Researchers: Google searches about symptoms, like loss of smell, can help find places where positive COVID-19 cases were missed and reveal undetected symptoms
They can also reveal symptoms that at first went undetected. I may have found a new one. — Mr. Stephens-Davidowitz is the author of …
Context & Ripple Effects
This study extends a line of pandemic-era work repurposing consumer data streams as epidemiological sensors: Kinsa Health's smart thermometers had already shown that everyday devices could flag flu-like spread before official counts, and weeks later researchers demonstrated an algorithm predicting outbreaks from Google, Twitter, and smartphone data. What is new here is the granularity — symptom-level search queries like loss of smell acting as a map of where confirmed cases were missed.
Google's own infrastructure makes the signal actionable: since 2016 it has listed possible health conditions directly in symptom searches (symptom results on mobile), and during the pandemic it began surfacing testing-center information for over 2,000 sites across 43 states alongside its dedicated COVID-19 site. The same query stream that serves users can, in aggregate, tell public health officials where testing fell short.
First-order effects
- Public health officials gain a free, near-real-time indicator of where positive cases were likely undercounted — regions with anomalous loss-of-smell searches become priority targets for expanded testing rather than waiting for case reports to accumulate.
Second-order effects
- Google's search product becomes de facto public health infrastructure: the company already surfaces testing centers and condition information in results, so aggregated symptom data gives it both the data and the distribution channel to shape where testing capacity goes.
Third-order effects
- If search-based signals prove out, outbreak detection shifts from reporting-lag-bound official counts toward passive digital surveillance — raising the standing question of how much epidemiology should run through a handful of platforms that hold the query logs.
The trend: Pandemic response is turning consumer digital exhaust — search queries, thermometer readings, social streams — into a shadow early-warning system that runs ahead of official case counts.