Using Generative Content in Experiences

Generative Experiences enables you to add real-time content generation to your Dynamic Yield campaigns. You continue to use the same audiences, targeting rules, experiences, and variations. But instead of preparing every content value in advance, you can define a prompt and let Dynamic Yield generate selected variables for each shopper using real-time context.

Dynamic Yield uses the campaign’s existing targeting and audience logic to select the appropriate experience and variation. This selection does not change. Only the designated content variables within the selected variation are generated in real time. Delivery, tracking, and measurement continue through the existing campaign workflow.

Why use Generative Experiences?

• Move from segment-based messages toward individualized content.
• Adapt content to each shopper's current intent and context.
• Reduce the effort required to create and maintain multiple variations.
• Use AI within existing Dynamic Yield delivery and measurement workflows.

→ See some common use cases

Configure generative content

Configure generative content directly in the experience editor:

1. Enable Generative Content for the variable group. 
   → Each variation can contain one generative variable group.
2. Write a prompt that clearly describes the content to generate.
   → Up to 800 characters
3. In Write in, select the language for the generated output. See Select the language for details.
4. In the Enrich with area, add context signals or evaluators to use for the generated content. 
5. Define the output variables that the AI should populate and add an example value for each variable.
6. Preview and save the variation.

Select the output language

Select the language for the generated content directly in the generative variable group. This provides predictable language output for both multi-language and single-language experiences.

• For multi-language sections, select Based on page context (default) to use the page language, or select a specific language.
• For single-language sections, select the required language.
• A selected language is always used for generation, regardless of the page context.

Enrich generated content

You can use various signals to generate personalized content. These can be based on context or evaluators.

Context signals 

Available context signals include: 

• Shopper affinities
• Recently viewed products
• Recently purchased products
• Items in the cart
• Current page and product details
• Shopper locale

Evaluators

Evaluators can add real-time page data to the prompt, extending personalization beyond shopper and product feed data. Use cases based on content available on the current page might include top reviews or the shopper's search query.

1. Create or edit the evaluator using the existing evaluator flow.
2. Enable your evaluator for generative content and add a short description of the data it provides.
3. Publish the evaluator, then select it from the enrichment picker in the variation editor. 

Generated content example: Personalized category labels

A homepage can use a generative variable group to create category labels for each shopper.

Different shoppers can receive different values from the same campaign configuration, based on the selected context.

Preview and validate

Preview as a test shopper

Use Preview as to test and refine generative content before saving or publishing the variation. Describe a test shopper in natural language, and Dynamic Yield creates simulated context for the selected enrichments to generate a preview.

This helps you validate the prompt, selected variables, enrichments, and language before launch. The preview uses simulated data for testing and does not represent a real shopper.

1. Select Preview as in the editor.

2. Describe a test shopper, such as “Fashion-forward college student”. 

   

3. Click Generate Preview and review the resulting content.
4. Update the prompt and run it again without saving as many times as necessary.

Preview on site

Use the variation Preview on site to validate the placement and structure of the generated variables before publishing.

Save

When you're done previewing and adjusting, save the variation. 

Common use cases

Dynamic Yield provides a selection of out-of-the-box templates for use in your Generative Experiences campaigns. You can identify them by the star label:

Content safeguards

Generated content is instructed to use only the configured prompt and available context. It should not invent unsupported product details or expose internal identifiers or personal information.

When personalization signals are unavailable, the experience uses the available product and page context to generate a relevant, more general response.

Monitor generated content

After launch, use the Generative Content Logs report to validate outputs, investigate unexpected results, and improve prompts over time. Log data is retained for 7 days. 

Go to Settings › Generative Experiences Logs to find the report, and do any of the following: 

• Review the prompt, selected context, and generated values.
  
• Filter by campaign, experience, variation, date, enrichment type, and cache status.
  
• Identify whether a response was newly generated or served from cache.
• Review readable inputs and outputs instead of raw JSON.

Credits and usage controls

Generative Experiences uses a credit-based pricing model. One credit represents one generative impression, providing a stable usage unit without exposing model-specific token consumption.

Monitor usage

Access the Credit Usage report from the Settings menu. In the report you can see:

• Credits used
• LLM requests
• Cached requests
• Usage by campaign and variation

Repeated requests can be served from cache, helping reduce unnecessary generation and credit consumption. Commercial allocations, overage rates, and billing terms are defined in your agreement. Contact your account team for assistance.

Set a monthly credit limit

You can configure a monthly limit for each section. Note the following:

• Only users with Admin permissions can modify the limit. Others users can view it.
• The limit resets on the first day of each month.