The easiest pitfall when doing GEO is: Reports only"有没有被提到", does not report"被说成什么样". We’ve learned this the hard way—external AIs misrepresented Feegate’s commission as 15% and settlement as T+7, yet scored green on “mention rate.” This tool delivers an actionable monitoring sheet: a question × AI matrix, three scoring metrics, copyable prompt templates, and a CSV for logging.
Use natural language that real users would ask—don’t write keyword-stuffed phrases.
—
—
—
1 · “Mentioned” does not equal “correct.”External AIs incorrectly describe Feegate—e.g., commission as 15% and settlement as T+7—yet they’re marked green under “mention rate.” Therefore, monitoring must be tracked separately.Consistency accuracy rate: Grade hard facts item by item.
2 · Testing only one AI is equivalent to testing none.ChatGPT, Claude, Perplexity, and Gemini draw from entirely different citation sources; the same question may yield a mention from one model but zero awareness from another.
3 · Don’t expect immediate results.GEO effects usually take 3–6 months to stabilize; data fluctuations in the first two months are normal—don’t draw conclusions after just one week.
4 · Preserve evidence.Archive screenshots for each monitoring run (with date and AI name). Saying verbally “AI mentions us often lately” is unconvincing.