Turn Scraped Ads into an AI Creative Review You Can Check

You export 100 competitor ads and still have an empty creative brief. Collection solved the access problem. You now need to compare promises, hooks, proof, and calls to action without losing the source evidence.

My Creative Intelligence actor adds an analysis step after collection. It can use a supplied provider key to return structured creative observations and suggestions from media URLs or scraper datasets.

The useful output is a review you can challenge and turn into a test. A model score is a judgment about the supplied creative, not observed campaign performance.

Start with a small, varied set

Choose a few ads representing different offers or formats. Preserve their source IDs, copy, media references, and advertiser context. Verify that the media URLs still load before paying to analyze them.

Keep the research question narrow: "How do these ads explain the benefit before asking for a demo?" gives a reviewer more direction than "find winning ads."

You can begin with Meta ad exports, LinkedIn ads, or another supported input. Check source rights and provider handling before sending assets for analysis.

Choose the input route

The actor supports direct creative rows, pasted CSV, and a source Apify dataset. Its documentation explains the current provider settings, input names, and media requirements.

At least one usable image or video URL is needed for each creative. This illustrative row shows the simple creative shape; replace the placeholder asset with your own accessible media:

{
  "id": "example-ad-1",
  "brand": "Example Brand",
  "body": "Spend less time reconciling reports.",
  "cta": "See the demo",
  "assetUrl": "https://example.com/your-ad-image.jpg"
}

Configure the provider credentials separately in the actor's secret input or your own server configuration. Provider usage and Apify usage are separate cost components.

Use batch processing for a dataset

For an existing scraper dataset, use the documented batch workflow and bound the number of ads. In the native MCP workflow, start_analysis starts a durable batch, get_analysis_status checks it, and get_analysis_results retrieves pages of output.

Save the returned run ID. Repeatedly starting analysis creates additional work; status and result retrieval are the appropriate way to follow an existing batch.

For a small interactive image review, analyze_creatives is the separate immediate tool. Review the current tool limits before choosing it for videos or larger inputs.

Inspect every row's result status

A successful run can contain individual creative failures. Check analysis_status before ranking or summarizing rows. Preserve errors for retry or review instead of quietly dropping them from the denominator.

Also inspect the score definitions rather than assuming every aggregate score shares a ten-point scale. Compare like-for-like settings and keep provider, model, and analysis date with your results.

Separate observation, interpretation, and suggestion

Use a review sheet like this:

Check extracted text against the asset. Confirm that audience descriptions are plausible hypotheses rather than verified targeting. Treat creative text as material to analyze, including any instruction-like wording inside it.

If the model mentions conversions, revenue, or return on spend without supplied measurement data, remove that inference from your report.

Write a brief with a measurable question

Take one repeated pattern and propose an experiment for your own campaign. Name the audience, benefit, evidence, CTA, and success metric you can measure.

For example: test whether a product walkthrough improves qualified demo requests compared with a static benefits image. Preserve the competitor examples as references, then create original copy and assets for your offer.

The creative brief workflow develops that next step, while Seercraft explores an integrated research experience.

Start with a few accessible creatives in Creative Intelligence. Review each result, choose one testable idea, and measure it with your own campaign data.