China Southern forecasts demand with AI and adds an AI travel assistant to its app
China Southern Airlines uses AI forecasting models developed with Huawei to predict passenger demand and plan its route network, with accuracy above 90% in its largest markets by Huawei's account. Its app now has an AI travel assistant, and AI translation screens help staff serve passengers at PKX and URC.
The operating picture
Planning the network with AI
Hamburg-Finkenwerder Airport China Southern Airlines Airbus A321-251NX B-32PU (DSC03313). Photograph: MarcelX42 / CC BY-SA 4.0 (cropped to fit). Illustrates the airline, not an AI deployment.
China Southern presented its AI applications at an innovation conference in Guangzhou in December 2025, including the Tianji forecasting model, the Tiance scheduling solver and the Tiandun safety model, Guangdong news reported. Its fourth innovation awards recognised 77 solutions across operations, marketing, safety and service.
The two featured implementations sit at either end of the business: network planning and the passenger's phone.
Selected implementations
The work behind the evidence
01 / Deployed
Forecasting demand to plan the network
The operational problem
Deciding which routes to fly and how much capacity to put on them depends on forecasts of demand months ahead, across hundreds of city pairs.
What the evidence reports
Huawei said that China Southern used its Pangu models to forecast passenger traffic and average fares 18 months ahead with 90% accuracy, AI Magazine reported on 10 October 2025. A Huawei case study describes the Tianji Prediction Model, developed jointly with the airline, which combines China Southern's production data with economic, demographic, weather, high speed rail and competitor factors. Its forecasts feed a route planning algorithm built on Huawei's OptVerse AI solver. Huawei puts accuracy above 90% for markets with millions of passengers and above 80% for markets with hundreds of thousands.
Operational significance
Network planning is among an airline's most valuable decisions, and linking AI forecasts directly to route and capacity optimisation puts AI in that decision.
Evidence still needed
The accuracy figures are Huawei's, and neither source says how accuracy is measured or how planners use the output. China Southern has not published revenue effects.
Huawei case study - smart airline (Tianji / OptVerse) · Evidence grade C
A source can verify an announcement without verifying its projected outcome.
02 / Deployed
A travel assistant in the app, translation at the airport
The operational problem
Booking, changes and check-in involve several screens in an airline app, and passengers who do not speak Chinese struggle to communicate with ground staff.
What the evidence reports
Guangdong news reported on 19 December 2025 that the Smart Travel Assistant in the China Southern app handles booking, check-in, changes, seat selection, services after the trip and finding the way around the airport through natural language. Project leader Deng Zhuonan said it can complete several services from a single command. AI translation screens deployed at Beijing Daxing (PKX) and Urumqi (URC) translate passengers' questions for staff and the replies back. Project lead Meng Yutong said passengers "can ask questions in their own language".
Operational significance
The assistant moves routine tasks into a conversation, and the translation screens address a practical gap at airports with many international passengers.
Evidence still needed
No usage figures have been published for either service, and the report does not say how many airports have the translation screens beyond PKX and URC.
A source can verify an announcement without verifying its projected outcome.
The wider register
Other tracked activity
Other records are shown separately, including non-AI technology and programmes at different stages of deployment. A register entry does not establish a measured AI benefit.
No other publicly listed records.
Editorial assessment
China Southern's AI reaches from network planning to the check-in desk, much of it developed with Huawei.
Independent evidence on forecast accuracy, and usage figures for the app assistant, would show how far the AI shapes the airline's results.
Live data from the Aerostack register. Status reflects published evidence only. See Our approach for how records are graded.
Aerostack register · 2 tracked deployments
Aerostack tracks AI deployments across airports and airlines, with the published evidence behind each record. Photography is real aviation imagery. Record thumbnails may show the named airport or airline; other photographs do not.
.jpg)