25 UGC Ad Hooks That Stop the Scroll (With Examples)

---

title: "25 UGC Ad Hooks That Stop the Scroll (With Examples)" description: "25 UGC ad hook formulas by angle with example lines, plus how to measure hook rate and hold rate, how many hooks to test, and what Hook Lab automates." date: "2026-09-24" tags: ["hooks", "creative-testing", "ugc", "hook-lab"]

The first two to three seconds of a UGC ad decide whether the other 27 seconds get watched. On Meta, roughly a quarter to a third of impressions make it past the three-second mark for a healthy ad; on TikTok the bar is measured at two seconds and the numbers run higher. Everything below the hook, the offer, the proof, the call to action, is invisible to the people who scrolled. Here are 25 hook formulas we see working in September 2026, grouped by angle, with an example line for each, followed by the measurement you need to actually pick a winner.

Key takeaways

• A hook is a promise about the next 20 seconds; the formula only matters insofar as it makes a promise the viewer cares about.
• Hook rate is 3-second views divided by impressions (2-second on TikTok). Median on Meta in 2026 is around 28 percent; 35 percent and up is strong (as of September 2026, from published benchmark aggregations).
• Hold rate tells you whether the hook lied. Test hook rate first, then hold rate, then conversion.
• Six to ten hooks per concept is the practical minimum to find a real winner; fewer than that and noise dominates.
• Hook Lab exists to make the boring part, generating and swapping 20 openers onto the same body, take minutes rather than a week.

25 hook formulas by angle

Curiosity

1. The withheld reveal. "I wasn't going to post this, but my sister made me."
2. The unexpected ingredient. "The reason my sleep tracker finally worked is a $12 pillow."
3. The strange rule. "There's one thing I never do before a flight anymore, and it's not what you think."

Pain point

4. Name the exact annoyance. "If your dog's water bowl turns into a swamp by Tuesday, watch this."
5. The 2 a.m. version. "It's 2 a.m., the baby is finally asleep, and the white noise app just crashed again."
6. The thing you gave up on. "I stopped trying to fix my posture years ago. Then this showed up."

Bold claim

7. The specific number. "This replaced four products on my bathroom shelf. Four."
8. The dare. "Show me another travel mug that survives a car door. I'll wait."
9. The reversal. "Everything you've heard about protein powder timing is wrong."

Social proof

10. The borrowed authority. "My physio told me to stop buying insoles and get this instead."
11. The crowd. "There's a reason 40,000 people bought this in August." (Only if true and documented.)
12. The sceptic converted. "I thought this was a scam. I was extremely wrong."

Pattern interrupt

13. The mid-action open. Start with the product already being used, no greeting: "...and that's the third time it's saved me this week."
14. The wrong setting. Talking about skincare from the driver's seat, or a laptop stand from a campsite.
15. The whisper. Deliberately quiet delivery on a loud feed: "Don't tell my roommate I found this."

POV

16. POV: the moment. "POV: you finally found jeans that don't gap at the back."
17. POV: the other person. "POV: your partner discovers you've been hiding the good coffee."

Question

18. The direct question. "Why does nobody talk about how loud air fryers are?"
19. The self-check. "Do you also re-wash the same load of laundry three times?"

Stat

20. The uncomfortable stat. "Most people replace their toothbrush head about half as often as dentists recommend. Here's what changed for me."
21. The personal stat. "I tracked every coffee I bought for 30 days. Then I did the maths."

Before and after

22. The split screen. Left: the drawer before. Right: the drawer now. "Same drawer, one weekend."
23. The timeline. "Day one, day fourteen, day thirty. Watch the difference."

Storytime

24. The confession. "Storytime: I almost returned this, and I'm so glad I didn't."

Us versus them

25. The tribe. "This is for the people who read the whole label. You know who you are."

A note on claims: hooks 7, 11, 20 and 21 are the ones that get ads rejected or brands into trouble. Anything numeric needs a source you can produce, and any "I used this" needs to be true of a real person. If the hook is delivered by an AI creator, "I've been using this for a month" is a fictional testimonial, and the FTC's consumer review and testimonial rule (in effect since October 2024) treats that as a problem. Rewrite it as a demonstration or a stated fact instead.

How to test hooks

Hook rate

Hook rate is the share of impressions that turned into 3-second views (Meta) or 2-second views (TikTok):

hook rate = 3-second video plays / impressions

Benchmarks vary by vertical and by whether the traffic is cold, but published aggregations for 2026 put the Meta median around 28 percent, with 25 to 35 percent healthy and above 35 to 40 percent in the top decile; TikTok runs five to ten points higher because of the shorter window (as of September 2026). Treat these as rough bands, not targets. Your own account's median is the number that matters, and the goal is to beat it consistently, not to hit an industry figure.

Hold rate

Hook rate only tells you the opener worked. Hold rate tells you whether the ad kept its promise:

hold rate = ThruPlay (or 15-second views) / 3-second views

A hook that gets 45 percent of viewers past three seconds and loses 90 percent of them by second ten is a bait-and-switch, and the algorithm will price it accordingly. The pattern you want is a strong hook rate and a hold rate that does not fall off a cliff. When you see a high hook rate and a bad hold rate, the fix is in the body, not the opener.

The order of operations

1. Freeze the body of the ad. Same script from second three onward, same creator, same edit.
2. Swap only the opening line and the opening shot.
3. Run each variant to roughly 1,000 impressions or a fixed daily budget for three days, whichever comes first, in an ad set that lets them compete on equal terms.
4. Rank by hook rate. Cut the bottom half.
5. Rank the survivors by hold rate. Cut the ones that collapse.
6. Only now look at cost per result, which needs far more data to be reliable than hook rate does.

The reason to test in this order is statistical. Hook rate stabilises with a few hundred impressions; conversions need thousands. If you judge hooks by purchases you will pick winners that are mostly noise.

How many hooks to test

Fewer than four and you are not really testing. Six to ten per concept is where most teams land: enough to cover three or four angles with two variations each, small enough to read the results in a few days. If you have the budget for more, spread it across concepts rather than piling 30 hooks onto one body. The best hook for a losing concept is still a loser.

Retest the winners quarterly. Hook fatigue is real, and the formula that worked in March is often the one everyone in your category is running by September.

What Hook Lab automates

Most of the work above is not creative; it is production and bookkeeping. Hook Lab in VidsFor.me takes one finished ad and generates a set of alternative openers, each as a separate variant with the body unchanged:

• Generates hook lines across the angles above from the product brief, so you are choosing between 20 written options rather than staring at an empty script.
• Renders each opener with the same Cast creator and voice, so the only variable is the hook. At provider prices that is a few dollars of talking-head time per opener via OmniHuman on fal.ai, about $0.20 per second (September 2026 list price), rather than a reshoot.
• Keeps captions, the AI disclosure label and Content Credentials consistent across the set so nothing needs re-checking per variant.
• Exports the set with hook text in the file name and metadata, and can push them to Meta or TikTok as paused drafts (beta) so the ad set is ready to switch on.

It does not tell you which hook will win. It makes the test cheap enough that you can find out.

Next step

If you want to run the six-to-ten-hook test on your own ad, the variants feature covers Hook Lab and what happens when you add creators, languages and formats to the matrix. The pricing page lists what is included on each plan.