Analysis Runs

A run sends your project's prompts to each enabled AI provider and collects the responses. This is the core action in GEOlyze -- each run produces a snapshot of your company's AI visibility at a specific point in time.

Starting a run

From the Runs page of your project, click Start New Run. GEOlyze will:

1. Send each enabled prompt to every active provider.
2. Analyze each response for entity mentions, website citations, competitor mentions, and sentiment.
3. Compile the results into a detailed report.
4. If website crawl is enabled, crawl your website and analyze GEO readiness, Schema.org markup, content structure, and keyword coverage.
5. Generate a diagnosis with health score, gap analysis, and recommendations.

Depending on the number of prompts and providers, a run typically takes a few minutes. You can track progress in real time with the progress bar showing completed prompts out of total.

Cost estimation

GEOlyze checks your balance before starting a run and shows a cost breakdown:

• AI analysis cost -- based on the number of prompts and providers.
• Website analysis cost -- if website crawling is enabled.
• Total estimated cost and your current balance.

If your balance is insufficient, you can top up before proceeding.

Delta runs

A delta run only re-tests prompts that have changed since the last completed run. When prompts have been modified, the Runs page shows a button indicating how many prompts have changed (e.g., "Re-run 3 changed"). Unchanged prompts inherit their results from the previous run.

Delta runs are useful for:

• Quickly checking the impact of prompt edits.
• Saving time and cost when only a few prompts have changed.
• Iterating on prompt wording without re-running the full analysis.

Run statuses

Monitoring a run

While a run is in progress, the run detail page shows:

• A progress bar with the number of completed prompts out of total.
• The current error count.
• Any billing errors.

The page automatically refreshes every few seconds to show updated progress.

Resuming a run

If a run has failed or was cancelled, you can resume it. Resuming retries the remaining unprocessed prompts without re-running those that have already completed successfully.

Run results

Each completed run shows a comprehensive report with the following sections:

Summary metrics

Four headline metrics displayed as cards:

• Entity mention rate -- percentage of responses that mention your company name (validated to exclude false positives from similarly-named entities).
• Website citation rate -- percentage of responses that cite your website URL.
• First mention rate -- percentage of responses where your company appears before any competitor.
• Total responses -- count of successful AI responses (with error count).

Per-provider breakdown

For each provider (Claude, OpenAI, Gemini, Perplexity), the report shows:

• The model used for this run.
• Entity mention rate and website citation rate.
• Average response time in milliseconds.
• Error count out of total responses.

Top cited domains

A ranked list of the most frequently cited websites across all responses. This helps you understand which domains AI systems prefer to reference and where your competitors stand.

Prompt-by-prompt details

Each prompt can be expanded to view every provider's response:

• Provider and model used.
• Mentioned badge -- appears if your company was mentioned in the response.
• Cited badge -- appears if your website URL was cited.
• Error badge -- appears if the request to this provider failed.
• Response time in milliseconds.
• Full AI response text rendered as formatted markdown.

Website analysis summary

If website crawling was enabled, a compact summary card shows:

• GEO Readiness score (out of 10) -- how well your website is optimized for AI discoverability.
• Schema.org score -- how well structured data is implemented.
• Content structure score (out of 10) -- quality of headings, paragraphs, and content organization.
• Keyword coverage -- percentage of target keywords found on your website.

A link takes you to the full website analysis page for detailed results.

Diagnosis summary

If diagnosis data is available, a summary card shows the overall health score and top-priority action recommendation, with a link to the full diagnosis page.

Managing runs

• Download as PDF -- export any completed run as a formatted PDF report.
• Cancel -- stop a running analysis. Completed responses are preserved.
• Resume -- retry a failed or cancelled run from where it left off.
• Delete -- permanently remove a completed, failed, or cancelled run and all its data.

Next steps

• View trends to see how your visibility changes over time.
• Check diagnosis for AI-powered gap analysis and recommendations.