Background: An internet-based survey of influenza-like illness (ILI) - the Great Influenza Survey or GIS - was launched in the Netherlands in the 2003-2004 influenza season. The aim of the present study was to validate the representativeness of the GIS population and to compare the GIS data with the official ILI data obtained by Dutch GPs participating in the Dutch Sentinel Practice Network. Method: Direct mailings to schools and universities, and repeated interviews on television and radio, and in newspapers were used to kindle the enthusiasm of a broad section of the public for GIS. Strict symptomatic criteria for ILI were formulated with the assistance of expert institutes and only participants who responded at least five times to weekly e-mails asking them about possible ILI symptoms were included in the survey. Validation of GIS was done at different levels: 1) some key demographic (age distribution) and public health statistics (prevalence of asthma and diabetes, and influenza vaccination rates) for the Dutch population were compared with corresponding figures calculated from GIS; 2) the ILI rates in GIS were compared with the ILI consultation rates reported by GPs participating in the Dutch Sentinel Practice Network. Results: 13,300 persons (53% of total responders), replied at least five times to weekly e-mails and were included in the survey. As expected, there was a marked under-representation of the age groups 0-10 years and 81->90 years in the GIS population, although the similarities were remarkable for most other age groups, albeit that the age groups between 21 and 70 years were slightly overrepresented. There were striking similarities between GIS and the Dutch population with regard to the prevalence of asthma (6.4% vs. 6.9%) and the influenza vaccination rates, and to a lesser degree for diabetes (2.4% vs. 3.5%). The vaccination rates in patients with asthma or diabetes, and persons older than 65 years were 68%, 85%, and 85% respectively in GIS, while the corresponding percentages in the Dutch population were 73%, 85% and 87%. There was also a marked similarity between the seasonal course of ILI measured by GIS and the GPs. Although the ILI rate in GIS was about 10 times higher, the curves followed an almost similar pattern, with peak incidences occurring in the same week. Conclusion: The current study demonstrates that recruitment of a high number of persons willing to participate in online health surveillance is feasible. The information gathered proved to be reliable, as it paralleled the information obtained via an undisputed route. We believe that the interactive nature of GIS and the appealing subject were keys to its success.