UGC Ad Longevity Study (Oct 2026)

How the data was collected

Source. Meta Ad Library, US delivery, accessed through a third-party Ad Library API. Snapshot date: October 20, 2026.

Advertiser selection.

• 48 DTC brands (well-known UGC-heavy brands plus challengers found by keyword search).
• 72 non-DTC advertisers found through US keyword searches. Examples: business coaching, real estate investing, business funding, roofing, HVAC, mortgage, personal injury lawyer, med spa, dentist, AI app, budgeting app.
• Advertisers were chosen, not randomly sampled.

Active snapshot.

• Up to about 50–63 active ads per advertiser, sorted by impressions (4,973 ads).
• Days running = full days between the ad's start date and the snapshot date, counting neither day.
• 4 telehealth advertisers returned no viewable media and were left out of the format shares.

Launch cohorts (3,134 ads).

• Every video ad that delivered in two launch windows, active and stopped: Mar 15–Apr 20 and Jun 1–Jul 12, 2026.
• We kept only ads whose own start date fell inside the window.
• For stopped ads, days running = full days between the start date and the stop date, counting neither day.

Coding.

• 365 distinct video creatives were coded by format, from the first frame, and by hook type, from the first spoken line in the transcript, or the ad copy when there was no speech.
• 142 were 90+ day long-runners (up to 2 per advertiser). The remaining 236 cohort creatives were sampled at random within each advertiser; a few creatives fall in both groups, which is why the totals add up to more than 365.
• One analyst coded everything, with no second-coder check.

Limits.

• Running long isn't proof of profit. Meta hides spend and results for US ads, so a long-running ad may simply be cheap or forgotten.
• Relaunches reset the clock. Re-posting a creative under a new ad ID makes it look younger than it is.
• Survival rates are probably inflated. Ads switched off within days seem under-returned by the Library.
• Big advertisers are under-counted in the cohorts. The video filter leaves out most auto-assembled dynamic ads.
• Some table cells are small. Treat differences under 10 points as noise.