Actively Promoting Active Share in the Age of AI: The Algorithmic Discovery Gap in Investment Trusts

alexanderdenny9
Aug 6
2 min read

By Alexander Denny

Why Traditional Marketing is Failing Closed-Ended Funds in AI Search

As wealth managers, financial advisers, and retail investors transition from traditional search engines to conversational Large Language Model (LLM) interfaces (such as ChatGPT, Gemini, Claude, and Perplexity), a critical commercial obstacle has emerged: The Algorithmic Discovery Gap.

While closed-ended Listed Investment Companies (LICs) and Investment Trusts offer unique operational tools—permanent capital, active gearing, unconstrained micro-cap access, and multi-decade dividend reserve smoothing—our empirical research reveals that conversational AI engines routinely filter out investment trusts in favor of open-ended OEICs and passive ETFs.

Key Findings from Our Multi-Model Benchmark

In our primary research paper, Actively Promoting Active Share in the Age of AI, Deeper Thinking Services benchmarked four leading AI architectures across 15 retail investment scenarios (evaluating 362 total fund entity recommendations):

Severe Parametric Bias: Frontier models demonstrate an overwhelming bias toward open-ended funds. OpenAI’s GPT-4o recorded an extreme Discovery Bias Index (DBI) of 0.86, failing to surface an investment trust as its primary lead recommendation in a single prompt across all 15 scenarios (0% lead primacy).
The Web-Search Fallacy: Live web search integration (Retrieval-Augmented Generation / RAG) does not solve structural bias. Perplexity Sonar Pro recorded a DBI of 0.69, demonstrating that RAG scrapers simply ingest and magnify the overwhelming volume of open-ended marketing collateral and platform SEO listings.
Model Heterogeneity: Not all AI models behave identically. Google’s Gemini Flash achieved near-neutral parity (DBI 0.09), regularly surfacing investment trusts alongside open-ended funds, whereas Anthropic’s Claude Sonnet 5 exhibited a moderate open-ended preference (DBI 0.38).
Illiquidity Recognition: Across all model architectures, AI engines only consistently recommend investment trusts when prompt language explicitly mandates access to private equity, venture capital, or illiquid assets.

Moving from SEO to Generative Engine Optimization (GEO)

Investment trust boards can no longer rely on legacy PDF factsheets and static broker distribution. If a trust’s structural capabilities—such as gearing discipline, revenue reserve cover, and historical performance metrics—remain locked in unindexed PDF attachments, AI engines will default to indexed open-ended alternatives.

To protect liquidity and retain share of voice in an AI-driven discovery landscape, boards must execute Generative Engine Optimization (GEO): structuring digital infrastructure into machine-readable JSON-LD schemas and raw HTML5 so LLMs can extract, verify, and cite closed-ended vehicles.