Short answer
Open on a person or a moment, or on one specific result or price.
In 43 US TikTok Shop affiliate videos we coded, 12 of 23 that reached 5M views opened on a person, a moment or a confession and 6 on a result or price, while none of the 14 under 1M views did either. Those 14 mostly praised the product, greeted the viewer or called them out ("stop scrolling if...").
What do TikTok Shop videos with millions of views open with?
Mostly with a person or a moment: 12 of the 23 winners (5M+ views) in our set opened that way, 6 opened on a specific result or price, and 4 opened with no words at all. The 14 videos under 1M views look like the mirror image: none opened either way, and 10 of them talked about the product or spoke to the viewer.
What 43 TikTok Shop videos opened with, one square per video
"Opened on" means the first on-screen text and the first spoken line, as an AI model (Gemini) wrote them down while watching each video. Each opening went into one family by what its words are about.
Which kind of person-or-moment hook should you film?
Any of the six below; each had at least 2 winners and none of them showed up under 1M views. The most common were someone else's line in a skit (4 winners) and a result (4). 2 winners simply opened by saying out loud that the video was an ad. Each card applies its template to a portable blender.
Six openings that won, written for one portable blender
Babe, did you just drink my whole smoothie?Someone else's line4 of 23 winners
On screen: "[his/her] [old thing] is so [bad]". Said by them: "[Babe], what are you doing?" over the hard way
- This is an ad for a blender I swore I'd never use
Confession3 of 23 winners
"This is an ad for a [product] I was told I didn't need", or "OK, I'm actually kind of annoyed" and the reason
- You're telling me I carried a shaker bottle for two years?
Everyday moment3 of 23 winners
"Oh no, we're out of [product] again", or "you're telling me I've been [old way] for [years]?"
- POV: blending at 6am without waking your roommate
POV or "me when"2 of 23 winners
"POV: you're trying to [do the thing] without anyone noticing", or "me the second [moment]"
- This and never paying $9 for a smoothie again
Result4 of 23 winners
"[Result] in [time], when [old way] couldn't", or "this and never [chore] again"
- This blender cost less than a week of smoothies
Price2 of 23 winners
"A whole [set] for only [price]", or a reaction with the price on screen
3 of the 6 result-or-price winners had a digit or a price in the words. The others proved it in the picture: a before and after, a pile of empty tubes. Use the free hook generator for more lines, then keep only the ones that fit a row above.
Do "stop scrolling if" and other viewer callouts work?
Not in this set, though the products explain part of it. 6 of the 14 videos under 1M views opened by speaking to the viewer, and 0 of 23 winners did. Tabcut's "visual problem statement" (its example: "If you're someone who struggles with bad breath") and the Hook Trifecta's line that calls out the buyer both recommend that kind of opening.
The fair reading is narrower. 5 of the 6 losing callouts were for products that had no winner in the set at all (gummies and chargers), so the product could explain the flop as well as the hook. The cleaner test is a product with both a winner and a loser, and there are 4. For a spin scrubber, the winners opened on a moment and a result, and the losers opened with a callout ("if you like a dirty house, this video isn't for you") and with unrelated text. For a skincare kit, both winners opened on a price; the loser listed what the product does.
Matched to openings in the set, Tabcut's relatable story is the best-supported advice on the results page, and its callout-style types the weakest.
The ranking guides' hook types, matched to the same videos
Should the text on screen say the same thing as your voice?
No: in every winner that used both, the two differed, though that's only 6 videos. That fits the Hook Trifecta's rule that the text and spoken hook must never be the same sentence.
Openings with text on screen and a spoken line
One video
The Trifecta's other claim, three hooks at once, didn't hold up as a requirement. Among the 23 winners, 7 opened on a spoken line alone, 6 on text alone, and 4 on the picture alone. One good line was enough for most of them.
Spoken length didn't separate the groups. The winners' first spoken lines had a median of 8 words, about 3.2 seconds at 150 words per minute (our script timer's pace), the losers' 9 words, and the longest winning line ran 18. On-screen text ran longer: a median of 10 words for winners.
What does one product look like with each kind of hook?
We wrote four openings for a portable blender, one per family, and rendered each on the same synthetic clip with UGCLive's export renderer: the hook as on-screen text in the app's hook style (yellow Anton in a dark box), the first spoken line as a caption. In A and B the voice says something the text doesn't, as in every winner that used both; C and D repeat the text, as 3 losers did. In a real A, the first shot would be the moment itself.
Our pick: film A or B first, and test a C or D opening only against one of them, because 18 of 23 winners opened like A or B and none of the 14 videos under 1M did, while C and D made up 10 of those 14. If a C or D opening beats your A and B on the same product, this pick is wrong for your niche. Our TikTok affiliate marketing guide covers how many orders a hook test needs before you call a winner.
How do you write a TikTok Shop hook from this?
Pick a row from the template table, write the line a person would say, and put any proof on screen.
- Pick the kind of opening
Choose a row above that is true for you: a confession, a result you got, a price you paid.
- Write the spoken line
Write what a person in that moment would say out loud, and leave the product's name out of it. If it runs past one breath, cut it.
- Write different on-screen text
Put the result, the price or the setup on screen while the voice tells the moment. Never put the spoken sentence on screen too.
- Test praise and callouts second
To try "stop scrolling if" or "this product is amazing", post it as a second version of the same video, after your A or B.
Where could this count be wrong?
The set is 43 US videos, picked from best-selling products' top affiliate videos plus some weak ones. We measured views because sales per video aren't public, and some reach was probably paid. The families are one coder's reading (ours) of an AI model's notes, with no second coder, so a few videos could move. 3 videos were in Spanish and none reached 5M; without them, 5 of 12 English losers still opened with a callout. The products differ too: the losers lean toward gummies and chargers. And we didn't post the four blender hooks, so the render shows what each family looks like, not which would sell.
