Sniffer dogs could contribute to efforts to prevent the spread of Covid as society reopens, according to scientists.
As part of a trial, dogs were trained to recognise a distinctive odour produced by people with the virus, but undetectable to the human nose.
This could come in useful for screening at airports or mass events.
But the dogs' findings would have to be confirmed by lab testing, the researchers said.
Although the dogs correctly picked up 88% of coronavirus cases, they also incorrectly flagged 16% of people who didn't have Covid as having the virus.
Dogs can have up to 100,000 times the smelling ability of humans and have long been used to sniff out drugs and explosives.
Recent research has shown dogs - particularly breeds like spaniels and retrievers - can detect the unique scents of diseases including cancer, Parkinson's and malaria.
As part of the current canine screening trial, six dogs were trained to recognise the smell produced by people with Covid-19 using worn socks, face masks and t-shirts of various materials.
They were rewarded with treats when they correctly identified whether the sample was from an individual who had tested positive or negative.
Some of the people in the negative group had common cold viruses, to make sure the dogs were able to distinguish Covid from other respiratory infections.
The dogs were able to sniff out the disease even when it was caused by different variants, and when the person had no symptoms or only had very low levels of the virus in their system.
Claire Guest, Chief Scientific Officer at charity Medical Detection Dogs, which trained the animals, said the results were "further evidence that dogs are one of the most reliable biosensors for detecting the odour of human disease".
They picked up roughly 88% of positive cases - meaning, for every 100 cases, the dogs failed to recognise just 12 infected people.
But out of 100 people who did not have Covid, the dogs wrongly suggested - via the sniff test - that 16 of them were infected.
So if one person on a plane of 300 passengers has coronavirus, the dogs are likely to correctly identify that person, but may also wrongly indicate that another 48 Covid-negative people are infected.
It means a proportion of infections will be missed, and more will be told they have the virus when they don't. This is the case for all tests to different extents, but the canine method incorrectly tells a lot more people they have the virus than the type you swab up your nose.
So the research team does not recommend dogs alone are used to sniff out positive cases.
But they believe the dogs could be an additional screening tool alongside more conventional tests. They say dog screening, followed by swab testing, will pick up 91% of infections.
The real potential advantage, though, is speed: even the quickest tests take 15 minutes to show a result, while dogs can sniff out the disease in seconds.
Two dogs could screen 300 people in half an hour, researchers say.
This could make the sniff test "a suitable method for mass screening", argues Prof James Logan at the London School of Hygiene & Tropical Medicine, which conducted the research alongside the University of Durham.
In theory, people could be screened as they queue for a flight or to enter an event, and anyone flagged up by the dogs would need to take a PCR test - the more accurate type of swab tests which is processed in a lab.
This could cut down on the numbers having to enter hotel quarantine.
Dogs could also potentially be used in areas where there isn't currently much screening, such as busy train stations, to help prevent a super-spreading event.
The research is at an early stage so it still needs to be reviewed by other scientists before it can be published and, in the next phase of the study, tried out on infected people - rather than bits of sock.