Researchers from the tech giant retraced the digital footsteps of millions of Bing search-engine users to see whether or not certain early queries could predict an eventual pancreatic cancer diagnosis. They started off by examining anonymized search-engine data clearly associated with those who have the disease and then followed the online trail back to each user’s previous searches about more low-key, but concerning, potential symptoms.

Click here to subscribe to our new Brainstorm Health Daily Newsletter.

The study authors found that there was some promise to the technique. “We showed specifically that we can identify 5% to 15% of cases, while preserving extremely low false-positive rates,” they wrote. “Signals in search logs show the possibilities of predicting a forthcoming diagnosis of pancreatic [cancer] from combinations of subtle temporal signals revealed in the queries of searchers.”

The next challenge will be harnessing such information to potentially warn at-risk users, Microsoft’s Eric Horvitz told the New York Times. “The question is, ‘What might we do? Might there be a Cortana for health some day?’” he said, referring to Microsoft’s equivalent to Apple’s (AAPL) Siri assistance software.

It’s unclear whether such early-risk detection techniques will be effective beyond certain specialized diseases, given that illness symptoms can be quite disparate. The most researchers may be able to hope for is suggesting that certain users check in with a medical professional.