News

Doubao’s extra 4 points: AI chat traffic priced like intent

From Aug 10, some Doubao→Douyin Merchant orders settle separately—hotels often ~12%, ~four points above the feed. Intent pricing, ledger splits, subsidy-in-base fees, and why ByteDance posts a high rate while volume is thin.

Same room, same Douyin Merchant settlement: 8% from the feed, 12% when the click comes from a Doubao chat. The only variable is AI dialogue as the door.

From August 10, 2026, some life-service orders that jump from Doubao into Douyin Merchant settle on a separate rate card. A 36Kr piece (from “强调 Next”) puts common categories around 9%–18%; hotels are often quoted as 11.4% software fee + 0.6% payment fee ≈ 12%. Sina Tech checked Douyin’s July 27 policy note: specific channels—including Doubao—take effect at 00:00 on August 10.

This is not another “commission hike” gossip. The product question is what ByteDance is pricing when it labels AI recommendation, and why it posts a high price while volume is still thin.

Mechanism: from “merged into organic” to “named channel”

Earlier, Doubao-sourced hotel orders were widely described as rolling into Douyin organic traffic at ~8%. With the new card, settlement lines show an explicit “Doubao” channel—the chat door becomes its own row in the ledger.

The ledger split matters more than the four points. While merged into organic, Doubao and feed orders were inseparable on rate—hard to price dialogue conversion alone, hard for merchants to run channel-level ROI. A named channel lets product and finance ask three questions: is this traffic expensive, is it worth it, should we push more supply. The AI door stops being an experiment and becomes a priced distribution channel.

Worse than the headline number: the fee base is “user paid amount + subsidies from non-merchant parties (e.g. platform coupons).” Platform vouchers can still expand the merchant’s commission base. For merchants, “platform subsidy lifts GMV” does not always mean relief—the commission base rises with it. Layer platform coupons into chat checkout, and the bill can feel sharper than the feed.

Doubao PR (summarized in follow-up 36Kr coverage) stresses: no paid promo / bid-for-rank for now—mainly a post-purchase channel fee, not OTA-style “commission + ads.” Product narrative: sell outcomes first, not shelf space. Users do not see an obvious auction shelf in chat; merchants save ad-slot budget short term and face a higher take rate on closed deals. Both stories are coherent; they land on different ledgers.

The “charging twice” debate grows here: users may already pay Doubao via membership or ecosystem, while merchants pay a higher take at checkout. 36Kr’s follow-up lays it out—fairness depends on whether Doubao is an independent acquisition channel or a split door inside the same Douyin system. If product definition holds “specific channel,” the premium rate stands; if users still feel “one Douyin family,” the four points read as a hike.

Business model: price first, volume later

Feed traffic guesses demand; chat lets users state time, place, preference—closer to search-ad intent density. ByteDance refuses cheap scale and prices dialogue in the upper travel-channel band.

Why can intent density support a higher take? Feed users often “just browsing”; chat users already stated constraints—dates, district, budget, kids or not. The path to conversion is shorter; some comparison and decision cost is absorbed in dialogue. The bet: higher conversion × higher rate can beat “cheap volume first.” If conversion does not actually pull away, 12% only scares supply off and empties the chat shelf—the most direct rebound of price-first strategy.

Volume is still small: ~200M daily users and ~¥10M daily e-commerce GMV in the cited snapshot means pennies per user per day. The raise is less about today’s cash than turning Doubao into an asset with a contracted price—every future order settles at that rate. In finance language: lock take rate, then bet the GMV curve. In product language: do not wait until volume is large to teach the market a high price—merchant bargaining power is stronger then, and hikes hurt more.

The same day, Qwen’s open platform courted service partners—36Kr’s contrast: ByteDance already has local-life supply; Alibaba’s supply is fragmented, so it opens free. Rent vs invite is hand strength, not AI theology. Doubao can post 12% because Douyin Merchant already holds jumpable hotel inventory and fulfillment; if Qwen charged first, merchants might have nothing to sell. Pricing power attaches to a supply loop, not model parameters.

Counter-intuitive: many assume AI doors must grow free, then raise. ByteDance flipped it—channel still small, rate already in the high travel band. Upside: avoid locking future margin into a low-price habit. Risk: merchants raise prices, hide cheap SKUs, or simply skip Doubao. If chat rates look systematically high, users bounce to Meituan/Ctrip—intent capture becomes lead-gen for rivals.

Merchant micro-moves reshape the product too. Raising prices or hiding cheap SKUs makes chat recommendations look “always a bit expensive”; skipping Doubao thins what the model can push. If the platform props conversion with vouchers, it hits “subsidy in the fee base”—the harder the promo, the worse the rate feels. A named channel makes that visible for the first time: under merged 8% organic, the conflict was averaged away; every “Doubao” settlement line is now a small price negotiation.

For AI-entry builders, the case also draws a boundary: you can usually take rate only on the segment you close or bind to settlement. Pure recommend, off-platform jump, someone else cashiers—you end up on lead or ad fees, not a durable 12% checkout take. ByteDance can post that price because fulfillment and settlement after the jump still sit in Douyin Merchant—intent captured in Doubao, money settled in Merchant.

Timing matters too: policy posted July 27, effective August 10—about two weeks for merchants to read rules and adjust quote/inventory. That “publish then enforce” window is itself a mechanism: the platform watches supply reaction before settlement flips overnight. If hotels pull dialogue-sellable rooms in the window, the platform can still soften with messaging or vouchers; if supply barely moves, 12% is already treated by some merchants as acceptable channel cost.

Sources

Comments0

No comments yet

Doubao local-life rates and AI intent traffic pricing | Clover Startup