Product overview
What PeakGEO does, what it delivers, and how a diagnosis runs end to end.
On this page
PeakGEO diagnoses brand visibility in Chinese AI search. It starts from the questions real users ask AI, puts each question to models such as Doubao, DeepSeek, Qwen, Tencent Yuanbao, Kimi, ERNIE, Zhipu Qingyan and iFlytek Spark, keeps every answer and its cited sources, measures whether your brand is mentioned and whether it is recommended, and turns that into evidence-backed actions you can retest with the same questions.
PeakGEO diagnoses, suggests and retests. It does not control model answers, and it does not promise that your brand will be recommended or that visibility will improve.
What PeakGEO answers
| You want to know | What PeakGEO gives you |
|---|---|
| How does AI see my brand today? | Organic mention rate, organic recommendation rate, risk mention rate, plus key findings (opportunities, gaps, risks) |
| How do users ask AI about my category? | A question set organized by theme, separating brand-free organic questions from brand verification questions |
| Why does AI recommend competitors? | Competitor analysis, share of voice, and the sources answers cite |
| What backs each conclusion? | Every finding links back to its questions, models and the original answers |
| What should I do next, and how do I verify it? | Actions, a content workflow, and like-for-like retests and monitoring |
Model coverage and availability
Formal diagnoses integrate Doubao, DeepSeek, Qwen, Tencent Yuanbao, Kimi, ERNIE, Zhipu Qingyan, iFlytek Spark, ChatGPT, Claude, and Grok. You can select several platforms per diagnosis. Available choices depend on the console’s current catalog and your account permissions; platforms that have not been enabled are not shown as available choices.
The website’s free quick check covers Doubao, DeepSeek, Qwen, and Yuanbao. Its scope differs from a formal diagnosis in the console.
Platform names identify sampling channels. Measurements use the corresponding public APIs and may differ from answers an individual sees in the consumer app. Model versions and search capabilities also affect results; keep questions, platforms, and sampling conditions comparable when retesting. See Metric definitions.
How a diagnosis runs end to end
- Start a new brand diagnosis: provide a website URL or upload brand / product materials, then confirm the brand details, target market, buyer persona and competitor list.
- Settle themes and questions: PeakGEO derives diagnosis themes and generates user questions in the background. Every question goes through an automatic check.
- Choose AI platforms and run: each question is asked independently on the selected models, and the full answer text and cited links are saved.
- Read the results: the Overview page shows the core metrics, key findings and next step; the insight pages let you drill into questions, models and sources.
- Act: verify the cause first, then carry out improvements once the evidence is clear. When new content is needed, Content produces drafts backed by sources.
- Retest and monitor: retest with the same questions and models, or schedule monitoring, and watch for changes within a comparable scope.
See the Quickstart for the step-by-step guide.
Console navigation
Once you open a brand, the sidebar is grouped in working order:
| Group | Pages | What they are for |
|---|---|---|
| Overview | Overview, Report | This run's conclusion, core metrics, key findings and next step; a report you can read online or download |
| Questions | Prompts, Tracking | Manage the question set and core retest set; compare each question's answers across models |
| Insights | Visibility, Competitors, Sources, Evidence | Metric trends and comparisons, competitor performance, cited sources, evidence gaps |
| Execution | Actions, Content, Monitoring & retests | Move actions forward, write and log published content, retest and monitor on a schedule |
| Configuration | Brand profile, Settings | Maintain the brand story and reference materials; workspace settings |
Core metrics at a glance
- Organic mention rate: the share of answers to brand-free organic questions that mention your brand.
- Organic recommendation rate: among all valid organic answers, the share where an organic recommendation question produced an explicit recommendation of your brand.
- Risk mention rate: the share of organic answers that express caution, negativity or risk about your brand.
Brand verification questions and custom questions are excluded from all three. Each rate carries a 95% confidence interval and with fewer than 20 valid answers the interval is wide, so treat the rate as indicative. With no valid answers, the console shows "No valid answers" rather than 0%. See Metric definitions for the full definitions and Cited sources and share of voice for citations.
Boundaries
Model answers vary from run to run. A change seen in a retest only means a change was observed this time; it does not prove that a particular action caused it. When the before and after confidence intervals overlap, the result is shown as indistinguishable. See Comparing runs and confidence intervals.
- PeakGEO never publishes content for you. Once a draft passes the fact check, you publish it on your own channels and log the link back in the console.
- A frequently cited source is simply one models use often; it is not a judgment that the source is accurate or reliable.
- Uploaded materials help PeakGEO understand your brand. They are not treated as public, and uploading them does not mean any model cites them.
Next steps
- Quickstart: from a new brand to your first diagnosis results.
- Question sets and measurement types: how questions are generated and how the four measurement types differ.
- Glossary: common terms used in the docs and console.