The covert battle over AI ride-hailing entrances between Didi, Qianwen, and Doubao

The so - called new opportunities in the era of disruption often involve re - measuring the old territory with new technologies.

In 2014, Didi and Kuaidi launched a subsidy war on the streets of Beijing. This not only reshaped the way Chinese people travel but also initiated the most important payment entry enlightenment campaign in the history of China's mobile Internet.

Twelve years later, a similar scenario unfolds again. In just a few months, Didi, Qianwen, and Doubao have successively invested heavily in the "AI taxi - hailing" function.

On the surface, this is an interaction upgrade that is not overly impressive. Users can simply state their needs, and the system will automatically analyze the intention and dispatch orders.

However, looking beyond the homogeneous technological appearance of voice recognition and large - model invocation, we can find that this is actually a fierce and covert battle centered around real - world data anchor points, the distribution rights of full - scenario life services, and preventing oneself from being completely reduced to a pipeline.

Taxi - hailing is never just a travel business. It is also the most cost - effective and efficient market education tool, as well as the core fulcrum for leveraging the next - generation traffic entry.

At the beginning of the mobile Internet era, taxi - hailing supported the duopoly pattern of mobile payment. Now, taxi - hailing is becoming the springboard for general AI to enter real life.

The Resonance of History: Why Taxi - Hailing is the Optimal Enlightenment Entry

The Didi - Kuaidi war in 2014 solved the problem of why users should bind their bank cards to a social software.

Breaking the mental barriers cannot rely on preaching. It can only be achieved through a scenario that is light enough, high - frequency enough, and provides a tangible experience, allowing users to complete their first action without any burden.

Taxi - hailing perfectly meets all these conditions. The unit price is low, and a few yuan in subsidies can encourage users to try. The feedback is immediate; users can enjoy the service right after payment and feel the value instantly. It is used frequently, with users in first - and second - tier cities taking taxis at least twice a week, which is enough to develop muscle memory in the shortest time.

The outcome has already been written in the industry history.

On February 7, 2014, the first working day after the Spring Festival, the peak order volume of Didi Taxi reached 2.62 million, among which the peak order volume of WeChat payment exceeded 2 million. The billions of subsidies burned by Didi, Kuaidi, Tencent, and Alibaba not only brought prosperity to the online car - hailing industry but also popularized mobile payment across the country in China.

Taxi - hailing is just a teaching tool, and payment is the real test.

After twelve years, the large - model industry today is at the same crossroads.

After years of iteration, the parameters and reasoning ability of large models are no longer bottlenecks. The real shortcoming of the industry lies in the narrow user mindset and the lack of practical application scenarios.

The majority of ordinary users' understanding of AI still remains at the level of online toys for writing copy, making PPTs, and chatting for fun, without generating the need for AI to be applied in real life.

No matter how much you talk to users about general artificial intelligence, multi - modality, or agents, it's not as effective as a tangible real - life experience.

Taxi - hailing is the optimal solution for AI to conduct national mental enlightenment.

Users don't need to learn prompt skills, understand the principle of large models, or even know that they are using AI. They can complete all operations just by speaking.

The trial - and - error cost is almost zero, and the price is the same as that of ordinary taxi - hailing. The value feedback is highly reliable; users will get results within a few minutes regarding whether the address is accurate and how long it will take for the car to arrive. Coupled with its essential and high - frequency nature, it is enough to implant the perception that AI can handle things into users' behavior habits in the shortest time.

In May 2026, the national online car - hailing order volume was 977 million, which is quite an attractive transaction frequency.

Of course, it must be admitted that the offline transformation of AI taxi - hailing is limited. Core issues such as insufficient capacity during peak hours, fluctuating dynamic pricing, and lack of service standardization remain untouched. The optimization of online interaction is also minimal, only reducing the number of screen taps, eliminating the need to manually enter addresses, and avoiding the need to compare prices across multiple apps.

However, even so, the significance of AI taxi - hailing is far more than just minor functional improvements. It is a battle to win over users' minds, replicating the historical path of mobile payment.

