Dating apps UGC ads by the numbers
- Winning ads
- 337from 7 brands
- Median run
- 91 daysall winners: 85
- Median copies
- 5all winners: 4
- Proven (90+ days)
- 51%16% still live
337 ads. 98% link to an app store, 2% to a website. Median video length 20 seconds (214 ads with a length). Data as of 9 November 2026: 337 winning ads from 7 brands, first started 16 July 2025.
What the winning dating apps ads have in common
- They run for months, not days. The median dating winner ran 91 days (all 1,922 winners: 85). 51% passed 90 days and 8% passed 180.
- Brands make more than one version. The median winner exists as 5 copies in the Ad Library (all 1,922 winners: 4), and 56% reached 5 or more: new cuts, captions or sizes of the same video.
- The copy stays, the creator changes. 89% of these ads share their exact ad copy with another ad from the same brand (all 1,922 winners: 69%). Brands keep a line that works and run it under several creators' videos.
- Talking head leads. 50% are talking head videos and 32% POV text. That is above the 42% share across all winners.
- A person is on screen. 90% show at least one person; 61% feature a woman and 18% a man.
- The ad copy is short. The median caption is 90 characters and 17% ask a question up top.
- Most send people to an app store. 98% link to the App Store or Google Play, so the video's job is the install.
- They are not only in English. 41% are in another language, led by Dutch, French and Chinese.
- One promise per ad. The copy is a single line against a pain: endless swiping, ghosting, chatting that never leads to a date.
- Street interviews scale. 15 winners are street interviews ("What does he do for a living?"), and the longest-running of them were copied into 15 versions.
- Localised, not translated. 41% run in other languages, led by Dutch, French and Chinese, each with local creators rather than dubbed English videos.
Brands behind these ads
| Brand | Winning ads | Median run | Longest run | Median copies | Top format |
|---|---|---|---|---|---|
| Breeze | 253 | 91 days | 454 days | 6 | Talking head |
| Hinge | 50 | 81 days | 127 days | 4 | Talking head |
| Coffee Meets Bagel | 24 | 132 days | 243 days | 1 | Talking head |
| Hunch | 6 | 130 days | 179 days | 1 | POV text |
| CheatEye | 2 | 269 days | 269 days | 2 | POV text |
| Paired | 1 | 138 days | 138 days | 2 | Review |
| Hily | 1 | 52 days | 52 days | 8 | Talking head |
Download this table as a CSV file (CC BY 4.0, link to this page). Brands with a linked name have a full UGC Brands page on how they run their creator program.
How we picked these ads
A winning ad here is a UGC-style video ad from the Meta Ad Library (Facebook and Instagram) that the brand ran for at least 45 days, that sits in the top 10% of that brand's video ads by impressions, and that the brand either scaled to 3 or more copies or kept running past 60 days. Proven winners ran 90 days or more. Each ad is checked as the brand's own: run by its verified page and linking to its site or app.
The Ad Library shows no spend, clicks or sales, so the signal is behaviour: a brand that keeps paying to run a video for months, and copies it into new versions, is telling you it works. "Copies" counts the near-identical ads the library groups under the same creative. Format (talking head, demo, POV text and so on), who is on camera and language are labelled from each ad's poster frame and copy by a model, then spot-checked; treat them as approximate.
Data as of 9 November 2026: 337 winning ads from 7 brands, first started 16 July 2025. Run days for live ads are counted to the export date, so they will grow. This page shows a sample picked for variety (the brand's longest-running ads first, a cap per brand and one version per line of copy). We never show the creator's name or handle; each card links to the public Ad Library entry.
These are observations about ads brands kept running, not proof that a format causes results. A brand that runs talking heads will have talking-head winners. Compare shares with the "all winners" figures before drawing conclusions.


