AI Chatbots Are Making People All Think the Same, Study Says

Paper published in Trends in Cognitive Sciences warns that widespread use of a small number of LLM chatbots by hundreds of millions risks homogenizing language, reasoning and reducing pluralism. Usage data cited: one-third of Americans used ChatGPT in 2024 (double 2023), two-thirds of teens use chatbots (nearly one-third daily), and 78% of organizations reported AI use in 2024 (up from 55% in 2023); authors argue this trend can erode creativity, collective problem-solving and exert social pressure on non-users to conform.

Analysis

LLMs are creating a common cognitive substrate: when a few models supply finished articulations at scale, the market of ideas shifts from a heavy-tailed distribution of idiosyncratic signals to a narrower, high-frequency mode. Practically, that compresses the variance in content quality and distinctiveness—I would expect measurable declines in click-through dispersion, A/B test lift, and virality tails over the next 6–24 months, which reduces marginal ROI for generic content producers and SEO arbitrageurs.

That compression creates a two-tier market: suppliers of compute, inference infrastructure, and fine-tuning pipelines (NVIDIA, major cloud providers) capture steady, sticky demand; platforms that can credibly certify provenance or monetize human-authored scarcity (subscription news, exclusive podcast platforms, creator platforms) gain a pricing premium. Conversely, ad-reliant aggregators and templated content marketplaces face margin compression and higher churn as their products become fungible with AI output, pressuring CPMs and freelance pricing within 12 months.

Key catalysts that could accelerate or reverse these dynamics are regulatory mandates for AI provenance/disclosure (EU/US trajectories, 6–24 months), major copyright or dataset-litigation outcomes that force retraining or narrower data coverage, and emergence of style-preservation tooling that preserves human distinctiveness. Near-term trade performance will pivot on enforcement timelines and a behavioral backlash that could either entrench homogenization or restore a premium for verified human voices; position sizing should reflect that binary risk with asymmetric option structures where possible.

Key Decisions for Investors

• Long NVDA (buy shares or a 6–12 month call spread). Rationale: continued GPU/HBM demand for inference and fine-tuning; target +25–35% in 6–12 months if AI adoption continues; downside -30–40% if regulatory constraints or cyclical GPU destocking occur. Position size: 2–4% NAV, use calls to cap downside if preferred.
• Pair trade: Long NYT (NYT) / Short BuzzFeed (BZFD) over 9–15 months. Rationale: paywalled, provenance-heavy journalism should capture higher willingness-to-pay vs ad-dependent, commoditized content. Target: NYT +15–25%, BZFD -30–50% if homogenization compresses ad CPMs. Size: 1–2% NAV net directional, hedge with equal notional dollar exposure.
• Buy MSFT 12–18 month call spread (buy lower-strike, sell higher-strike) to play enterprise LLM monetization. Rationale: OpenAI tie-up + Azure capture of fine-tuning/inference revenue; structured spread limits capital while keeping 2:1–3:1 upside on premium. Time horizon: 9–18 months around product launches/regulatory clarity.
• Long Veritone (VERI) or other public AI provenance/authentication vendors (select small-cap) with 12–24 month horizon. Rationale: regulation and enterprise demand for provenance/authenticity tools could drive >2–3x upside if mandates appear; high idiosyncratic risk—use <1% NAV per name and set disciplined stop-losses.