When does flu season start and end?
"October through May" is technically accurate as an average — but it's the kind of answer that's useless when you're trying to make a decision in January. Flu seasons vary by four to six weeks in either direction, peak at different times in different parts of the country, and can be mild or severe independent of timing. Here's how to actually read a season as it unfolds.
How the CDC defines flu season
The CDC tracks flu activity using a surveillance system called ILINet — a network of thousands of outpatient providers who report the percentage of visits attributed to influenza-like illness (ILI) each week. Flu season is considered "active" when ILI activity rises above the national baseline of approximately 2.5%.
The official flu season runs from week 40 of one year (early October) through week 20 of the next (mid-May). This is the reporting window, not a guarantee that flu is active the whole time. Most seasons see meaningful activity for 12 to 16 weeks within that window, with a clear peak somewhere in the middle.
The CDC also tracks flu seasons by severity — not just when they happen, but how bad they are — using hospitalization rates, ILI peaks, and pneumonia and influenza mortality data. A season can be early and mild, late and severe, or any combination. Timing and severity are largely independent.
Month by month: what a typical season looks like
Why seasons vary so much
The timing and severity of flu seasons are notoriously difficult to predict, even for the CDC. Several factors drive variation:
Which strains dominate. Influenza A (particularly H3N2 subtypes) tends to cause more severe seasons than Influenza B or H1N1. H3N2-dominant seasons often peak earlier and hit older adults harder. The specific strains circulating in a given year affect both timing and who bears the burden.
Vaccine match. When the vaccine strains chosen in February (for the following winter's vaccine) match the circulating strains well, more of the population is protected and spread is slower. In mismatched years — when the dominant strain drifts after the vaccine is formulated — more people get sick, activity rises faster, and peaks can be higher.
Population immunity. A severe previous season leaves more people with natural immunity to similar strains, which can reduce the following year's activity. A mild season does the opposite.
Temperature and humidity. Influenza viruses survive longer and spread more efficiently in cold, dry air — which is why flu is a winter phenomenon in temperate climates. Years with early cold snaps in the South tend to see earlier season onset; mild Decembers can delay the peak.
Travel and behavior. Holiday travel reliably accelerates spread. School calendars matter too — kids are major vectors, and the return from Thanksgiving and winter breaks consistently shows up in ILI data two weeks later.
How recent seasons have compared
| Season | Peak month | Dominant strain | Severity | Notable |
|---|---|---|---|---|
| 2017–18 | December–January | H3N2 | Very High | One of the worst in a decade; vaccine poorly matched to H3N2 |
| 2018–19 | February | H1N1 / Flu B | Moderate | Late season; Flu B dominant in second half |
| 2019–20 | January | H1N1 | High | Season ended abruptly in March with COVID-19 mitigation measures |
| 2020–21 | — | Minimal circulation | Very Low | Masks and distancing nearly eliminated flu; historically anomalous |
| 2021–22 | February–March | H3N2 | High | Late, compressed season; H3N2 dominant; children heavily affected |
| 2022–23 | December | H3N2 / Flu B | High | Early, severe season; tripledemic with RSV and COVID |
| 2023–24 | January–February | H1N1 | Moderate | Moderate severity; vaccine reasonably well-matched |
| 2024–25 | January | H3N2 / H1N1 | High | Above-average season; dual A subtypes circulating simultaneously |
How the CDC tracks the season in real time
The CDC publishes a weekly flu surveillance report called FluView, released every Thursday for the prior week. It includes national and regional ILI percentages from ILINet, clinical lab data on strain typing and test positivity, hospitalization rates from a network of hospitals across the country, and mortality data from cities and states.
This site pulls from the same underlying data via the Delphi CMU Epidata API, which mirrors the FluView numbers in near-real-time. The activity level you see on the homepage reflects the most recent week available — typically data through the prior Saturday, published the following Thursday.
The 7-day data lag
One thing worth understanding about all flu surveillance data, including what this site displays: there is always roughly a one-week lag between real-world conditions and reported numbers. Providers report to ILINet weekly, the CDC processes and publishes Thursday, and this site reflects those published numbers. If flu is exploding in your city today, the data won't show it until next Thursday.
This is a property of the surveillance system, not a flaw in how we display it. The practical implication: during a rapidly rising season, the numbers may understate how bad conditions are right now. Treat the current activity level as a floor, not a ceiling, when the season is actively accelerating.
The best leading indicator during a rising season is clinical lab positivity — the percentage of flu tests coming back positive. This number rises ahead of ILI percentages because it reflects testing patterns in real time. It's shown on the homepage alongside the ILI activity level when the season is active.