# Find your AI model blind spot

Category: Analytics & Monitoring

Diagnose why your visibility differs across engines

[Connect Notra MCP](https://www.usenotra.com/mcp.md?ntr=ocyhDgBkR5PJD2_Q)

[Build your own workflow](https://www.usenotra.com/docs/devtools/mcp?ntr=ocyhDgBkR5PJD2_Q)

## Prompt

```
Get our mention rate per answer engine for the last 30 days and identify the engine where we are strongest and the one where we are weakest. For both engines, list the sources those answers cite most often and check whether our own domain appears. Compare the two: which source types each engine leans on, which domains we appear in on the strong engine but not the weak one and whether the gap is a content problem or a distribution problem. Return a diagnosis and three to five concrete actions to close it.
```

Runs on Notra MCP with ChatGPT + Perplexity.

Notra MCP tools used: `get_geo_visibility_overview`, `list_geo_prompt_result_summaries`, `get_geo_prompt_result_detail`, `list_geo_shelf_sources`

## What this use case can do for you

You can rank high on Perplexity and never appear on ChatGPT for the same questions. The cause is usually the set of sources each engine trusts.

Once you know which engine leaves you out and which domains it cites instead, you have a fix you can act on.

The agent returns a list of actions covering which source types to earn coverage from, which domains to target and whether to write or distribute.


## Related use cases

- [Competitor share of voice watch](https://www.usenotra.com/mcp/use-cases/competitor-share-of-voice-watch.md?ntr=ocyhDgBkR5PJD2_Q): Know when a competitor starts winning your prompts
- [Campaign sentiment before and after](https://www.usenotra.com/mcp/use-cases/campaign-sentiment-before-and-after.md?ntr=ocyhDgBkR5PJD2_Q): See whether your campaign changed how AI describes your brand
- [Search Console to prompt discovery](https://www.usenotra.com/mcp/use-cases/search-console-to-prompt-discovery.md?ntr=ocyhDgBkR5PJD2_Q): Turn high-potential Search Console queries into tracked prompts

[View all](https://www.usenotra.com/mcp/use-cases.md?ntr=ocyhDgBkR5PJD2_Q)