Taxi - hailing is a touchstone. It pries open not only a corner of the travel market but also a crack for general AI to enter the daily lives of the public.

Three Players Enter the Game: Non - overlapping Tracks, Gambling with Different Intentions

Different from the direct confrontation between Didi and Kuaidi back then, the three players entering the AI taxi - hailing market today seem to be targeting the same function, but in fact, they come with completely different entry tickets and are heading for different destinations. This is not a price war at the same level but an ecological position - taking battle where each side meets its own needs.

Didi is a conservative, evolving from a target to a player.

Twelve years ago, Didi was the entry carrier for payment giants to compete for. Riding on the wave of the subsidy war, it grew into the absolute leader in the travel industry. However, for Didi today, it is an unavoidable survival defense war, and it must enter the AI field on its own.

If users get used to hailing a taxi with a single command on the AI assistant in the future and no longer actively open the Didi App, the brand assets that Didi has accumulated over the decades will quickly weaken, and it will gradually degenerate into a back - end capacity provider, just like how banks became mere fund channels after the popularization of mobile payment.

Losing direct contact with users means losing the pricing power, brand premium, and initiative for business expansion, ultimately becoming a factory that earns meager profits.

Therefore, Didi's core goal in developing AI is to hold on to the user entry and keep the initiative in interaction in its own hands. Its chips are the densest capacity network in the country, the most mature dispatching system, and the established travel user mindset.

For Didi, AI is an efficiency tool for its existing business and a moat to prevent itself from being reduced to a pipeline.

Qianwen is a connector for an ecological closed - loop.

Alibaba's foray into AI taxi - hailing aims to revitalize its entire set of life - service ecosystems. This can be regarded as a grand review of the full AI transformation of the consumer ecosystem. Gaode's maps and capacity, Ele.me's local catering, Fliggy's hotel and travel services, Taobao's e - commerce, and Alipay's payment links are scattered across different apps. In the past, users had to jump between apps, and these services never formed a synergy.

AI taxi - hailing is the connector for the entire ecosystem.

When a user says, "Go to XX shopping mall for hot pot at 7 p.m.," the AI can simultaneously complete restaurant selection, group - buying, car - hailing, navigation, and even recommend post - meal entertainment, all within the conversation flow without the need to jump to any other apps.

Taxi - hailing is the lowest - threshold and highest - frequency contact point, used to educate users that AI can handle an entire itinerary, ultimately locking users firmly within Alibaba's payment and service closed - loop.

Doubao is a game - changer aiming for a breakthrough.

ByteDance's entry into the market has the same offensive flavor as WeChat Pay back then. Doubao's model capabilities are on par, but it has been trapped in the circle of content - creation tools. By entering the taxi - hailing market, it is following the same path as WeChat Pay to break into new markets. Using a necessary scenario, it pulls AI from a creation tool into a life - service entry. First, it pries open users' minds, and then gradually penetrates into full - scenario services such as ordering food, booking tickets, hotels, and group - buying.

ByteDance holds a large amount of public - domain traffic from short - video platforms and search engines. Once users' minds are opened, the efficiency of traffic conversion will be extremely high.

To put it more deeply, what Alibaba and ByteDance are gambling on is not the 411.9 - billion - yuan taxi - hailing market but the chief entry position in the entire life - service AI field. This is a high - risk, high - reward gamble. Taxi - hailing is the ticket, and the ecological feast after entering the market is the real goal.

The three players have three different strategies. One is to defend the travel market, one is to connect the ecological closed - loop, and one is to seize the entry position. Although they seem to be competing in the same arena, they actually have their own battlefields and survival lines.

The Battle for the Entry: The Transfer of Distribution Rights and the Underlying Logic of Winner - Takes - All

The real battle for the entry is never about the access path but the distribution rights of services. This is the core essence of this war.

In the past decade or so, the logic of the Internet was that users sought services. Users first had a clear need to take a taxi and then actively opened apps like Didi, Gaode, or Baidu Maps, and the platforms received the orders. During this stage, the power was in the hands of vertical service platforms, and the entry was just a channel for attracting traffic.

