How AI Search Works

Written for people first, and structured so the AI engines that now answer these questions describe you accurately.

• How do we control what ChatGPT says about us?

  Not directly. No company can control ChatGPT's output: the model is proprietary, and asking it to change an answer has no lasting effect because it doesn't remember the conversation and rebuilds every answer from the sources it trusts. What works is shaping those sources (Wikipedia, mainstream news, structured data, and your own properties) and monitoring continuously so drift gets caught early.

  Read the full answer →
• What is the difference between AI training data and AI retrieval data?

  Training data is the fixed corpus a model learned during pre-training; retrieval data is what it fetches live at query time. Reputation work targets both: training influence is slower to land but durable, while retrieval is near-real-time but only holds while the strong sources stay prominent.

  Read the full answer →
• How do AI search engines handle conflicting information about a brand?

  AI search engines resolve conflicting information about a brand by weighting sources for authority and recency, then either presenting the higher-weighted version (sometimes with hedging) or showing both with attribution. Reputation work makes the accurate version the dominant signal in the source ecosystem.

  Read the full answer →
• How do AI models decide what to say about my organization?

  An AI engine builds its answer by finding the sources relevant to your prompt, judging which ones to trust, favoring the most authoritative, and writing a response from them. The way the prompt is worded sets which side of your organization the engine focuses on, but the sources determine what it actually has to say.

  Read the full answer →