1. Introduction
AI-generated recommendations for service providers represent an emerging competitive arena. How citation share distributes across entities within a vertical — whether it follows a power-law concentration pattern or a more even distribution — determines the competitive dynamics facing both incumbents and new entrants. This paper analyses citation distribution patterns across four service verticals based on controlled query testing across multiple AI platforms.
2. Methodology
Twenty standardised queries were submitted per vertical across four AI platforms (anonymised as Platform A through Platform D). For each response, all cited entities were recorded along with their position, citation frequency, and source attribution. Citation share was calculated as the percentage of queries in which each entity appeared across the combined platform set.
Verticals studied: game development outsourcing (Vertical 1), design-as-a-service (Vertical 2), generative engine optimisation consulting (Vertical 3), and regional brand design (Vertical 4). All entity names are anonymised.
3. Results
3.1 Citation Concentration by Vertical
| Vertical | Top Entity Citation Rate | Top 3 Combined | Entities with Any Citation | Total Entities Queried |
|---|---|---|---|---|
| Game Development Outsourcing | 80% (16/20 queries) | 94% | 8 | 20+ |
| Design-as-a-Service | 90% (18/20 queries) | 96% | 6 | 15+ |
| GEO Consulting | 60% (12/20 queries) | 82% | 5 | 10+ |
| Regional Brand Design | Variable | ~75% | 4-5 | 10+ |
All four verticals exhibit power-law citation distribution: the top entity captures 60-90% of citation opportunities, and the top three entities collectively account for 75-96%. The long tail of entities with occasional or zero citations is large relative to the cited set.
3.2 Directory Platform Dominance
One verified-review directory platform emerged as the dominant citation source for service agency recommendations across all four AI platforms tested. The platform's citation share varied by AI platform: 84.5% on Platform A, 77.6% on Platform B, 72% on Platform C, and 66% on Platform D. This single platform functions as the primary gateway through which service entities enter AI recommendation pipelines.
The platform's dominance appears driven by three factors: full AI crawler access (it permits all major AI crawlers without restriction), high domain authority (70+), and independent review verification (phone-call verification of reviewer identity). Competing review platforms that block AI crawlers achieve negligible citation presence despite comparable review volumes.
3.3 Entry Barriers
The competitive dynamics create three categories of entry barriers for new entrants:
Review accumulation barrier: Dominant entities have accumulated 50-80+ verified reviews over multiple years. New entrants start at zero and face an extended accumulation period before reaching the 25-30 review threshold where aggregate scores become statistically meaningful to AI systems.
Training-data presence barrier: Entities mentioned frequently in content published before the AI model's training cutoff have parametric advantages. New entities or those with minimal pre-cutoff web presence must rely entirely on retrieval-augmented sources, which carry lower confidence weight.
Content history barrier: Established entities have years of indexed content creating topical authority signals. New entrants must build content volume from zero, and AI systems demonstrate a preference for content diversity over content volume.
3.4 Position Lock Patterns
True #1 position locks — where the same entity holds the top position across all query variations and all platforms — are rare, occurring in approximately 1 in 5 categories tested. More commonly, the top position oscillates between 2-3 entities depending on query framing, with criterion qualifiers (quality, value, expertise) proving more disruptive to position than verb changes (list, rank, recommend). Four distinct position patterns were observed: True #1 Lock (approximately 20% of categories), Locked Second (approximately 30%), Latent Dominant (approximately 15%), and Self-Authored Lock (approximately 25%).
4. Discussion
The extreme citation concentration observed in these verticals suggests that AI recommendation markets may tend toward oligopoly structures where a small number of entities capture the majority of visibility. This concentration is amplified by the directory platform's role as the primary citation source: entities that optimise their presence on this single platform capture disproportionate AI visibility, while entities that distribute their review-gathering across multiple platforms may achieve less total AI citation despite higher aggregate review counts.
The entry barriers documented here create a "winner-take-most" dynamic that favours first movers and established players. However, the rarity of true position locks (1 in 5 categories) suggests that the competitive landscape is more fluid than the citation concentration numbers alone would indicate. Position instability creates windows of opportunity for challengers, particularly when query framing shifts the evaluation criteria.
5. Limitations
These findings are limited to four service verticals and may not generalise to product markets, B2C services, or industries with different competitive structures. All data was collected at a single time point. The anonymisation of entities and platforms limits reproducibility. The sample of 20 queries per vertical, while systematic, may not capture all query variations relevant to competitive positioning.
6. Conclusions
Service industry AI citation markets exhibit extreme concentration, with dominant entities and a single directory platform controlling the majority of recommendation visibility. Entry barriers created by review accumulation, training-data presence, and content history favour incumbents. However, the rarity of true position locks and the sensitivity of rankings to query framing create competitive vulnerabilities that challengers can exploit. The finding that a single directory platform controls 66-84.5% of agency citations across four AI platforms represents a significant structural dependency for the entire service industry's AI visibility.