Brandi AI Research: Most-Cited Human-Written Pages Earn Nearly Six Times More AI Citations Than Most-Cited AI-Written Pages

Brandi AI's Q2 2026 study of approximately 900,000 AI answers found that the 25 most-cited human-written webpages averaged 682 citations each, 5.7x the 119 average for the top AI-written pages. Human-written pages represented 23% of all citations versus 4% for AI-written pages among the top 25, while the top 1% of human-written content captured 27% of citations compared with less than 1% for AI-written content. The findings suggest publishers and brands seeking generative-engine visibility should prioritize original research, proprietary data and expert analysis over scaled AI-generated output.

Analysis

The investable implication is not that all human-produced content gains value; it is that scarce, proprietary source material gains pricing power as generic text becomes commoditized. RELX and Thomson Reuters (TRI) are better positioned than ad-funded publishers because their workflows embed proprietary data, expert curation and recurring enterprise subscriptions—assets that are difficult for answer engines to replace without licensing or attribution risk. Over 6-18 months, this should support relative multiple resilience for information-services incumbents versus digital publishers whose traffic can be disintermediated by AI summaries.

For GOOG and MSFT, higher-quality retrieval requirements raise the strategic value of licensed and authoritative content, but the economics are mixed. Better answers can improve retention and enterprise adoption, yet each answer that satisfies intent without a click pressures legacy search/referral monetization and may increase content-acquisition costs. The near-term financial effect is likely immaterial versus cloud and advertising drivers; the relevant catalyst is whether AI-answer usage causes measurable query monetization dilution or forces large publisher licensing commitments.

The study is promotional, uses inferred authorship rather than verified production methods, and measures citations rather than referral traffic, conversion, or revenue. Consensus may overread citation concentration as a direct publisher benefit: citations can create brand authority while still reducing pageviews. The more durable trade is ownership of proprietary datasets and professional workflow distribution, not a broad long in media or "GEO" marketing vendors.

Key Decisions for Investors

• Maintain a 6-18 month long RELX / short broad digital-media basket posture rather than buying publishers on AI-citation headlines. RELX has recurring-data monetization and lower dependence on click-through traffic; invalidate if enterprise organic growth decelerates below management's medium-term range or AI-driven product investment materially compresses margins.
• Keep GOOG and MSFT unchanged on this item; treat it as a watch signal, not a catalyst. Add to diligence dashboards: AI-answer query share, search revenue per query, referral traffic trends, and disclosed content-licensing expense over the next 2-4 earnings cycles.
• For TRI, consider accumulating only on market-wide software/information-services weakness, with a 12-month horizon. The thesis requires continued Legal and Tax & Accounting organic-growth durability plus evidence that AI features are monetized as premium workflow tools rather than bundled at no incremental price.
• Avoid a directional long in ad-funded publishers solely on expected AI citations. Reassess only if third-party data demonstrate that cited placement produces measurable referral conversion or licensing revenue sufficient to offset zero-click traffic loss within 1-3 quarters.