PressPlay uses different generative AI paired with the Research Methods in order to create assets that will increase conversion rate.
Instead of manually designing each variant, you provide parameters and let the AI generate optimized assets based on proven app store optimization principles and best practices.
PressPlay's AI can generate the following asset types:
App Icons (IC): 1024x1024px PNG icons with various design styles
Feature Graphics (FG): 1024x500px landscape graphics for store listing headers
Short Descriptions (SD): Promotional taglines up to 80 characters
Long Descriptions (LD): Full descriptions up to 4,000 characters with keyword optimization
Not supported: Screenshots, Promo Videos, and Multi-Asset experiments require manual upload. See Manual Image Upload for details.
The AI experiment generator follows a sophisticated process:
Analysis: Reviews your current store listing assets and metadata
Pattern Recognition: Identifies successful patterns from similar apps and categories
Variant Generation: Creates multiple variations optimized for different testing hypotheses
Quality Assurance: Ensures all generated assets meet Google Play requirements
Visit the Settings > AI Workspace
Guarantee that each Research Methods you pretend to use are complete with the correct information
Navigate to your app's dashboard
Click the "Create" tab in the main navigation
Select "AI-Generated Experiments"
Configure what you want the AI to generate:
Asset Types: Choose which assets to generate (icons, feature graphics, short descriptions)
Target Locales: Select one or more markets for generation
Number of Variants: How many variations per asset type (typically 3-6-9)
Research Methods: Select which method, or methods, you want to use as base for the asset creation
Model: Select from the dropdown the best-suited model
For Specific Art Style or Character alignment use models like Flux-Kontext Pro/ Qwen /Seededit-3.0. Once one of those methods is selected it will open a new field called Reference Image per default it will present your default Icon/Feature Graphic. If you want to iterate on this character select “character” on the Asset Focus section, if you want to follow only the style than select Style. You can always upload a different reference image, for example a new character a new splash-art.
Specific Instructions: Optional guidance on design direction or tone
Click "Generate Experiments" to start the AI process
Wait 30-60 seconds while the AI creates variants
Review all generated experiments in the review queue
The AI creates separate experiments for each variant, ready for approval
The AI generates icon variations that:
Test different visual styles and compositions
Explore color palette variations
Experiment with symbol vs. text-based designs
Maintain brand recognition while testing new approaches
Generated feature graphics include:
Different layout compositions and focal points
Various text placement and messaging approaches
Alternative color schemes and visual treatments
Different calls-to-action and benefit highlights
AI-generated descriptions test:
Different hooks and opening lines
Feature-focused vs. benefit-focused messaging
Various tones of voice (professional, casual, enthusiastic)
Different value propositions and key messages
After generation, all AI experiments enter the review queue:
Review Generated Assets: Examine each variant carefully
Check Quality: Ensure assets meet your standards and brand guidelines
Edit if Needed: Modify hypotheses, settings, or assets before approval
Approve or Reject: Accept variants you want to test, reject those that don't fit
Organize by Priority: Set execution order for approved experiments
Speed: Create multiple experiments in seconds instead of hours
Variety: AI explores design directions you might not consider
Data-Driven: Based on patterns from successful apps
Consistency: All variants meet technical requirements automatically
Scalability: Generate experiments for multiple locales simultaneously
Provide clear context: Better input on Research Methods leads to better generated variants
Review everything: AI is powerful but still requires human oversight to guarantee the assets have the required quality.
Edit freely: Use generated assets as starting points, not final products
Test AI suggestions: The AI might identify opportunities you overlooked
Combine with manual: Mix AI-generated and manually created experiments
After generation, you can modify experiments before deployment:
Edit hypotheses to better reflect your testing goals
Adjust experiment duration and priority
Modify or replace generated assets
Change locale or CSL assignments
Update traffic allocation settings
AI generation is ideal when you:
Need to create many experiments quickly
Want fresh perspectives on store listing optimization
Are exploring different creative directions
Need to test across multiple locales
Want to overcome creative block or writer's block
Brand compliance: Always verify generated assets match brand guidelines
Cultural sensitivity: Review locale-specific generations for cultural appropriateness
Legal review: Ensure generated text complies with advertising regulations
Quality variance: Some generated variants may be stronger than others
Creating Manual Experiments - Alternative hands-on approach
Understanding Asset Types - Comprehensive guide to all 7 asset types
Multi-Locale Testing - Generate experiments for multiple markets
Experiment Review Process - How to review and approve experiments