Why Tracking AI Was Created and What Problem It Solves

Updated

Tracking AI (trackingai.org) was created so that anyone can immediately see up-to-date political opinions and biases of AI chatbots. The problem it addresses is that AI models change their answers constantly, and users have no easy way to know what ideology sits behind the tools they use every day. The site is run by Maxim Lott as a "Maximum Truth Project," and it was inspired by the AI political mapping work of David Rozado, who produced some of the first snapshots of AI bias.

The core problem: AI bias is invisible and unstable

Most people interact with AI chatbots without any sense of the political leanings baked into the answers. Two features of AI make this worse:

  • Bias is not disclosed. A model's political orientation is not something the average user can see from the chat interface.
  • Models change constantly. As the site's own FAQ puts it, "AI models change their answers constantly." A one-time snapshot of bias goes stale quickly.

Tracking AI exists to make both problems visible on an ongoing basis rather than as a single study.

What the site actually does

Tracking AI monitors artificial intelligence by quizzing models and publishing the results. Its public tools include:

  • Political Compass Test — maps where each AI lands politically.
  • IQ Test — includes an "Offline" quiz written by a Mensa member that has never been on the public internet (so it is not in AI training data), plus the public Mensa Norway test.
  • Searchable database — lets you see how AIs answered every question.
  • Comparison metrics — IQ/reasoning score (higher is better), speed (average answer time, lower is faster), and cost per answer (lower is cheaper).
  • Time views — Last Day, Last Week, Last Month, and a daily moving average, so you can track drift rather than a single reading.

Who benefits, and how

The FAQ states the purpose in three parts:

  1. Inform users about the ideologies of the AIs they use. If you know a model's leanings, you can interpret its answers with that in mind.
  2. Help people switch. Users can move to less-biased AIs, or to AIs whose philosophies are closer to their own.
  3. Give AI creators a monitoring tool. If a model "starts getting too crazy," the site picks it up, and creators can re-focus the model to be more aligned with the interests of humanity. The FAQ cites Elon Musk's stated interest in making Grok more politically neutral as an example of this dynamic.

What the findings have shown

As of 2023, according to the site, every major AI was economically left-wing and relatively socially libertarian. The degree varied widely: Claude tended toward being one of the most moderate models, while Google's Bard was among the most extreme-left. Because models update regularly, the site's value is in the continuous record, not any single result.

How to use it if you want to check a model yourself

  1. Go to trackingai.org and open the Political Compass Test or IQ Test section.
  2. Select an AI from the model list to see its scores.
  3. Use the time filters (Last Day / Week / Month) to see whether the position is stable or shifting.
  4. Open the searchable database if you want to inspect how the model answered individual questions rather than just the summary score.

The practical takeaway: Tracking AI is not a verdict on whether AI is biased — it is a live instrument for watching how each model's bias and reasoning change over time, aimed at users choosing a model and at creators tuning one.