with model updates and new content publication.
• Observational methodology: This study identifies patterns but cannot establish causal mechanisms for Platform Alpha's dominance.
• Conservative citation counting: Only explicit citations were counted. Implicit directory influence (information used without attribution) would increase actual dependency.

6. Conclusions

We observe extreme citation concentration, with a single directory platform accounting for 66-84.5% of AI agency citations across four platforms in a 500-search study. This dominance is consistent across all AI systems tested despite architectural differences between them. The finding that Platform Alpha's paid placement model does not materially reduce its citation share challenges the assumption that AI systems naturally discount commercially influenced sources.

The service industry AI citation market is effectively monopolised by a single directory, creating concentration risk for service providers and information quality risk for users receiving AI-generated recommendations.

Confidence: MODERATE. Based on a 500-search empirical study across four AI platforms. Single time point, single industry vertical. Causal mechanisms hypothesised but not experimentally confirmed.