Baidu Search API Drops to ¥4.9/1,000 Calls, Seeking New Monetization Paths Amid Declining MAU
nashnova research
Baidu slashed its search AI API promotional price to ¥4.9 per 1,000 calls, seeking a new B2B revenue line as monthly active users fell 12.4% year-on-year and ad revenue dropped 19% — a bet that search can monetize even when users leave the app.
What exactly is Baidu selling at ¥4.9 per thousand calls?
Baidu launched two search AI API tiers: an Enhanced tier at ¥4.9/1K calls (list price ¥40) and a Flagship tier at ¥9.9/1K calls (list price ¥72).
This means → the promotional price is roughly one-eighth to one-tenth of list — essentially a subsidy to pull developers onto the platform and build ecosystem volume first.
The Enhanced tier caps cumulative usage at 1 million calls; the Flagship tier requires an invitation code. Both revert to list price once the quota is exceeded — the discount has a ceiling.
Why the urgency to cut prices now?
Baidu's Q2 2026 earnings: online marketing revenue was ¥13.1 billion, down 19% YoY; AI-native marketing revenue was just ¥2.6 billion in the same period — nowhere near enough to close the gap.
The starker signal is user attrition: Baidu App MAU hit 644 million in June 2026, down roughly 91 million from 735 million a year earlier — a 12.4% YoY decline.
In plain terms = ad revenue depends on traffic, and traffic is leaving. Baidu needs a path that earns money even when users never open the Baidu app — the API is that path.
What is the commercial logic behind a search API?
The search API — Baidu's retrieval capability packaged as a pay-per-call interface for external developers — targets third-party AI applications, not end consumers.
This means → if a user queries someone else's AI assistant and that assistant calls Baidu's retrieval back-end, Baidu still earns revenue. Monetization decouples from its own app traffic.
This reflects a fundamental shift: from "users must come to us" to "our capability can embed anywhere" — a rewiring of Baidu's core business model.
What did the technical upgrade actually deliver?
The API now uses a four-layer architecture (content, framework, engine, application). Baidu claims data reliability rose 50%, Ernie Bot integration improved 20%, and uptime reached 99.999%.
The company did not disclose testing methodology — these figures are official claims only and cannot be independently verified.
On the consumer side, Ernie Bot added multi-turn search, proactive alerts, and tool-based task completion across 170+ scenarios; the AI camera now supports 9 dialects.
The whole search market is shrinking — can the API fill the ad gap?
CNNIC data: as of December 2025, China had 782 million search engine users, down roughly 95.76 million from the prior year. Usage rate fell from 79.2% to 69.5%.
In plain terms = Baidu is not the only one losing users — the entire search pie is contracting. This is structural, not company-specific.
This means → whether the B2B API path can scale enough to offset declining ad revenue is Baidu's single most important proof point going forward. Low prices can attract developer trials, but retention and renewals will decide if this path actually works.
市场有风险，内容仅供研究参考，不构成投资建议。