Across 110 Billboard No. 1s from 2004 to 2026, songs that spent just one week at No. 1 lost search interest at the same speed as long-running hits: all four groups fell below a quarter of their peak within a median 4 months. A year after the peak, one-week No. 1s held a median 5.8% of their peak, against 3.2% for 2–4-week runs, 3.3% for 5–9 weeks and 4.9% for 10+ weeks. Five years on, every group sits at about 1–2%. Because each song is measured against its own peak, a long run means a higher, flatter peak; the drop afterwards looks just as steep.
Decay by weeks at No. 1 (medians, % of each song's own peak)
| Weeks at No. 1 | Songs | Months to <25% | Year-after level % | Year-5 level % | Songs with 5 yrs of data |
|---|---|---|---|---|---|
| 1 week | 58 | 4 | 5.8 | 2.1 | 44 |
| 2–4 weeks | 24 | 4 | 3.2 | 1.1 | 21 |
| 5–9 weeks | 18 | 4 | 3.3 | 1.2 | 18 |
| 10+ weeks | 10 | 4 | 4.9 | 1.9 | 8 |
By decade-ish era
| Era | Run | Songs | Year-after % | Year-5 % |
|---|---|---|---|---|
| 2004–07 | 1 week | 10 | 5.9 | 2.3 |
| 2004–07 | 2+ weeks | 23 | 3.1 | 1.6 |
| 2008–11 | 1 week | 16 | 4.6 | 1 |
| 2008–11 | 2+ weeks | 8 | 4.7 | 1.1 |
| 2012–15 | 1 week | 3 | 5.4 | 2.8 |
| 2012–15 | 2+ weeks | 9 | 3.7 | 1.1 |
| 2016–19 | 1 week | 12 | 8.9 | 2.8 |
| 2016–19 | 2+ weeks | 7 | 2.7 | 1 |
One-week No. 1s with the longest tails (year-5 level, % of peak)
| Song | Artist | Year | Year-5 % |
|---|---|---|---|
| You're Beautiful | James Blunt | 2006 | 14.2 |
| Girlfriend | Avril Lavigne | 2007 | 11.6 |
| Psycho | Post Malone featuring Ty Dolla Sign | 2018 | 10.6 |
| The Way You Move | OutKast featuring Sleepy Brown | 2004 | 7.3 |
| That's What I Like | Bruno Mars | 2017 | 4.9 |
| What Goes Around...Comes Around | Justin Timberlake | 2007 | 4.7 |
| Sunflower | Post Malone and Swae Lee | 2019 | 4.6 |
| Watermelon Sugar | Harry Styles | 2020 | 4.5 |
What it means
Every song on these charts started as a hook someone wrote down. If you have a hook of your own, an AI song generator can turn it into a full track with vocals in a few minutes. For a soundtrack to your own day, an AI music generator can score it to whatever mood you need.
Method
- Google search interest, US, web search, one query per song for "<title> lyrics", Jan 2004–Sep 2026 (monthly points). Each song's series is scaled to its own peak (=100), so the metrics are shapes: share of a song's own peak, not volume.
- Sample: 143 of the 300 Billboard Hot 100 No. 1s first reaching No. 1 between Jan 2004 and 2026 (all 16 Songs of the Summer 2010–2025 included). Chart dates from the public Billboard No. 1 list.
- Ambiguous titles removed (33 songs, leaving 110): a song is dropped if its all-time search peak falls more than 12 months from its first week at No. 1 (another song or meaning dominates, e.g. "Seven", "Up", "Rockstar"), or if searches in months 13–36 before its peak already averaged ≥10% of the peak (common phrase, e.g. "So What", "Savage"). For 2004–2006 hits, a hand list of generic titles ("Burn", "Yeah!", "Run It", "I Believe") was dropped.
- Year-after level = mean of months 12–23 after the peak month; year-5 = months 60–71.
- "One-hit wonder" is defined by weeks at No. 1.
Cite and download
Free to use and cite with a link back to this page.
aisong.lol Research. (2026). The one-hit-wonder curve: how long the world keeps searching for a song after it leaves the chart. aisong.lol. https://aisong.lol/research/the-one-hit-wonder-curve-how-long-the-world-keeps-searching-for-a-song
Use this data. Every table in this study as one spreadsheet-ready file.
↓ Download the data (CSV)
