Below I describe how to set up the fully integrated AI in JASP. All of the individual steps are straightforward. If you get lost or if anything is unclear, you can simply ask your favorite LLM for advice, and it will help you out. Let’s go!

Step 1: Activate the AI in the JASP Preferences Menu

Open JASP. From the the blue “hamburger” menu icon  in the top-left (i.e., the three horizontal lines) go to “Preferences” and then to “AI”:

At this point you will be asked whether or not you want to activate the AI. Those users who have developed an allergy to all things that reek of AI: rest assured, you can simply keep using JASP without activating the AI. Other users may be curious to see the AI in action and they can confirm their choice. A pop-up screen presents the usual warnings that come with any AI application. Moving on gets us to the next step: the AI settings.

Step 2: Select your Provider & Model

Under the AI Settings you have to indicate a provider and a model. In general, AI costs energy, “tokens”, and hence money. Most providers and models require that you pay for the work that the AI is doing. Once you pay, you can obtain an “API key” that is needed for JASP to get the show on the road. Better models are usually more expensive. However, Gemini 3.1 flash-lite can be run without any payment whatsoever. As you see from the screenshot below, I have selected “Google Gemini” as the provider and “Gemini 3.1 Flash-Lite” as the model: [NB. Thomas Langkamp just attended me to the fact that there are many more free LLMs around, and some are better than Gemini 3.1 flash-lite. We might do a blog post about these other options in the near future!] 

All we need now is that API key. This brings us to the next step.

Step 3: Obtaining the API Key

How you obtain an API key depends on the provider (when in doubt, ask your LLM). For Gemini, we have to go to Google AI Studio and log in with your Google account. As you can see from the screenshot below, on the bottom-left of the screen, just to the top-right of your name, there sits the icon of a key. Click that icon:

Clicking the icon of the key opens a new screen, and under “key” you will find “Default Gemini API Key”, as shown in the screenshot below:

Clicking the blue string with symbols shows you the entire key, which you can then copy. With the API key now in your possession, we can move on to the next step.

Step 4: Entering the API Key

Return to the JASP application and paste the API key into the “API Key” field. You can then press “Test Connection”. If it worked, you should see the green text “Connection successful”, as shown below:

Step 5: Use the AI in JASP!

Your AI is now hooked up, and you can use it to assist your work. The screenshot below provides a rough overview of the AI functionality that is available by default. On the bottom-right there is the icon of Alfred (“A”), our default AI assistant. Clicking Alfred brings up the JASP AI Chatbox (“B”) that allows users to communicate with Alfred: ask for advice, give assignments, etc. In the input panel to the left, you will also see a small version of Alfred (“C”), next to the “R” icon. Clicking the Alfred icon executes an “annotation prompt”, where Alfred is asked to provide explanations and interpretations for each element of the output. Future posts will elaborate on the ways in which AI can support statistical work.

A Word of Warning

Selecting a particular AI model to run in JASP is not a trivial endeavor. Yes, Gemini 3.1 flash-lite is free, but it is not the strongest model around. We have experimented the most with DeepSeek; this model is relatively cheap and performs well. However, most of these LLMs may use your data and interactions to train their network; consequently, these models are not suited for the analysis of sensitive data — for instance, patient data or company data. If you want to use the JASP AI and keep your data safe as well, you have to use a provider that guarantees the data will not be used for model training.  In order to set things up properly, we recommend that teams and institutions contact JASP Services B.V. for full support.
NB: AI can make mistakes. We recommend verifying AI advice for important decisions.

Authors

Eric-Jan Wagenmakers (CEO) and the JASP Services BV Team.