AI Experiments make it easy to scale your experimentation by automatically generating and launching optimized variations across multiple web pages, all in just a few clicks. Instead of manually creating each experiment, you can quickly test AI-generated variations focused on improving key elements like product descriptions or marketing copy.
Our AI is designed to understand your website’s context to deliver relevant, brand-aligned suggestions. For example, a luxury retailer will receive different recommendations than a fast fashion brand, ensuring each variation fits your audience and tone.
User benefits
- Save time by launching experiments on multiple pages simultaneously.
- Rely on AI-generated copy improvements focused on marketing and engagement.
- Quickly review, validate, and push changes live, without needing to build each experiment manually.
Workflow overview
Start by defining your optimization’s scope—like product pages or a specific category—by setting URL conditions. Our AI then analyzes the selected pages and generates tailored, text-based suggestions for each one.
Once you review the suggestions, you can finalize the experiment directly in Kameleoon. All changes are automatically grouped into a single variation and applied across the relevant pages. You stay fully in control, with the ability to adjust targeting, simulate the variation, and track results, just like any standard Kameleoon experiment.
Prerequisites
Ensure Kameleoon’s bot can access your website to perform screenshot analysis. If access is restricted, you can whitelist the bot using one of the following methods:
- Whitelist User-agent: Kameleoon-Visual-Inspector/1.0
- Whitelist IPs:
- 65.21.88.143
- 65.21.88.142
- 65.21.88.141
- 65.21.88.140
- 65.21.88.139
- 65.21.20.21
- 65.21.20.20
Availability and limitations
- Available only for Web Experimentation projects.
- You can run as many AI Experiments as you want — no limit.
- Each variation will include all selected suggestions from a specific analysis.
- The AI suggests improvements to your existing text, keeping the length about the same.
- Works for websites in multiple languages, as long as each language has its own separate URL.
- The code automatically adapts for both desktop and mobile devices.
- For Single Page Applications (SPAs), the AI will only suggest changes for the landing screen of the specified URL.
Create an AI Experiment
1. Define your experiment scope and customize your suggestions
Navigate to AI Copilot > AI Suggestions in your Kameleoon dashboard and select the website you want to optimize.
Refine your scope by setting URL filters:
- Contains: Target pages containing specific keywords (for example, “clothing”).
- Regex: Use Regular Expressions for more precise page targeting.
- CustomData: (Coming soon)
Note: We recommend you refine the scope of your analysis to similar types of pages.
Add specific instructions to guide the AI in its suggestions, depending on your objectives. For example:
- “Add seasonal context based on [Easter/Christmas/Thanksgiving] promotions.”
- “Focus on sustainability benefits for eco-conscious shoppers.”
- “Highlight urgency with copy like ‘Only a few left in stock’.”
- “Adapt tone to match playful brand voice.”
2. Review and validate suggestions
After the AI analyzes your pages, you’ll receive a set of suggestions.
- Review each suggestion and preview visual changes.
- Confirm that the suggestions align with your optimization goals.
- If a specific suggestion doesn’t fit your needs, you can exclude it from the experiment individually before launching.
3. Launch your experiment
Once you’re satisfied with the AI suggestions:
- Click Create Experiment to proceed to the finalization steps.
- By default, your targeting will include all pages that received at least one suggestion.
- You cannot individually edit suggestions due to the scale of AI Experiments.
- Simulate your experiment to verify changes. The simulation mode will automatically select one page as a representative sample.
Finalizing your experiment
Click Edit Campaign to finalize your AI Experiment. From here, the steps are the same as any other experiment you’ve created.
Complete your experiment’s configuration by:
- Refining targeting: add visitor segments, device types, or specific user conditions.
- Set your traffic distribution, goals, and integration preferences.
- Confirm and launch your experiment. Variations will automatically deploy across all selected pages.
Configuration options per project
Admins can configure project-specific settings including:
- Analysis Period: Specify how many past days should be included when retrieving and filtering target URLs
- URL Cleaning: Define parameters (for example, product model or color) that should be retained during optimization to ensure unique page variations are preserved.
- Blocked Areas: Exclude specific website sections from AI suggestions using CSS selector.
- Pre-treatment script: Use JavaScript (via the Kameleoon API) to prepare the page for AI analysis (for example, open accordions, close pop-ups, or reveal hidden content) so elements are in the correct state before suggestions are generated.
These settings are accessible via the dedicated settings area from the AI suggestion page.
Analysis and results
Monitor your experiments and analyze performance as usual, comparing AI-generated variations against original content to measure effectiveness and conversion impacts.
Upcoming features
Kameleoon continues to enhance AI Experiments with planned features:
- Suggestion Selection: Select specific AI suggestions to include.
- Tone Customization: Adjust the AI-generated copy’s tone (formal, casual, playful).
