Google's AI was asked about 15-minute cities. One in eight times, it treated the conspiracy as a fair fight.

Google's AI now argues the cookers' case for them.

Researchers at Queensland University of Technology fed Google's AI the exact language cookers use for two conspiracies, chemtrails and 15-minute cities, against a week of plain searches.

A 15-minute city is meant to be simple: shops, a school, a GP, within a short walk or ride of home.

Melbourne runs its own version, a 20-minute neighbourhood, policy since 2017.

Cookers don't hear "closer shops."

They hear zones, permits, cameras stopping them leaving their own suburb, a control grid dressed as good planning.

For chemtrails, the system has learned.

Google demoted the worst chemtrails results in 2015 after researchers called it out, and the AI barely mentions the theory now.

15-minute cities got no such lesson.

12.5 per cent of the AI's answers on that topic argued the cookers' side outright.

Asked "15 minute cities versus smart cities control," Google blurred convenience and surveillance into the same sentence, a coin toss between a planning term and a surveillance state.

We know exactly what that sounds like here.

In 2023, a group calling itself Concerned Citizens for Wyndham stood up in Melbourne's outer west and demanded the removal of traffic cameras they said were CCP spy equipment, installed under the 20-minute neighbourhood plan to track residents' movements.

The QUT team found the same script running through Google's AI three years later.

A planning idea, laundered into a surveillance plot.

Where it does debunk, researchers found it barely clears the bar.

It calls a claim misinformation and stops there.

No argument.

No evidence.

Nothing a cooker actually has to reckon with.

This is not a one-off glitch.

The same QUT researchers found in June that Google's AI summarised Philip Morris International as transitioning to smoke-free products, no mention of decades lying about smoking's harms.

"Essentially a regurgitation of Philip Morris International's PR materials," University of Sydney public health professor Becky Freeman called it.

The AI does not check who is telling the truth.

It checks who has the tidiest website.

That's a commercial problem for a tobacco company.

It's a public health problem for a conspiracy already primed by five years of lockdown grievance.

A German court has ruled Google can't hide behind "it's just search."

In May, Munich's regional court found Google directly liable for false AI Overview claims, its own statement, not a neutral link.

That case involved two publishers wrongly accused of running scams.

Nobody has tested that principle against a machine calling a bike lane a surveillance state.

Google says it invests heavily in AI Overview quality.

It has not explained why Wyndham's traffic cameras are still one search away from making their case all over again.