AI heating uses 18% less gas in Guardian journalist’s two-winter home test

A warmer home did not mean a bigger gas bill in Samuel Gibbs’s two-winter heating test. Writing in the Guardian on November 7, the consumer technology editor reported that a Tado X system with AI features enabled used 18% less gas than his strict manual setup. The result came from one home, not a multi-household trial.

The starting point was already frugal

Gibbs used his 1930s solid-brick house, which has a newer insulated loft conversion. His original system paired a 10-year-old Worcester Bosch gas combination boiler with a hallway Hive thermostat, its smart features disabled, and 13 radiators with standard thermostatic valves. Those valves let him limit heat in individual rooms, but adjustments were manual.

He scheduled heating for a couple of hours each morning and evening, shut off radiators in unused rooms and turned the heating off when leaving. The routine was laborious and sometimes left family members cold. It was also a demanding baseline: Gibbs said consumption was already about 10–15% below earlier years, when he had been less restrictive.

Across 213 days from November to May, that manual regime used 14,316.7 kilowatt-hours of gas for heating. Gibbs estimated daily hot-water consumption from summer 2025 meter readings, when the heating was off, and used that baseline to separate hot water from heating demand.

From one hallway thermostat to room-by-room control

In summer 2025, he replaced the controls with Tado X while keeping the boiler. The installation included 13 smart radiator thermostats, wireless thermostats in the two largest rooms and a programmer wired to the boiler. Tado lent him the equipment.

Each smart radiator valve can request heat from the boiler and regulate the hot-water flow into its radiator. Rooms that need no heat keep their radiators off. Compatible boilers can also receive instructions to lower the water’s flow temperature through a connection standard called OpenTherm. Gibbs’s boiler did not support that feature.

The subscription adds automation that learns how quickly rooms warm and cool, shuts off heating when an open window is detected and responds to residents leaving or returning. Gibbs enabled all the AI features and set room schedules. During the first week, the energy-tracking feature flagged a radiator heating an uninsulated porch; he then cut its schedule substantially.

Lower consumption, but an upfront bill

After the porch adjustment, Gibbs said the system ran without further intervention. Rooms warmed faster and the house felt more comfortable, even though the boiler seemed to run more often. He recorded monthly meter readings but avoided calculating consumption during the test, resisting the temptation to intervene.

The gas reduction equated to £205.70 at the energy price cap cited in his article, which runs through December 31. That is an estimate at those rates, rather than a fixed annual saving. The tested setup cost about £930, with roughly £80 more for an engineer to wire the boiler connection.

The AI features require £3.99 a month or £29.99 a year; monthly subscriptions can be paused when heating is not needed. The comparison also changed radiator controls, schedules and automation together. It therefore measures the complete upgrade in this house, rather than isolating how much of the reduction came from AI alone.