Goals and Methodologies

Understanding our approach to measuring LLM sentiment and its impact on human market behavior

Our Goals

Four core objectives drive our research into LLM market sentiment

01

UNDERSTAND LLM SENTIMENT INFLUENCE

Our primary goal is to understand LLM sentiment patterns and how they influence human behavior in financial markets. By analyzing systematic biases in AI responses, we can better predict how these recommendations might shape investment decisions across millions of users.

02

FOCUS ON HIGH-IMPACT MODELS

We follow the 5-10 most popular LLM models by user count, focusing on those with the potential to impact the greatest number of humans. This includes top companies being used by various applications in the USA and China, ensuring our data reflects the models that matter most to market sentiment.

03

SURFACE LATENT SENTIMENTS

We aim to uncover the hidden biases and preferences embedded within LLM training data and response patterns. These latent sentiments often reveal systematic preferences for certain assets, companies, or market sectors that may not be immediately apparent but can significantly influence market dynamics.

04

MAXIMIZE FREE ACCESS TO INSIGHTS

We are committed to providing as much free data, insights, and access as possible to democratize understanding of LLM market influence. However, running queries against multiple models is costly, and some advanced features and real-time data will necessarily be limited to paying customers to sustain the platform's operations.

Our Methodologies

Systematic approaches to gathering and analyzing LLM sentiment data

STANDARDIZED QUERY PROTOCOL

Periodic Querying

We periodically query all LLMs we are following with identical prompts to ensure consistency and comparability across models. This systematic approach allows us to track sentiment changes over time and identify patterns that might indicate market influence.

Query Categories

Our standardized queries focus on several key areas:

  • Predicted appreciation and depreciation of various securities
  • Asset rankings and comparative analyses
  • Investment recommendations across different time horizons
  • Sector and thematic investment preferences

Data Consistency

By using exact same prompts across all models, we eliminate variables that could skew results and ensure that any differences in responses reflect genuine model biases rather than inconsistent questioning approaches.

Research Purpose Only

Our methodologies are designed for academic and research purposes to better understand AI influence on market sentiment. This data should not be used as the sole basis for investment decisions.