Overview
Content Intelligence tracks the performance of your creatives across channels and uses AI to generate actionable insights and improvement suggestions. It identifies which subject lines, copy patterns, and creative variants drive the best engagement — and recommends specific changes to underperforming content.
Navigate to AI > Content Intelligence in the sidebar.
Content Intelligence automatically tracks creative performance using data from the Respond API. Every time you record an outcome (impression, click, conversion, dismissal), the system aggregates metrics per creative and per channel.
Tracked metrics include:
| Metric | Description |
|---|
| Impressions | Number of times the creative was shown |
| Click rate | Percentage of impressions that resulted in a click |
| Conversion rate | Percentage of impressions that resulted in a conversion |
| Dismiss rate | Percentage of impressions that were dismissed |
| Revenue per impression | Average revenue generated per impression |
Performance data begins accumulating as soon as you start recording outcomes via the Respond API. No additional configuration is required.
Insights Generation
Click Generate Insights to run an AI analysis of your creative performance data. The AI examines patterns across all your creatives and channels to produce insights.
Channel-Level Analysis
The AI compares performance across channels and identifies:
- Which channels produce the highest engagement for each creative type
- Whether certain copy styles perform better on specific channels (e.g., short copy for push notifications, detailed copy for email)
- Channel-specific trends over time (improving, declining, or stable performance)
Individual Creative Analysis
For each creative, the AI evaluates:
- Performance vs. peers — How it compares to other creatives in the same category
- Trend direction — Whether performance is improving, declining, or flat
- Strengths — What aspects of the creative are working well
- Weaknesses — What aspects could be improved
Subject Line Patterns
For email and push notification creatives, the AI analyzes subject line patterns:
- Which subject line structures (questions, urgency, personalization) produce the highest open rates
- Optimal subject line length for your audience
- Words and phrases that correlate with higher or lower engagement
Rewrite Suggestions
When the AI identifies underperforming creatives, it generates specific rewrite suggestions:
- Original text — The current copy
- Suggested rewrite — An improved version based on patterns from your top-performing creatives
- Rationale — Why the change is expected to improve performance
- Expected improvement — Estimated lift based on similar pattern changes in your data
Applying Rewrites
When you click Apply on a rewrite suggestion:
- A new creative variant is created as a draft, linked to the original creative
- The variant contains the suggested copy changes
- You can review, edit, and activate the variant
- Optionally, set up an A/B experiment to test the variant against the original
Use rewrite suggestions as starting points rather than final copy. The AI bases suggestions on engagement patterns, but your brand voice and compliance requirements should always inform the final version.
Best Practices
- Accumulate data first — Wait until creatives have at least 100 impressions before generating insights. Small sample sizes produce unreliable analysis.
- Run insights regularly — Creative performance shifts over time. Generate fresh insights weekly or after major campaign changes.
- Test before replacing — Use A/B experiments to validate rewrite suggestions before fully replacing existing creatives.
- Segment by channel — Channel-level insights are often more actionable than aggregate insights. Focus on one channel at a time.
Advanced Parameters
Each content analysis run can be fine-tuned using the Advanced Parameters panel. Expand it to adjust:
| Parameter | Default | Description |
|---|
| Min Impressions | 100 | Times content must be shown before judging performance |
| Metric Weights (CTR / CVR / Revenue) | 0.33 / 0.34 / 0.33 | How much weight to give each metric when scoring |
| Confidence Level | 0.95 | How sure we need to be before recommending a change |
Large Dataset Warning
When the creative performance dataset contains 5,000 or more rows, a confirmation dialog appears before analysis begins. The dialog shows:
- Accuracy comparison — ML Worker uses TF-IDF and Random Forest feature importance on the full dataset vs. LLM heuristics
- Estimated cost — Token count and approximate cost if proceeding with LLM
- Speed comparison — ML Worker processes locally in seconds vs. LLM round-trip
You can choose Use ML Worker (recommended for large datasets) or Proceed with LLM (uses sampled data).
For large creative datasets, the ML Worker identifies statistically significant content patterns that LLM sampling may miss. See ML Worker Setup for deployment instructions. Next Steps
