What it is in one line
Turning the "search a template → download → edit it yourself" document workflow
into "describe it, get a first draft" — money still comes from memberships and
downloads, but the user now gets an assistant instead of a file.
Who built it
Baidu. Wenku is Baidu's incumbent content platform (launched 2009). Since 2025 it
has been operated jointly with Baidu Drive under the shared GenFlow agent; on
2026-08-14 GenFlow was officially renamed "Kuku AI" with a standalone client
launched. Models underneath: Ernie plus DeepSeek and others.
Read: Wenku adding AI is defense, not offense — the real moat is not the AI
capability but the decade-old document stock and the paying-user habit already
built.
What it actually does
- AI generation → PPT, copy, reports, Word/Excel handling; one-sentence drafts
- Q&A over documents → ask questions against the authoritative corpus inside Wenku
- AI notes / mind maps → videos and documents converted to structured notes in one click
- Search and creation fused → search is not the end point; search now ends in a finished artifact
- Kuku AI standalone → PC client, web, mini-program, enterprise edition
What it deliberately does not do: no general-purpose chat. Wenku's AI is
entirely anchored to documents — generation, Q&A, and notes all hang on the
document asset.
What old behavior it replaces
Making a PPT or report used to follow this flow: search Wenku for a sample →
find a good one → download → pay to unlock → spend half a day editing. The core
was "humans find documents, humans edit documents."
The AI version compresses it to: describe the need → get a draft → polish. It
replaces not a single tool but the act of "hunting for documents in order to make
documents" — students writing theses and office workers building decks, one of
the highest-frequency AI-work scenarios in China.
Business model
- Traditional: VIP membership (downloads, document unlocking) and per-document payment
- AI: 40M+ paying users (official figure, April 2026, covering AI-related
subscriptions)
- With Kuku AI's standalone client, AI-office work monetizes through subscription
and membership going forward
Read: Wenku is the clearest example of a business model rewritten by AI — the
same users, the same willingness to pay, but the object purchased shifts from "a
document" to "an assistant that produces documents." Far cheaper than building an
AI-office product from zero.
Hard numbers
- Web visits 23.77M/month, +4.83% MoM (traffic board, 2026-08 measurement)
- Official (April 2026): Wenku AI MAU over 97M, 40M+ paying users, 1.8B+
authoritative professional documents, #1 globally in monthly smart-PPT visits
- August 2026: GenFlow MAU passed 100M, of which AI-office MAU is 25M+
- The traffic board web-visit figure and the official "97M AI MAU" differ hugely — the
measures are different (official includes app/embedded/mini-program), and the
source series must be stated when citing
Four-way read
| Dimension |
Call |
| Founder-product fit |
A corporate incumbent product, not a founder's project |
| Product insight |
Converting a content platform into a creation entry point; inherently defensive |
| Execution quality |
Ernie plus multi-model; generation quality is adequate and switchable |
| Timing |
AI office is one of the hottest paying scenarios; Wenku holds both user and content stock |
The call
Wenku is the standard sample of an incumbent content platform adopting AI — and
it did it smartly. It did not spin AI into a standalone product to fight the
giants; it grafted AI onto the existing "search documents → produce documents"
path. Zero switching cost, and the payment reason did not change — the purchase
just upgraded.
The moat is documents, not AI. 1.8B authoritative documents are a decade of
accumulation that WPS AI, Tencent Docs, and every AI-PPT tool cannot take away;
nor can they take the 40M paying users' habit. AI capability is actually the
easiest piece to fill in — which is exactly the right playbook for an incumbent:
re-activate old assets with AI instead of betting on new capability.
One self-cannibalization risk deserves naming: once generation quality is good
enough, the need to "download an existing document" falls, and traditional
download revenue gets eaten by its own AI. How Wenku balances the old and new
revenue lines is the most interesting internal game to watch over the next year.
Growth signal is healthy: web visits +4.83% MoM, mild growth among top tools;
the official AI-office MAU of 25M is climbing fast.
What to watch next
① The revenue mix between traditional download payments and AI subscriptions —
is AI eating the old line, and how is it balanced
② Whether the Kuku AI standalone client (PC) actually acquires new users — the
key move that turns AI-office from "a Wenku add-on" into "a standalone product"
③ Whether the 25M AI-office MAU keeps climbing — the test of the
"document asset + AI" growth curve
What you can take from it
Product logic: graft AI onto the path users already walk instead of starting
over. An incumbent content library is the most overlooked ammunition in the AI
era — upgrading "search for documents" into "ask documents, generate documents"
keeps the user path intact while the product changes generation. Applies to any
platform holding historical data or content.
Pricing structure: running AI subscriptions in parallel with legacy VIP — old
users keep paying for downloads, new users pay for generation, the two lines do
not disturb each other. That is the safest collection design for an incumbent
adopting AI.
Verdict
Worth watching. Data is real and growth is healthy — a positive specimen of
incumbent-platform AI-ification. The story hinges on the old/new revenue balance
and the standalone client's fate: whether Wenku just defends its old business or
actually grows a second curve.