METHODS: A lay person entered between 1-5 clinical features, for each of 24 frequently missed diagnoses, into both Isabel & Google.

RESULTS: In 7 / 24 [29%] cases Google did not include the missed diagnosis in 30 returned results. In 24 / 24 [100%] cases Isabel returned the missed diagnosis in 30 returned results.

CONCLUSIONS : Isabel performs with acceptable accuracy as a symptom checker for non-clinicians look for diagnosis possibilities.