No category leader: AI recommendations are universally fragmented

Short answer: AI engines do not crown category winners. In Obsurfable's research corpus, every one of 325 categories with sufficient mention volume is fragmented — no category has a single brand above 50% of mentions, and none has a top-three block above 60%. The median #1 brand holds 2.9% of mentions in its category. The most concentrated category in the entire corpus — payments infrastructure — tops out at 9.9% (Stripe).

If you are tracking "market leader in AI recommendations," the data says that leader does not exist at the category level.

What we analysed

We measured brand-mention concentration across Obsurfable Explorer's research corpus: buyer-style prompts grouped by vertical and subvertical (e.g. technology · API platforms, media · developer media). For each prompt we keep the latest observation per platform family — ChatGPT, Gemini, Claude, Perplexity, and Grok — and aggregate every brand named across those captures.

We classify each category's mention landscape using two thresholds:

• Dominant: the #1 brand holds ≥50% of mentions
• Contested: the top three brands hold ≥60% combined (but no single brand ≥50%)
• Fragmented: everything else

We also report the Herfindahl–Hirschman Index (HHI) on brand mention shares — lower values mean more fragmentation. Traditional antitrust "highly concentrated" markets typically exceed HHI 0.25; our categories median 0.007.

Browse live category rankings on Obsurfable Explorer.

What we found

1. Every measured category is fragmented

Among 325 categories with at least 50 brand mentions:

No category — from data infrastructure to CRM to payments — has a single brand that commands even one-tenth of AI recommendation mentions.

2. Category leaders are weak by traditional standards

Distribution of #1 brand share across all 325 categories:

85.5% of categories have a #1 brand below 5% mention share. The median category contains 286 unique brands; the largest — data infrastructure — has 2,314 distinct named products across 744 prompts.

3. High-volume categories are just as fragmented

The five categories with the most brand mentions:

Even API platforms — the most concentrated high-volume category — spreads mentions across 968 brands. Its top five (Postmark 9.1%, Twilio 9.0%, Mailgun 8.3%, Amazon SES 7.0%, Resend 6.6%) account for only 40% of mentions combined.

4. The "most concentrated" categories still have no winner

Among categories with 100+ mentions, the highest #1 shares in the corpus:

Stripe at 9.9% is the ceiling — not the floor. In a category with 161 named brands, the most-mentioned product still appears in fewer than one in ten recommendation slots.

5. Engines name many brands per answer — but spread them thin

Average brands named per answer in the highest-volume categories:

Engines produce long shortlists — typically 6–11 brands per answer — but those mentions distribute across hundreds or thousands of products at the category level. Volume of naming does not translate to concentration.

Worked examples

Technology · API platforms — the corpus's most concentrated major category: explorer.obsurfable.com/categories/technology/api-platforms

Five vendors split the top of the leaderboard within a few percentage points, and 960+ other brands still accumulate mentions. Compare platform-specific splits: ChatGPT vs Perplexity — concentration differs by engine, but neither produces a runaway winner.

Technology · data infrastructure — the highest mention volume, lowest concentration: explorer.obsurfable.com/categories/technology/data-infrastructure

Bright Data leads at 4.8%, but the top ten brands account for only 19.9% of mentions. Over 2,300 unique products appear across 744 prompts — a long tail that no single vendor can dominate.

Why this is surprising

Market-share reports in traditional software often show clear leaders — Salesforce in CRM, Stripe in payments, AWS in cloud. AI recommendation data tells a different story:

1. Long shortlists, flat distribution. Engines name many brands per answer (median ~7–10 in top categories) but no single brand accumulates enough share to qualify as a category leader.

2. "Leaderboard" framing misleads. A brand ranked #1 in a category may hold under 3% of mentions. Reporting "#1 in AI recommendations" without the denominator implies dominance that the data does not support.

3. Concentration varies by category shape, not category size. API platforms (968 brands, 9.1% top share) is more concentrated than data infrastructure (2,314 brands, 4.8%) — but both are firmly fragmented. Volume of prompts does not create winners.

4. Complements cross-engine disagreement. Our prior work showed that matched prompts yield different brand shortlists across ChatGPT, Gemini, and Claude. Fragmentation at the category level means even aggregating across engines does not produce a dominant leader — it produces a wider, flatter field.

Limitations

• Mention threshold: We include categories with 50+ brand mentions. Smaller categories (below threshold) may show different patterns but lack statistical stability.
• Latest observation per engine: We aggregate the most recent capture per platform family per prompt, not full time-series or all historical runs.
• Brand extraction: Mention counts reflect Obsurfable's entity extraction on model outputs. Generic terms and parent-company vs product naming may affect counts.
• Category taxonomy: Vertical/subvertical groupings follow Explorer's category tree. Prompts assigned to broad categories (e.g. data infrastructure) mix sub-topics that may have different competitive dynamics.
• Not citation analysis: This report measures brands named in answers, not URL citations or domain authority.

Methodology

Corpus: Buyer-style research prompts tracked across major AI engines in Obsurfable Explorer.

Platform families: Raw platform values mapped to public families — e.g. ChatGPT UI and OpenAI API → ChatGPT; Gemini UI and Google API → Gemini; Claude UI and Anthropic API → Claude.

Deduplication: For each (prompt, platform family) pair, keep the latest observation; prefer UI captures over API runs at equal timestamps.

Concentration metrics: For each category, count every brand in brands_mentioned across deduplicated observations. Compute share = brand count / total category mentions. HHI = Σ(share²).

Landscape thresholds: Dominant = top-1 share ≥50%. Contested = top-3 share ≥60% and top-1 <50%. Fragmented = all other cases with 50+ total mentions.

Exclusions: Customer-scoped prompts and observations on unmapped platform keys omitted.

Reproduce category rankings at explorer.obsurfable.com.

Conclusion

Across 325 buyer categories and 38,941 observations, AI recommendation landscapes are universally fragmented. The median category leader holds 2.9% of mentions; the corpus maximum is 9.9%. Zero categories qualify as dominant or even contested by conventional concentration thresholds.

Tracking a single "AI visibility leader" per category overstates what the data supports. Measurement should report mention share with denominators, track concentration metrics (HHI, top-N share), and treat category rankings as long-tail distributions — not winner-take-all markets.