The image provided with this headline is pretty misleading. The dogs in the study were smelling saliva or tracheobronchial secretions, not people's skin/cloths. This definitely has an effect on how useful dogs would be as a screening tool. Dogs do have a great sense of smell though.

I believe that CNBC got a bit over-enthusiastic about this one. The headline said dogs could sniff out “infections” but the paper itself said dogs differentiated between saliva sample from *hospitalized* patients and others. A person who has been hospitalized may well have many differences from a person on the street, as well as a much higher viral load of COVID than an ordinary infection. The paper itself notes that there were large differences in dogs' performance, which raises concerns about repeatability and sensitivity to more subtle levels of infection. The paper concludes by saying the findings are preliminary and much more work is needed before this could be used in the field. Because the reported results seem likely to be so heavily impacted sample acquisition, I believe some of the fault of over-hyping the results lie with the original study authors. I think the biggest fault is that CNBC was credulous, a button that I'd love to see Credder include. The likelihood that anything comes of this seems quite small because I don't doubt that dogs can smell differences between people who've been receiving heavy-duty (oxygen, drugs, bed rest) medical care but can't imagine how lower viral loads and non-hospital provenance of samples will achieve anything close to the 94% accuracy claimed.