See how AI models censor and bias.
GPTfake independently monitors censorship, bias, and transparency across ChatGPT, Claude, Gemini, and other LLMs — with a reproducible methodology and open data for researchers, journalists, and developers.
- Independent
- No vendor funding
- Reproducible
- Open methodology
- Multi-model
- Side-by-side
Independent & transparent
No vendor funds us, no model gets a pass. Every censorship and bias finding is reproducible, sourced, and published in the open.
Cross-model coverage
ChatGPT, Claude, Gemini, Mistral, Qwen and more — tracked side by side with one consistent methodology so results are comparable.
Censorship monitoring
We probe how each model handles sensitive topics over time and flag refusals, deflections, and silent policy shifts as they happen.
Bias detection
Structured tests surface political, cultural, and demographic skew — measured with documented metrics, not vibes.
Longitudinal research
Model behavior drifts. We re-run the same prompts across releases to chart how transparency and guardrails change over time.
Open data & API
Pull our monitoring results programmatically. Datasets, methodology, and an API built for researchers and journalists.
GPTfake