Prompt Movements

Prompt movements are the most actionable data in the trends dashboard. While the progress summary tells you how much your visibility changed, prompt movements tell you where exactly those changes occurred -- which specific queries now return your brand and which no longer do.

Status types

Every prompt in your project is classified into one of the following statuses when comparing two runs:

Gained

A prompt is classified as gained when your company is mentioned in the latest run but was not mentioned in the comparison (previous) run. This is the strongest positive signal -- it means an AI provider has started associating your brand with this query topic.

Gained prompts appear with a green indicator and are the first category highlighted in the movements section.

Lost

A prompt is classified as lost when your company was mentioned in the previous run but is no longer mentioned in the latest run. This is a negative signal that requires investigation -- something caused an AI provider to stop referencing your brand for this query.

Lost prompts appear with a red indicator and are shown alongside gained prompts for quick comparison.

Stable visible

A prompt is classified as stable_visible when your company is mentioned in both the previous and latest run. These prompts represent your consistent visibility footprint -- the queries where AI systems reliably associate your brand.

Stable visible prompts are your foundation. A high count here means your core visibility is solid.

Stable invisible

A prompt is classified as stable_invisible when your company is not mentioned in either the previous or the latest run. These prompts represent persistent visibility gaps -- queries where AI systems consistently do not associate your brand.

While not immediately actionable as "losses," stable invisible prompts are important for long-term planning. They represent untapped opportunities.

Changed

A prompt is classified as changed when your company is mentioned in both runs, but the nature of the mention has shifted. This can mean:

• Provider coverage shift -- you were mentioned by different providers in the two runs. For example, Claude and OpenAI mentioned you previously, but now only Perplexity does.
• Sentiment change -- the tone of mentions shifted (e.g., from neutral to positive, or from positive to negative).
• Citation change -- your website was cited in one run but not the other, even though your entity was mentioned in both.

Changed prompts require careful review because the aggregate visibility appears stable, but the underlying dynamics are shifting.

Top gains and top losses

The prompt movements section highlights the most important changes at the top:

Top gains (up to 5)

The prompts that transitioned from invisible to visible are listed with:

• Prompt reference -- the identifier (e.g., "P-7") that links back to the prompt in your project.
• Prompt label -- the text or shorthand label of the prompt.

These are your recent wins. Each gained prompt represents a query where your GEO and content efforts are producing results.

Top losses (up to 5)

The prompts that transitioned from visible to invisible are listed with the same reference and label format.

These are alerts. Each lost prompt should trigger an investigation into what changed -- on your side, on the competitor side, or in the AI provider's behavior.

Why only 5?

The top-5 limit keeps the summary focused on the most impactful changes. If your project has 100 prompts and 15 gained visibility, the top 5 are the ones most worth celebrating and studying. The full list of all movements is available in the detailed prompt movements breakdown below the summary.

Prompt references

Each prompt movement entry includes a clickable reference that takes you back to the specific prompt within your project. This connection is important because:

• You can review the exact prompt text to understand what type of query gained or lost visibility.
• You can see the full provider-by-provider response history for that prompt.
• You can check whether the prompt is tied to specific keywords that you have been optimizing.

This linking ensures that prompt movements are never abstract numbers -- they always connect back to specific, reviewable queries.

Changed status details

Prompts with a "changed" status deserve special attention. The detail view for a changed prompt shows:

Provider coverage shifts

A before-and-after comparison of which providers mentioned your brand:

In this example, you gained Gemini but lost Claude. The overall mention status is "stable visible," but the provider mix has shifted.

Why provider shifts matter

A shift from generative-model mentions to RAG-only mentions (or vice versa) can indicate:

• Training data rotation -- a generative model may have been updated and no longer includes your content prominently.
• SEO changes -- a change in search rankings could affect RAG-based providers while leaving generative models unchanged.
• Competitor displacement -- a competitor may have gained prominence with a specific provider.

Sentiment changes

If sentiment analysis is active, the changed status also shows whether the tone of mentions shifted:

• Positive to neutral -- AI still mentions you but with less favorable framing.
• Neutral to positive -- your reputation signals may be improving.
• Any direction to negative -- an urgent signal that requires immediate attention.

Using prompt movements for GEO optimization

Prompt movements are your most direct feedback mechanism for optimization work. Here is how to use them strategically:

Analyzing gained prompts

When a prompt moves to "gained," ask:

1. What content did I recently create or update that relates to this prompt's topic? If you can identify the content, you have evidence of what works.
2. Which provider(s) now mention me for this prompt? If only one provider gained you, the cause is likely provider-specific (e.g., a Perplexity gain suggests your page now ranks for a related search query).
3. Are other prompts on the same topic also gaining? A cluster of gains on related queries confirms a broader improvement, not just a random fluctuation.

Analyzing lost prompts

When a prompt moves to "lost," investigate:

1. Did I change or remove content related to this prompt's topic? Accidental content removal is a common cause.
2. Did a competitor publish strong content on this topic? Check Competitor Trends for correlated competitor gains.
3. Is this loss provider-specific or cross-provider? A single-provider loss may resolve on its own (model update noise). A cross-provider loss is more significant.
4. Was the previous mention marginal? If you were only barely mentioned (e.g., in a list of many alternatives), the loss may reflect natural AI variance rather than a real decline.

Prioritizing action on stable invisible prompts

Stable invisible prompts are your roadmap for growth:

• Group them by topic to identify content gaps. If multiple prompts about "enterprise data analytics" are all invisible, you likely need a content hub on that topic.
• Cross-reference with the diagnosis to see if these prompts map to content gaps, GEO gaps, or SEO/GEO gaps. The Diagnosis page provides specific recommendations for each gap type.
• Set priorities based on business value. Not every invisible prompt is worth optimizing for. Focus on the prompts that represent queries your target audience actually asks.

Tracking movement patterns over time

Look at movement data across multiple runs, not just the latest comparison:

• Oscillating prompts that flip between gained and lost across runs indicate borderline visibility. You need to strengthen your position on these topics to achieve consistent visibility.
• Consistently gaining prompts over multiple runs confirm a sustained upward trend driven by real improvements.
• Consistently losing prompts over multiple runs indicate a structural problem that needs to be addressed, such as competitor growth or content decay.

Connecting movements to actions

For maximum value, combine prompt movements with the Action Correlation feature. When you log an action like "Published new FAQ page about X" and the next run shows gained prompts related to X, you have established a causal link between your effort and the result.

This evidence-based approach to GEO is what separates systematic optimization from guesswork.