Ad Testing

Ad Testing compares model-generated reactions to headlines, copy, images, and calls to action across AudiAInce's structured AI audience archetypes. Use the output to refine creative and choose candidates for real-customer or live-market testing.

Ad Testing Features

Quick Test

A single-ad model review can show:

• model-generated appeal and clarity scores
• call-to-action reactions
• differences across defined archetypes
• objections and improvement suggestions

Tournament

The tournament workflow compares multiple variants inside the simulation:

• variants compete in configured rounds
• model-generated rankings are recorded
• archetype-level differences can be reviewed
• new candidate variants may be generated

A tournament winner is the model's leading hypothesis, not proof of better campaign performance.

Simulation

Some views may display estimated engagement-style metrics or other model diagnostics. These are synthetic model outputs and do not establish future clicks, conversion, reach, or budget efficiency without separate live evidence.

Creating an Ad Test

1. Go to Ads in the app.
2. Click New Ad or paste ad content.
3. Select a test type.
4. Configure the audience and comparison criteria.
5. Run the simulation.

Supported Ad Types

| Type | Elements Reviewed | | --- | --- | | Display | Image, headline, body, CTA | | Video | Thumbnail, hook, messaging | | Social | Copy, hashtags, visual | | Search | Headlines, descriptions |

Running a Tournament

1. Add the ad variants.
2. Configure tournament settings.
3. Start the model comparison.
4. Review bracket and archetype-level results.
5. Select candidates for a live or human test.

Mathematical repetitions or model responses are not human respondents and do not create human-population statistical power.

Generated Suggestions

Tests can generate candidate changes to:

• headlines
• body copy
• calls to action
• audience hypotheses

Review suggestions for brand, legal, factual, and platform-policy requirements before use.

Brand Context

Where Brand Memory is enabled, prior product, brand, or test context may be supplied to later model calls. Context reuse can make suggestions more consistent with the supplied brand information; it does not prove automatic learning or improved predictive accuracy.

Best Practices

1. Test before production spend, while changes are inexpensive.
2. Change one major variable at a time when possible.
3. Read the reasoning, not just the score.
4. Treat model rankings as hypotheses.
5. Validate the leading candidate with real customers or live campaign outcomes.

Related Articles

• Products - Setting up product context for ads
• Discovery - Exploring audience hypotheses
• Compare - Side-by-side model comparisons