What it is in one line
ByteDance's consumer AI all-in-one: chat, search, writing, translation, coding, and images
behind one entrance — and the #1 AI-native app in China by monthly actives.
Who built it
ByteDance. The Doubao business is run by Zhao Qi, who led ByteDance's growth middle office
and management research institute, then took over the Flow team's HR, then Doubao itself.
It sits at the center of ByteDance's three-pillar strategy — AI, information platforms, and
transaction services — with Doubao expected to feed search, e-commerce, and office the way
Douyin did.
Entry relationship: this listing (appid, MAU metric) is the main Doubao app; the separate
"Doubao | Douyin" listing (webid, visits metric) is the Douyin-pool channel version of the
same product.
Read: Doubao is not "an app made by the best model" — it is "an app made by the strongest
channels and the best productization." Its rivals are never short on models; they are short
on this growth and distribution engine.
What it actually does
- Chat and Q&A → the daily high-frequency action of conversational AI; voice and a
human-like persona are its standout differences
- Search → built-in search, replacing opening a separate browser
- Writing / translation / coding → a general productivity toolkit
- Image capabilities → generation and editing folded into the assistant
- Multi-app matrix → Jimeng (image/video), Doubao Aixue (education), Trae (coding IDE)
form a ByteDance AI matrix with Doubao as the main entrance
What old behavior it replaces
It replaces the daily combination of search + Q&A + translation + writing tools: a fact
check used to mean opening a search engine, a question meant asking a person, an English
passage meant opening a translator, and a piece of copy meant opening a document. Now all
of it happens in one dialog box. It is the smoothest domestic execution of this
substitution — not because its model is the strongest, but because it is fast, personable,
and within reach through the channels.
The ByteDance matrix replaces another habit: one app per job. Jimeng for images, Doubao
Aixue for homework, Trae for coding, and Doubao as the front door — the "app factory"
playbook carried into the AI era.
Business model
Free-first, with paid subscriptions just launched. On June 24, 2026, Doubao shipped
its pro tier: Basic free; Standard at ¥68/month continuous (¥688/year); Plus at ¥200/month
(¥2048/year); Pro at ¥500/month (¥5088/year). Paid tiers target office task modes.
Read: the significance is not the revenue, it is the signal — the largest free AI product
in China is probing paid tiers, which means the "all-AI-is-free" era is ending for the
leaders. How many of 300M+ actives pay ¥68 a month is the number the whole industry now
watches.
Hard numbers
- Traffic-board (this pool): MAU 324.31M, -1.71% MoM, on the domestic overall, global,
and domestic growth boards
- QuestMobile: 345M MAU in March 2026 with DAU peaking above 150M; ~382M in June 2026,
+172.1% YoY
- Another source (Quanmeiti Stars): ~502M MAU in April 2026, -4% MoM
- Doubao models: 180 trillion daily token calls by June 2026, >1,500x growth from the
120B at launch in May 2024
- Competitive comparison (QuestMobile, March 2026): Doubao 345M vs Qwen 166M vs
DeepSeek 127M — Doubao alone exceeds the other two combined
Four-way read
| Dimension |
Call |
| Founder-product fit |
Zhao Qi came out of the growth middle office; putting the company's best growth person on the app that needs growth most is a near-perfect match |
| Product insight |
The model is not the advantage; experience closes the gap — persona and response speed are real perceived differences |
| Execution quality |
180T daily tokens proves engineering and infrastructure at top scale, not a demo |
| Timing |
The general-assistant layer is settled; Doubao is the locked-in leader, but it faces high-base growth exhaustion |
The call
This is the end-state of China's general-assistant layer. 324M MAU in this pool,
more than the #2 and #3 combined by QuestMobile's count — the top position is
indisputable and corroborated across sources. The -1.71% MoM is ordinary noise at this
base, not an anomaly. It no longer needs a growth story; it needs to hold.
Read: Doubao proves that in Chinese consumer AI, channels and productization matter more
than models. ByteDance does not run the best model; it combined "good-enough model +
best distribution + strongest productization" into a position nobody can catch. The
implication for founders is direct: the general-assistant lane is closed; do not enter.
The paid subscription is this year's real story. The ¥68+ tier cuts at office
productivity, essentially spending Feishu's decade of office-scene understanding on
raising Doubao's average revenue. How much it collects and how many pay will set the
reference frame for "China's AI app monetization ceiling."
The risk underneath: media-reported ByteDance net profit fell sharply in 2025,
mainly on AI investment; Li Liang of Douyin Group disputed the accounting basis.
Compute cost grows linearly with usage — the structural burden of free products, and one
reason Doubao rushed to open subscriptions in June.
What to watch next
① Whether MoM MAU turns positive next month — two consecutive declines would warrant
concern; -1.71% is normal for now
② Paid penetration one quarter after launch (whether officials disclose member counts) —
the first real data point on China's AI payment ceiling
③ Whether the channel gap between Doubao main and the Douyin channel version
converges — how much incremental traffic the channel playbook still feeds
What you can take from it
Product logic: when your model is not ahead, close the experience gap with persona,
response speed, and scene coverage — Doubao does these three well enough that users feel
it is the best. Small teams should do the same: do not fight the model war, fight for
completion in one scene.
Pricing structure: keep the base free for the mass, tier the productivity scenes by
demand intensity (¥68/200/500 continuous monthly) rather than by feature gating —
pricing compute quotas instead of locking features triggers less resistance.
Verdict
Worth watching. The landscape is settled, the data is real, and its position needs no
defense. The actual watch point is the paid subscription — whether it goes from "free #1"
to "paid #1" will define the payment ceiling for the entire Chinese AI app industry. Let
subscription penetration be the judge in three months.