Introducing BrandGPT - Our BrandGuard Interface

A fine-tuned model to chat with brands.

We recently announced BrandGuard, our ensemble of models that check the outputs of generative AI to ensure the outputs are high quality, error free, brand safe, and on brand. Today we are announcing a new tool called BrandGPT, which is a chat interface designed to answer questions about brands and branding topics. BrandGPT is your brand’s representative for your partners, marketing staff, or customers.  Think of it as your brand’s style guide brought to life. You can present BrandGPT with ads or content and determine if the content is on-brand or off and why.

BrandGPT was trained on thousands of style guides and brand examples and is a fine-tuned model on top of GPT-4. Here are some examples of what it can do.

In the example above, BrandGPT is answering questions about Ben and Jerry’s brand identity. This ties into our BrandGuard models and ensures that, were Ben and Jerry’s using generative AI for text or imagery, these brand guidelines would be followed, and only approved outputs would be generated.

Below is another example, where I’ve asked BrandGPT about the brand ambassadors for Puma.

We built this interface because we’ve become increasingly concerned with brand safety and AI. BrandGPT has the ability to capture chat interactions with brands, and we hope brands can experiment and learn from those interactions.

If you are a brand manager and are interested in a free monthly report of what questions people ask about your brand, you can sign up for that here.