What 50,000 Competitive AI Queries Taught Us About Who Actually Wins Brand Visibility

When we built the competitor benchmarking feature into LLM Metrix, we expected it to confirm what most AI visibility practitioners already believed: bigger brands with more content win, and smaller brands are left to fight over the scraps. The data turned out to be considerably more interesting than that.

Over the past six months, we’ve been analyzing anonymized competitive query data from LLM Metrix customers — approximately 50,000 tracked queries across 34 product categories, spanning brands that range from well-funded category leaders to early-stage startups. Here’s what we found.

Finding 1: Category leader ≠ AI visibility leader in most categories

In 21 of the 34 product categories we analyzed, the brand with the highest AI visibility score was not the brand with the highest estimated market share or search traffic dominance.

This was the finding that generated the most internal discussion. We expected a strong correlation between market position and AI visibility — and for a handful of categories (cloud storage, general productivity, CRM), it existed. But in the majority, the AI visibility rankings and market share rankings diverged, often significantly.

The divergence pattern was consistent: the AI visibility leaders were brands that had invested heavily in educational content and category definition — regardless of how large they were by revenue or headcount. They were the brands that published “what is X” and “how to choose X” content early, built topic authority around the category, and maintained that content investment over time.

Market leaders who had coasted on brand recognition and distribution found that AI systems, unlike search engines, don’t weight brand familiarity. They weight content authority, citation patterns, and retrieval relevance.

The implication: If you’re a challenger brand with disciplined content investment, AI visibility is the most level competitive playing field you’ve had in a decade. The incumbents’ brand advantage doesn’t transfer here the way it did in paid search or traditional SEO.

Finding 2: Win rate concentrates on a small number of queries — for everyone

Across our dataset, the average brand won first mention on just 24% of tracked queries in its category. The median brand won first mention on only 18%. But the distribution wasn’t flat — it was sharply concentrated.

The typical pattern: a brand achieves first mention on roughly 60–80% of the queries it does well on, and is absent or buried on the rest. Competitive positions are not spread evenly. They cluster around specific topic areas where the brand has established retrieval authority.

This means that most brands have a small set of “topic strongholds” — the 10–20% of queries they consistently win — and a much larger set of queries where they’re competitive but not dominant, or where a specific competitor has clearly won.

The implication: The most effective AI visibility strategy isn’t trying to win everywhere. It’s identifying your natural topic strongholds, defending them, and systematically converting the adjacent contested queries — the ones you appear in but don’t win — rather than trying to recover losses from your most firmly established competitors.

Finding 3: Content recency effects differ sharply by category

We already knew recency matters for AI retrieval. What surprised us was how much the recency effect varies by category type.

In categories characterized by rapid change — AI tools, cybersecurity, developer infrastructure — content older than three months showed significantly reduced citation frequency. These are categories where “current” and “last year” represent genuinely different realities, and AI retrieval systems appear to calibrate accordingly.

In evergreen categories — accounting software, project management, HR tools — content from 18–24 months ago performed almost as well as recent content in retrieval, provided it remained factually accurate. The “information clock” ticks at different speeds by category.

This has a direct consequence for how brands should allocate their content refresh budget. If you’re in a fast-moving category, stale content is an active liability — refreshing it is higher priority than creating new content. If you’re in a stable category, your refresh cycle can be longer, and the ROI from new content creation is higher relative to refresh.

How we measured it: For each brand in our dataset, we looked at the publication dates of the pages appearing in their citation traces and compared the citation frequency of pages in different age cohorts, controlling for domain authority and content quality signals.

Finding 4: Share of voice diverges significantly from share of first mention

These two metrics — share of voice (how often you appear across all query responses) and share of first mention (how often you’re first) — are often treated as similar, but in our data they told very different stories about competitive position.

We found brands where share of voice was high (appearing in 70–80% of responses) but share of first mention was low (first in only 15–20% of those). These brands were a consistent listed presence across the category but rarely the leading recommendation. Their content was sufficient to be retrieved and included, but not authoritative enough to anchor the recommendation.

We also found the reverse: brands with moderate share of voice (40–50%) but very high share of first mention (first in 55–65% of responses). These brands appeared less often overall, but when they appeared, they led. Their category positioning was narrower but their authority within that position was much stronger.

The second profile was consistently more commercially valuable. Appearing often as a third or fourth mention has limited brand discovery value. Appearing less often but anchoring the recommendation is the position that drives awareness and consideration.

The implication: Share of voice as a standalone metric can be misleading. The brand tracking their visibility by “how often am I mentioned” may feel like they’re competing, when they’re actually just making the list. First-mention rate, or share of first mention, is the metric that reveals whether you’re actually winning the query.

Finding 5: Competitive gains are faster — and more fragile — than most brands expect

One finding that surprised us was the speed of competitive change at the query level. We measured how frequently brands exchanged first-mention positions on specific queries over a rolling 8-week period.

In categories with high content publishing velocity (AI tools, marketing technology, developer platforms), approximately 35% of first-mention positions changed hands at least once over 8 weeks. In stable categories, that figure dropped to around 12%.

The takeaway: in fast-moving categories, the competitive landscape is much more fluid than brands realize. A competitor you’re losing to today may have lost that position in six weeks — and a strong position you hold may erode in the same window if you stop refreshing.

This also means that brands in high-velocity categories who are actively investing in content are seeing faster results from that investment — the competitive positions aren’t as entrenched as they appear. The brands waiting to build content authority “once the category matures” are making a strategic error: they’re ceding ground during the period when it’s cheapest to take.

Finding 6: Third-party citation authority is the hardest competitive advantage to overcome

When we looked at the competitive gaps that were most persistent — queries where one brand had held first mention for 6+ months while challengers came and went — the most consistent differentiator was third-party citation authority, not owned content quality.

The category leaders in persistent first-mention positions were almost always the brands most frequently cited by review platforms, industry publications, and comparison sites. Not by large margins in any single source, but by consistent, distributed citation patterns across many sources.

Owned content quality (comprehensiveness, recency, structure) was the determining factor in contested positions — where multiple brands had reasonable content. Third-party citations were the factor that created separation at the top.

This maps to the underlying retrieval dynamics: when multiple brands publish good content for the same query, the retrieval system’s tiebreaker is authority signals from the broader web. Third-party citations are those authority signals.

The implication: If you’re executing well on owned content and still losing to a dominant competitor on your highest-priority queries, the gap is almost certainly citation authority. The path to closing it is earning citations from the authoritative third-party sources in your category — review platforms, trade publications, comparison sites — not publishing more owned content.

---

The competitive picture that emerges

Stepping back, the data paints a consistent picture of how competitive AI visibility actually works:

Market leadership doesn’t transfer automatically. Category definition through educational content matters more than brand recognition. Competitive positions concentrate around topic clusters, not categories at large. The gap between a sophisticated and unsophisticated approach to AI visibility is larger than in any previous search channel — and the gap is growing as more brands wake up to it.

The brands that will dominate AI visibility in two years are building that dominance right now through consistent content investment, deliberate third-party citation strategy, and tight monitoring of where their competitive positions are being won and lost. The brands waiting to act are not just falling behind — they’re watching the competitive moats of their better-prepared competitors deepen.

We’ll publish follow-up analysis on how these patterns vary by company stage (startup vs. scale-up vs. enterprise) in a future piece. If you’re interested in how your brand’s competitive position compares to these benchmarks, the competitor benchmarking feature in LLM Metrix gives you the query-level data to run this analysis against your own category.