FAQ & troubleshooting

My analytics are empty — why?

Adding a question only configures it — you have to run a collection to generate answers. Press Run now on Brand Exposure or Brands. After a run, give it a moment: metrics come from a second analysis pass that lands shortly after the raw answers. Also check the top-bar filters — a 7-day date range hides older data, and a platform filter hides every other engine.

Why does Brand Exposure show fewer answers than Questions?

Questions counts raw answers; analytics count analyzed ones. Empty or failed model replies are stored but never analyzed, so the analyzed count is smaller. Expected, not a bug.

Citations stay empty.

Citations come from source links inside answers, and many models rarely cite sources. Your citation data can stay sparse even while every other metric is fully populated. Nothing to fix.

Which AI engines does GEOX query?

Up to ten, one flagship model each: Anthropic (Claude), OpenAI (ChatGPT), Google (Gemini), xAI (Grok), DeepSeek, Alibaba (Qwen), Moonshot (Kimi), Z.ai (GLM), MiniMax, and Xiaomi (MiMo). The three core engines run by default; enable the rest per brand on Settings.

How often does GEOX collect data?

Whenever you tell it to: manually with Run now or a question’s ▶ button, or automatically when a brand has a schedule — daily, weekly, monthly, or a one-time run, in your local time zone.

Paused vs. deleted — what’s the difference?

Pausing a brand or question keeps its history and skips it on future runs — you can resume anytime. Deleting removes the configuration. Raw responses are the source of truth and are never thrown away.

What exactly is the GEO score?

A 0–100 grade of how quotable a piece of content is for AI engines. Five signals are measured from the text (direct answer up top, sentence length, factual density, structure, schema readiness) and two are judged by an AI auditor (entity coverage, authority). It always comes with a ranked list of concrete fixes.

Do page tests affect my analytics?

No. Page tests ask the engines directly and store results separately — they never touch the responses and metrics your dashboards are built on.

Are the volume and difficulty numbers real search data?

Depends on the page. Keyword Research and Domain Overview use real search-market data. The volume and difficulty on Question Mining results are AI estimates for relative comparison — cross-check promising questions against Keyword Research.

What does tracking a mined question do?

It promotes the question into your monitored set, so collections measure it on your schedule alongside hand-written questions. Untracking pauses monitoring; history is kept.

Where do tasks come from?

Anywhere with an “add to tasks” action: the Task column in Question Mining, “Publish task” on a Content Studio draft, “Improve this page” on a page test, outreach targets in Citation Sources, “Close this gap” in Brand Exposure — or manually on Tasks. Each task deep-links back to its source.

What consumes credits?

Anything that calls a live model or data provider: collection runs, question mining, content generation, page tests, keyword and domain lookups. The meter at the bottom of the sidebar shows your balance; Settings → Credits has the activity ledger. Cheapest levers: pause unused questions and disable engines you don’t need.