Visual Recommendations

Instead of only telling you what to change, Neurons shows you what better could look like, using generated visual examples grounded in predicted ad impact.

Here’s a short video to show you how it works:

Watch the walkthrough here.

 

What are Visual Recommendations?

Visual Recommendations is an advanced feature in the Neurons Platform. In short, Visual Recommendations turn the written insights and recommendations into new visual examples.

Visual Recommendations generate inspirational visual mockups that show how specific improvements to an ad could look in practice.

They are based on:

• Your creative
• Neurons’ predictive models of attention, engagement, cognition, and memory
• The insights and recommendations identified in Neurons
• An AI workflow translating the recommendations into actions on your behalf to create and curate examples

Each run produces curated examples that help you move from insight to action faster.

 

How do Visual Recommendations work?

At a high level, Visual Recommendations run the full Creative AI Loop:

1. Predict how people are likely to respond to your original ad
2. Suggest what to improve based on predicted impact
3. Generate visual examples that apply those suggestions
4. Curate only examples that are expected to improve impact

Behind the scenes, the system generates multiple visual interpretations, evaluates them, and only shows you a small number of the strongest, most relevant examples.

You’ll typically see:

• Up to 4 optimized visual examples
• Sometimes a Wildcard

 

What is the Wildcard?

The Wildcard is a concept-aligned visual variation that is not tied to a specific recommendation.

Its purpose is to:

• Break out of your current creative frame
• Introduce ideas outside your usual focus
• Spark discussion and inspiration

You may not always see a Wildcard. It’s only shown when there’s an improvement in impact.

 

How do I use Visual Recommendations in practice?

Getting started

1. Log in to Neurons
2. Open or upload an asset
3. Make sure the asset is fully set up (objectives, AOIs, etc.)
4. Go to the Optimize page
5. Click Generate Visual Recommendations
6. A full run typically takes up to 10 minutes

Saving and sharing

You can:

• Download individual examples
• Compare an example with the original
• Save a variation as a new asset to continue working with it

If you regenerate recommendations, previously generated visuals are replaced unless you’ve saved them.

Best practices, tips & tricks

Learn a few ways to maximize value with Visual Recommendations on our Tips & Tricks page.

 

Data Privacy & Compliance

All assets uploaded to Neurons are handled under strict security and privacy standards.

• Uploaded content is not used to train models
• Data is processed only by approved vendors
• Assets are deleted after short retention periods
• Industry-standard security certifications apply

 

Visual Recommendations Pricing, Credits & Usage

Visual Recommendations is a paid add-on feature to the Neurons Platform.

Visual Recommendations operate on a credit-based system. Your remaining credit balance is visible directly inside Neurons in the "Settings" page under "Usage".

Should you meet the limit please reach out to your Customer Success Manager. They’ll guide you through the available options.

 

Common questions

Why do I sometimes see fewer than 4 examples?
Only examples with improved predicted impact are shown. If fewer qualify, fewer are displayed.

Why do some examples look very similar?
This can happen when the system finds a narrow improvement path. You can regenerate to explore alternative directions.

Why didn’t my Impact Score increase after saving an example?
In some cases, missing or mis-detected Areas of Interest (AOIs) can affect the score. Always review AOIs when saving a new asset.

 

How do Visual Recommendations work behind the scenes?

Visual Recommendations are not created by simply sending your ad to an image generator.

Behind the scenes, Neurons runs a structured process that translates recommendations into visual examples, evaluates them, and only shows you options that are expected to improve impact.

The process starts with the text recommendations in Neurons, which are based on predicted attention, memory, and engagement. These recommendations are translated into a structured format that image models can interpret consistently.

Neurons then generates multiple visual versions based on those recommendations. Each version is automatically checked to confirm that a visible change was made and that it reflects the intended recommendation. Versions that don’t meet these criteria are discarded.

The remaining examples are evaluated using Neurons’ prediction engine. Objectives and Areas of Interest from the original asset are applied, and a new Neurons Impact Score is calculated. Only versions expected to improve predicted impact are considered.

From these, a small set of the strongest and most distinct examples is selected and shown in the interface.

This layered process is what separates Visual Recommendations from generic image generation: the output is guided by predicted human response, filtered for quality, and focused on helping teams make better creative decisions.