HERETIX

Understand the AI answering your questions.

Heretix shows how likely the AI thinks your claim is true—and explains why.

Select models

How Heretix works

Heretix isolates what an AI really believes. We ask your claim in precise ways, measure how confident the model is, and explain what shaped its answer.

Step 1 — Ask

Ask

Type any claim.

Step 2 — Choose a model

Choose a model

Pick which AI model you want to examine as each has its own perspective and training.

Step 3 — Add context

Add context

Decide whether to keep the model in its own world, or let it pull from the internet for new evidence.

Step 4 — Understand

Understand

See how confident the model is in your claim. Inspect the explanation to further understand how the answer was derived.

Example outcome

67.8%

“Do tariffs cause inflation?”
  • Tariffs raise costs of imported goods for businesses and consumers.
  • Higher import costs often pass through to prices when demand stays strong.
  • Exceptions happen when currency shifts or suppliers absorb the tariff shock.

Ask the AI what it really believes

Heretix makes bias visible. Input a claim (maximum 180 characters) and see what the model thinks, how confident the model is in the answer, and what items supported its conclusion

Start a claim

Questions people ask

What does the percentage mean?

This is how confident the AI is in your claim. It’s not absolute truth, it’s a reflection of what the model believes, shaped by its training and bias.

Does Heretix browse the web with my claim?

Your choice. Select "Internet Search" to let the model pull current information from the web to supplement its reasoning. If you only want to see the output of the model's internal training, select "Internal Knowledge Only".

Is this politically biased?

We show the model’s belief, not a judgement from us. Heretix reveals consensus the AI already absorbed, so you can inspect it and judge for yourself