Editorial Policy

Last updated: March 25, 2026

StockScreener.finance publishes two types of content: algorithmic trading signals and AI-powered news analysis. This page explains how both are produced, what rules govern the AI models, and how we maintain quality.

Trading Signals

All trading signals are generated by a fully automated mathematical pipeline. The system scans 3,000+ NASDAQ-listed stocks daily through seven stages: market regime detection, sector screening, fundamental filtering, technical signal identification, quality scoring, AI review, and risk management.

Every decision in this pipeline is based on formulas, financial metrics, technical indicators, and backtested data. No human judgment is involved in signal generation. The AI review at the final stage is purely advisory — it provides context for the user but does not influence whether a signal is published.

Signals are not investment recommendations. They are algorithmic outputs that users should evaluate independently before making any trading decisions.

AI News Analysis

Every financial news article on this platform goes through the following process:

  1. Source collection. Articles are collected from public financial news sources via RSS feeds and news APIs. The original article is always the source of truth.
  1. AI analysis. Each article is sent to four leading AI models — Claude by Anthropic, Gemini by Google, ChatGPT by OpenAI, and Grok by xAI. Each model independently analyzes the article, assesses its potential market impact, and forms a position (bullish, bearish, or neutral).
  1. AI debate. The models then engage in a structured discussion where they challenge each other's conclusions, defend their positions with arguments, and may change their stance when presented with stronger evidence.
  1. Publication. The original article summary, individual model analyses, and the full debate are published together.

Rules for AI Models

All four AI models operate under the same prompt and the same rules, with no exceptions:

  • One prompt for all. Every model receives identical instructions. No model gets special treatment, additional context, or modified framing. This ensures fair comparison and prevents bias toward any particular model's strengths.
  • Source fidelity. Models must base their analysis on facts present in the original article. If a model cannot verify a claim or does not have reliable information about a topic, it is forbidden from using unverified statements as arguments.
  • No hallucination. Models are explicitly instructed not to fabricate facts, statistics, quotes, or events. If a model is uncertain about something, it must acknowledge the uncertainty rather than invent supporting evidence.
  • No editorial direction. There are no topics, companies, sectors, or narratives where models are asked to lean in any particular direction. The analysis is driven entirely by the content of the original article.
  • Transparency. Every model's full analysis and reasoning is published. Nothing is edited, filtered, or cherry-picked.

Human Oversight

StockScreener is a one-person project built and maintained by Maksym Misichenko. Human oversight includes:

  • Designing and maintaining the signal generation algorithm
  • Writing and updating prompts for AI models
  • Monitoring translation quality across 16 languages
  • Reviewing system performance and fixing errors
  • Maintaining data pipeline integrity

The AI-generated content (model analyses and debates) is published automatically without human editing of individual articles. The human role is in system design, prompt engineering, and quality assurance at the platform level — not at the individual article level.

Data Sources

Financial data is obtained from publicly available sources through APIs and data feeds. News articles are collected from established financial media outlets. We do not guarantee the accuracy or completeness of upstream data and recommend users verify information through sources they trust.

Corrections and Feedback

If you find an error in any content on this platform, please contact us at [email protected]. We take accuracy seriously and will investigate and correct confirmed errors.

Languages

All content is available in 16 languages. Translations are produced using AI language models. While we monitor translation quality, slight variations in nuance may occur across languages. The English version is the authoritative source.

---