GEO Website Optimization Diagnostic Tool

Problem

Brand mentions in AI answers are not the same as traditional SEO visibility. Engineers need to know how a target brand is cited, compared with competitors, and supported by real user-facing questions. This skill turns a GEO diagnosis into a repeatable workflow instead of a one-off prompt summary.

How It Works

The minimum inputs are brand_name, website, industry, and target_customer. It first completes missing fields, then validates inputs, researches the brand, generates a question map, executes searches, analyzes citations, and computes metrics. At least 70% of questions must omit the target brand name, and at least 30% should target recommendation, comparison, or purchase intent, reducing self-referential bias.

Key steps include:
- using web_search for real results, or switching to offline_import when search is unavailable
- extracting brand, competitor, source type, sentiment, and recommendation signals
- calculating numeric metrics with deterministic scripts rather than manual reasoning
- generating prioritized opportunities and content briefs after quality checks

Boundaries and Notes

Cited URLs must come from actual search results or user input. Claims that cannot be verified should be marked unverifiable. The workflow does not guarantee rankings, inclusion, or stable AI-answer placement, and it must not bypass login, paywalls, or access controls. If search success falls below the threshold, treat the run as failed or partial.