LLM Visibility Monitoring

LLM Visibility Monitoring tracks whether AI assistants recommend your site when users ask about your niche. Every week RankAscend probes multiple language models with niche-relevant questions and records whether your site appears in the response, where it ranks, and what the model says about it.

Enabling monitoring

Go to Settings → Integrations → LLM Visibility and toggle Enable weekly scans.

No OAuth or API keys are required. RankAscend uses your site's niche configuration to generate prompts automatically. If your site has no niche set, configure one in Settings → General first.

When you enable the feature, RankAscend checks whether a scan is overdue. If your last scan was more than seven days ago (or never run), it triggers one immediately.

How scans work

Models probed

Each scan queries three models in parallel:

  • Name
    OpenAI
    Type
    model
    Description

    Probed via the Pollinations API proxy with the openai model.

  • Name
    Gemini
    Type
    model
    Description

    Probed via the Pollinations API proxy with the gemini-fast model.

  • Name
    Perplexity
    Type
    model
    Description

    Probed via the Pollinations API proxy with the perplexity-fast model.

Configuration

  • Name
    Enable weekly scans
    Type
    boolean
    Description

    Turns the weekly Sunday 6:00 AM UTC scan schedule on or off. When disabled, no automatic scans run, but existing data is preserved.

  • Name
    Mention keywords
    Type
    string[]
    Description

    Keywords RankAscend looks for in model responses. Defaults to your site name and domain. Add product names, brand variants, or specific URLs you want to track. Separate entries with Enter or a comma.

Data is retained for 12 weeks by default. Older probe records are deleted automatically during each scan run.

Error handling

Individual model failures do not stop a scan. If one model times out or errors (RankAscend waits up to 20 seconds per probe), that probe is skipped and logged, and the scan continues with the remaining models and prompts. Partial results are saved normally.

ScenarioWhat happens
One model times outThat probe is skipped; others complete normally
Site has no niche configuredScan is skipped for that site; no error shown
Mention keyword list is emptyFalls back to site name and domain as defaults
Data retention pruning failsLogged as a warning; new probe data is still saved

Was this page helpful?