Onboarding

Help users understand how best to interact with AI by setting the right expectations during their first interaction with your product.

Data consent

When onboarding new AI features, be transparent about data and privacy consent. Users should clearly understand when and how their data may be used, and especially how they will benefit from it.

Ash’s app lets the user know how their information will be used, before they initiate a conversation with their AI.

Superhuman mentions upfront how user input and data will and won't be used.

Provide suggestions

Use suggested prompts or actions to inspire users and make it clear what they can use this product for. Avoid any prompt writer's block by offering a way forward with relevant and hands-on examples.

FigJam provides prompts that show the variety of situations their app is useful in.

Perplexity’s home page shows a revolving carousel of various prompts to help the user get started.

Provide structure

Use templates to make it easier for users to get the results they want without having to type each word themselves. AI models often need a lot of context and a template can help users to provide more context in an easy way.

Elicit helps users get started by through a variety of AI-powered options, each with their own interaction.

UX Pilot helps users create detailed prompt templates through an auto “enhance” feature.

AI Notice

Provide an “AI notice” to help explain the AI’s capabilities, making clear its limitations and what is generated by AI as it can be wrong or misleading. Adding a ‘Preview’ or ‘Experiment’ badge and presenting it with consistent “AI styling” can help signal to users to be wary of results and provide caution.

Anthropic’s Claude app displays an AI notice after logging in.

Mozilla clearly shows how their AI help feature is powered.

Next guideline

Interactions