However, AI taxi - hailing has completely reversed this logic, making intention the first step and service the subsequent one.

Users don't need to decide which platform to use for taxi - hailing first, or even be clear that they want to take a taxi. They only need to express their final destination, and the AI will handle the rest, including address recognition, vehicle - model matching, capacity selection, and order - placing and payment.

This means that the distribution rights of services have completely shifted from vertical travel platforms to AI entries.

Whether the AI assigns the order to Didi or Caocao Chuxing, users won't notice the difference and don't really care. In the long run, the brands of traditional travel platforms will be infinitely weakened, and they will eventually degenerate into invisible back - end capacity providers.

An even more core chip than the distribution rights is the dual interception of data and transactions.

Travel data is the highest - quality real - behavior anchor point among all life - service data. It comes with time, geographical location, consumption ability, and travel trajectory, and is deeply bound to users' daily lives. Once the AI obtains this data, it can draw an accurate user life profile, which in turn can optimize the service - matching ability and form a positive cycle.

This barrier is something that pure online large models can never catch up with.

The closed - loop of the transaction link means that the entry party completely controls the initiative for commercial monetization.

Users make decisions and complete payments entirely within the dialogue box without jumping to the service platform. The bulk of the income from commissions, advertisements, and value - added services will ultimately be concentrated at the entry end.

It is also worth noting that AI entries are naturally highly exclusive. Just as users only keep one main input method and one main browser on their phones, they will only use one most convenient AI assistant in daily life. Once a path - dependence is formed where users speak out when they have a need, all life - service needs will be concentrated at this entry, ultimately resulting in a winner - takes - all situation.

This also determines that the outcome of this war does not depend on the travel industry itself.

Winning in the taxi - hailing scenario is just getting the entry ticket. The real decisive battle is who can extend the taxi - hailing usage habit to broader life - service scenarios such as catering, hotel and travel, and retail.

However, there is a huge hidden concern in this industry. The travel scenario involves strong offline operations, strict supervision, and heavy - service requirements, which are vastly different from the error - tolerance rate of pure online text tools.

Most large - model manufacturers do not have the genes for offline service operations.

Currently, Qianwen relies on Gaode's aggregated capacity as a backup, and Doubao uses Caocao Chuxing to handle the fulfillment. In essence, both are outsourcing their offline services. Didi is the only one that shoulders all the costs on its own, from model training to driver training, safety management, and complaint handling.

In the short term, the aggregated model is light and appealing. However, in the long run, when service consistency, safety bottom - lines, and emergency response become the key factors for competition, Didi, which has its own capacity and direct - management experience, is likely to widen the gap with players that rely on outsourcing.

This hidden factor will gradually emerge as AI taxi - hailing moves from the trial - use period to regular use.

For the time being, there will be no absolute winner in this AI taxi - hailing competition.

The long - term dividing line will appear in two dimensions. One is who can turn AI taxi - hailing from a novelty into a habit, forming a stable path - dependence. The other is who can perfectly balance the lightness of AI online decision - making and the heaviness of offline fulfillment.

AI taxi - hailing is not a travel revolution. In essence, it is just another stop in the Internet traffic war, evolving from text - and - image content to short - videos, from short - videos to live - streaming e - commerce, from live - streaming e - commerce to instant retail, and now to instant travel.

Twelve years ago, a taxi - hailing war brought mobile payment into the hands of hundreds of millions of people and initiated the golden decade of China's mobile Internet. Twelve years later, the same scenario is playing out again. This time, what needs to be popularized is using AI to schedule daily life.

A simple car - hailing command is a door.

On this side of the door is the mobile Internet era with numerous apps. On the other side is the next - generation life - service ecosystem dominated by conversational AI.

Technology changes its appearance every few years, but the story of competing for entries, data, and users' decision - making power remains the same.

This article is from the WeChat official account "TrueView", and is published by 36Kr with authorization.