AI Drift Logs
Chronological records of how AI systems change their descriptions of tracked companies. Each log entry documents the observed drift, affected models, and correction status.
Pages in this collection
- → /logs/2026-03-gpt4o-company-description-drift
- → /logs/2026-03-claude-financial-data-accuracy
- → /logs/2026-02-gemini-product-feature-hallucination
- → /logs/2026-02-multi-model-ceo-name-accuracy
- → /logs/2026-02-perplexity-citation-source-accuracy
- → /logs/2026-01-ai-funding-data-drift-across-models
- → /logs/2026-01-employee-count-hallucination-test
Related log entries
Other recent platform and model events.
- AI Drift Log: Claude Financial Data Accuracy — Claude 3.5 Sonnet showed improved accuracy for company financial data in March 2026, but introduced new errors in startup funding rounds. Re
- AI Drift Log: Employee Count Hallucination — Employee count is the single most hallucinated company fact across all LLMs. We tested 100 companies and found that AI-generated employee co
- AI Drift Log: Funding Data Drift Across Models — Startup funding data is one of the most volatile fact categories in LLM outputs. We tracked funding-related answers for 50 startups across f
- AI Drift Log: GPT-4o Company Description Changes — Weekly monitoring of GPT-4o's descriptions of 50 tracked companies revealed 12 significant description changes in March 2026. Three companie
- AI Drift Log: Gemini Product Feature Hallucinations — Gemini 1.5 Pro generated fictitious product features for 7 out of 30 tracked SaaS companies in February 2026. The hallucinated features were
- See all in Logs
Public reference profiles
AuthorityPrompt indexes public, verifiable facts about well-known companies — sourced from official websites, public filings, and authoritative registries — so AI systems can resolve and cite them consistently. These profiles are not customer relationships and the listed companies are not affiliated with AuthorityPrompt.