AI Brand Visibility Monitor — Chinese + Western engines

China AI search visibility: DeepSeek, Qwen, Kimi, GLM (+ ChatGPT, Gemini, Claude)

See what AI engines say about your brand — DeepSeek, Qwen, Kimi and GLM alongside ChatGPT, Gemini and Claude — and watch it change over time.

When someone asks an AI "what's the best running-shoe brand?" or "is Brand X any good?", that answer is the new shelf placement. This Actor measures it. Give it your brand and the real questions your customers ask, and for every prompt × engine it returns:

• Mentioned? — is your brand named in the answer
• Sentiment — positive / neutral / negative + a −1…1 score
• Rank — where you sit vs the competitors you list
• Share of voice — you vs rivals in that answer
• Cited sources — the URLs the engine surfaced (search-enabled engines)
• Answer snippet — the evidence behind the score

This is a monitor, not a one-off lookup — and that's the point

A single snapshot of "does ChatGPT mention me" says little, because AI answers are volatile: the same question can name different brands from one run to the next. The value is the trend — did you just drop out of DeepSeek's answer? did a competitor overtake you on Qwen this week? Run it on an Apify Schedule with deltaMode on, and each run returns only what changed since the last — so a quiet week costs almost nothing and a real shift surfaces the moment it happens.

A real program isn't two test prompts. It's your full buyer-intent prompt set (10–50 questions) × several engines × your competitor set, on a weekly or daily cadence — that's where the signal (and the value) lives.

Why this one: the China blind spot

GEO / "AI visibility" tools mostly focus on the big Western engines (ChatGPT, Gemini, Claude); fewer cover the Chinese AI engines (DeepSeek, Qwen, Kimi, GLM). If you sell into China, those answers matter too. This Actor tracks brand visibility across Chinese and Western engines in one run.

	Engines	API key
China	DeepSeek · Qwen/Tongyi · Kimi/Moonshot · GLM/Zhipu	your own key
West	ChatGPT · Gemini · Claude	your own key

Quick start

1. Set brand (e.g. Nike) and a handful of prompts (the questions your customers actually ask).
2. Pick engines — exact JSON values: deepseek, qwen, kimi, glm, openai, gemini, claude. Add your own keys in apiKeys for the ones you want, e.g.:

   { "deepseek": "sk-...", "qwen": "sk-...", "openai": "sk-..." }
   

3. Add competitorBrands for share-of-voice, turn on deltaMode, and attach a Schedule.

Each engine bills against your key (pure pay-as-you-go for you, no subscription). Get keys from each provider's console: DeepSeek, Alibaba DashScope (Qwen), Moonshot (Kimi), Zhipu (GLM), OpenAI, Google AI Studio, Anthropic.

Example input

{
  "brand": "Nike",
  "competitorBrands": ["Adidas", "Li-Ning", "Anta"],
  "prompts": [
    "What are the best running shoe brands?",
    "Which sportswear brand has the best quality?",
    "Is Nike worth the price?",
    "Recommend a running shoe for marathons"
  ],
  "engines": ["deepseek", "qwen", "kimi", "glm"],
  "apiKeys": { "deepseek": "sk-...", "qwen": "sk-...", "kimi": "sk-..." },
  "deltaMode": true,
  "deltaStateKey": "nike-weekly"
}

Who runs it (and keeps running it)

• GEO / SEO / market-research agencies — a per-client, multi-engine visibility feed they bill monthly. It's their recurring deliverable, so it runs on a schedule by definition.
• Brands selling into China — track your AI visibility on DeepSeek, Qwen, Kimi and GLM alongside ChatGPT, Gemini and Claude; perception there shifts constantly, so they monitor continuously.
• Cross-border / market-entry teams — track how the engines your China customers actually use describe you and your rivals, week over week.

Agencies: one tracking stream per client

Running this for clients? Use one deltaStateKey per client (e.g. client-nike-weekly) so each client has an independent tracking stream, schedule one Task per client, and export the dataset to Sheets/Looker Studio for a branded report. Large programs: the maxChecks safety cap defaults to 200 per run — raise it for big brand × prompt × engine grids, or split per-client Tasks.

Pricing

Pay-per-event, and you bring your own LLM API keys — so you pay each provider directly for inference and pay Apify only for the analysis:

• $0.25 per visibility check (1 brand × 1 prompt × 1 engine)
• +$0.05 per check in delta mode
• +$0.05 per competitor scanned per check

Program	Checks / month	Your Apify cost
One-off audit — 1 brand × 10 prompts × 4 engines	40	~$10
Brand monitor — 1 brand × 10 prompts × 4 engines, weekly	~170	~$43 max — with deltaMode usually less (only CHANGED answers bill; $52 = worst case if everything changes)
Agency — 5 brands × 15 prompts × 4 engines, weekly + 3 competitors	~1,300	~$325 base ($585 = worst case with delta + competitor add-ons)

(Monthly figures use 4.33 weeks/month; 1 check = 1 brand × 1 prompt × 1 engine.) A 20–50-prompt program across several engines weekly comes to roughly $150–$600/mo on Apify. Here inference is billed by each engine provider on your own key, on top of the $0.25 check.

Output

One record per brand × prompt × engine: mentioned, sentiment, sentimentScore, rank, shareOfVoice, competitorsMentioned, citations, answerSnippet, and changed (in delta mode). A SUMMARY record totals mentions and checks. Export JSON / CSV / Excel or pull via API into your BI.

Notes

• What it measures: how engines answer about your brand. Mention/sentiment/rank work on every engine; cited sources appear when the engine does live web search (e.g. Qwen, GLM).
• Privacy: your API keys are read per run and never logged or stored.
• Want another engine (Doubao, ERNIE, Yuanbao)? Open an issue on the Actor page.

FAQ

Do I need my own API keys? Yes — you bring a key per engine you select, so you pay each provider directly for inference (cheapest path for heavy use) and only pay Apify for the analysis. Run it with no key and you get a free labeled sample so you can see the exact output first.

What does one $0.25 "check" buy? One brand × one prompt × one engine, fully scored: mentioned, sentiment, rank vs your competitors, share-of-voice, cited sources and the answer snippet.

How is this different from Profound / Otterly / Peec? It covers the Chinese engines (DeepSeek, Qwen, Kimi, GLM) alongside ChatGPT, Gemini and Claude, and it's pay-as-you-go on your own engine keys rather than a subscription.

Do citations work on every engine? Cited source URLs appear on the engines that do live web search (e.g. Qwen, GLM). Mention / sentiment / rank / share-of-voice work on every engine.

Does a quiet week still cost a lot? No. With deltaMode on a schedule, a run that finds nothing changed returns (and bills) almost nothing — you pay for signal, not silence.

What if one engine's key is wrong? That engine's checks are surfaced as clearly-labelled error records and are never charged — only successful checks are billed.

Are my keys safe? They're read per run and never logged or stored.

Support & maintenance

Built and maintained by zhorex — part of a suite of Chinese-platform intelligence Actors. Want a new engine (Doubao, ERNIE, Yuanbao…) or found a bug? Open an issue on the Actor page.