You and your chat assistant

Add AIQT to your assistant.

Your assistant sounds just as confident when it is guessing as when it knows. AIQT 1.0.0 fixes that, and it is available now. Pick the AI you use below. You do not need a command line, a developer account, or any knowledge of GitHub.

Pick your assistant

Choose the AI you use

The same skill works the same way in every assistant that accepts one.

While you are here

We also recommend Clean Language

AIQT keeps your assistant honest. Clean Language keeps its writing clean: a free editorial standard, from the same family, that strips out the habits that make AI prose grating, the robotic tone, the filler, the fake enthusiasm, and the predictable vocabulary. It installs the same easy way, one skill you add to your assistant, and we recommend adding it alongside AIQT. Visit cleanlanguage.ai to install it.

Step one

Download the AIQT Skill

Everything is in one file. Download it and keep it as it is.

Your browser saves a file named aiqt-skill.zip. Do not unzip it. The Skill is prose instructions and reference material only: it contains no executable code and makes no network calls. You cannot break anything with it.

Step two

Add it to your assistant

Claude (the recommended method)

  1. Turn on code execution first. The Skills section does not appear without it. In Claude, open Settings, then Capabilities, and enable code execution and file creation. Install from a desktop browser; the mobile apps are not documented to support Skill upload.
  2. Open your Skills, and create one. Go to Claude's Skills page, choose to create a new Skill, and pick "upload".
  3. Upload the file. Select the aiqt-skill.zip you just downloaded, then make sure AIQT is switched on in your Skills list.
  4. That is it. Start a chat and every conversation is held to the standard. You can switch the Skill off at any time.

ChatGPT, Copilot, and other assistants

For an assistant that supports its own Skills or a custom agent, add the downloaded Skill the same way that assistant adds any other. For anything else, use the portable instruction file below: open it, copy it, and paste it in wherever your assistant lets you set its instructions.

What you get

What changes once it is on

The irritations that brought you to this page are exactly what stops.

No more confident bluffing

Every factual claim points back to a source, and when a guardrail catches something, your assistant names the guardrail and what it caught.

No more guessing to please you

When your assistant does not know, it says so. When a decision is yours, it pauses and asks, instead of filling the gap with a plausible invention.

No more "done" that was not

A "done" rests on a check that actually ran, so every finished claim comes with a check you can point to.

No more quiet changes

No weakened checks, no half-built work dressed up as finished, no invented names or citations. Failures are surfaced to you first.

Beyond chat

Writing code?

The development assistant (1.1.0, in development) brings the same standard into your repository: you give your coding assistant a short setup prompt, it fetches the pack from GitHub, and a wizard walks you through the rest. A team can also share one standard across every assistant it uses, on the Teams & Enterprise path.