# We Analyzed 705 Top TikTok Ads: The First 3 Seconds — Cut Marketing

> Original 2026 data on 705 TikTok Top Ads from 28 countries: length, what appears in the first second, on-screen text, formats and captions, plus what to copy.

# We Analyzed 705 Top TikTok Ads: The First 3 Seconds

Original 2026 data on 705 TikTok Top Ads from 28 countries: length, what appears in the first second, on-screen text, formats and captions, plus what to copy.

Published 2026-11-12 Last updated 2026-11-16 12 min read [Cut Marketing Team](/editorial)

![AI-generated overhead view of a research table with twelve smartphones in a grid showing blurred vertical video stills, three marked with lime stickers, beside a printed contact sheet with circled frames, a stopwatch and coffee](/samples/blog/tiktok-top-ads-study-2026.webp)

Top TikTok ads in 2026 are vertical, longer than most advice suggests, and show the thing being sold almost immediately. We collected 705 ads that TikTok lists as Top Ads in 28 countries and labeled the opening of 662 of them. The median ad ran 26 seconds. In physical-product ads, 76% put the product on screen within the first second. A person’s face was the most common first frame, but hands and products together opened more product ads than faces did.

**Key findings**

-   **97% are vertical 9:16.** Only 21 of 705 ads used another ratio (square, landscape or a shorter portrait such as 4:5).
-   **Short is not the rule.** The median length was 26 seconds; only 20% ran 15 seconds or less. Product ads: median 23 seconds.
-   **The product shows up fast.** 76% of product ads showed the product by the 1-second mark, and 89% showed the product, a face, or both. Only 11% showed neither.
-   **Faces lead, hands follow closely.** Product ads opened on a face 39% of the time and on hands or a body without a face 27% of the time; 19% opened on the product alone.
-   **About half put text on the first frame (54%), but almost none open with a question (4%).**
-   **Hands-on demos lead.** 37% of product ads were a person showing or using the product, ahead of product-only footage (21%) and talking to camera (11%).
-   **Half of product ads cut to a new shot within 2.5 seconds (50%).**
-   **Captions sell plainly.** The median caption was 80 characters; 28% mention a discount, price or free item.

We make Cut Marketing (Cutma), an AI ad creative tool, and our Ad Finder reads the same public TikTok pages. This is a description of what the listed ads have in common, not proof of what causes performance. Limits are listed in the method section.

## How did we collect and label the ads?

We read 2,506 entries from TikTok’s public Top Ads lists on 12 November 2026, kept the 705 unique ads, and labeled three opening frames of each with AI vision under one written rubric, with hand spot checks.

**Source.** [TikTok Creative Center Top Ads](https://ads.tiktok.com/business/creativecenter/inspiration/topads/pc/en), the public library where TikTok lists ads by country, period, industry and sort order. We read the first page (up to 20 ads) of each list for the last 30 days, on 12 November 2026.

**Lists.** 28 countries × 4 sort orders (For You, likes, CTR, 6-second view rate), plus 12 product industries × 8 large markets. That gave 2,506 list entries and 705 unique ads, because the same ad often appears in several lists.

**Labels.** For each ad we extracted three frames: at 0.05 seconds, 1.0 second and 2.5 seconds. AI vision models labeled each ad under one written rubric: what fills the first frame, whether a face and the product are visible by 1 second, the type of overlay text, the overall format, whether there is a cut by 2.5 seconds, and the hook type of the caption. We spot-checked samples by hand. 43 ads had no readable video, so the frame-based numbers use 662 ads; length, ratio, objective and caption numbers use all 705.

**Product ads.** 472 ads (452 labeled) belong to TikTok’s physical-goods industries: apparel, beauty, food and beverage, health, home improvement, household products, tech, appliances, baby, pets, sports and e-commerce. The rest are apps, games, services, finance, travel, education and entertainment.

**Download the data:** [per-ad labels (CSV, 705 rows)](/samples/research/tiktok-top-ads-2026.csv). It contains TikTok ad IDs, lists, industry, objective, length and our labels, with no captions, videos or advertiser assets.

**What the sample is.** Apparel (22%), beauty (13%) and news & entertainment (13%) were the largest industries. Campaign objectives: conversions 36%, product sales 19%, lead generation 13%, reach 12%, traffic 12%, video views 9%.

## Finding 1: Top ads are longer than “keep it under 15 seconds”

The median top ad ran 26 seconds, and only 20% ran 15 seconds or less. Product ads were a little shorter, with a median of 23 seconds.

| Length | All ads (705) | Product ads (472) |
| --- | --- | --- |
| 15 s or less | 20% | 24% |
| 15–30 s | 37% | 38% |
| 30–60 s | 30% | 31% |
| Over 60 s | 13% | 8% |
| Median | 26.0 s | 23.0 s |

The middle half of all ads ran between 15 and 44 seconds. Apps, services and entertainment ran longer (median 30 seconds; 24% over a minute), which fits demos and explainers.

Median length in countries with at least 30 ads in the sample:

| Country | Median length | Ads |
| --- | --- | --- |
| Mexico | 19 s | 32 |
| United States | 20 s | 86 |
| United Kingdom | 22 s | 59 |
| Germany | 23 s | 51 |
| Korea | 24 s | 38 |
| Indonesia | 30 s | 102 |
| Brazil | 31 s | 44 |
| Japan | 43 s | 50 |

**What to take from it:** length did not separate these ads from each other; the opening did. Make the first seconds work and let the rest run as long as the demo needs. If you want a shorter cut, test it as a variant rather than assuming it wins. Current length and ratio limits per placement are in our [TikTok, Reels and Shorts spec sheet](/blog/tiktok-reels-shorts-ad-specs-2026).

## Finding 2: In product ads, the product appears in the first second

By the 1-second mark, 76% of product ads showed the product and 89% showed the product, a face or both. For other ads the figures were 42% and 73%.

| By the 1-second mark | Product ads (452) | Other ads (210) |
| --- | --- | --- |
| Product visible | **76%** | 42% |
| A face visible | 42% | 47% |
| Both | 29% | 16% |
| Product or face | 89% | 73% |
| Neither | 11% | 27% |

For apps and services, “product” means the app screen, the game or the service result. Those ads more often open on a face, a scene or a title card and show what is being sold later.

**What fills the very first frame?**

| First frame | Product ads | All ads |
| --- | --- | --- |
| A person’s face | 39% | 40% |
| Hands or body, no face | 27% | 20% |
| The product alone | 19% | 15% |
| A scene or place | 10% | 14% |
| A text card | 6% | 9% |
| An app or screen | 0% | 2% |

Two out of three product ads (65%) opened on a person, but not always a face: a hand already holding, opening, applying or pouring the product was the second most common first frame. A full-screen title card or logo opened only 6% of product ads.

## Finding 3: Half use first-frame text, and very few ask a question

54% of the 662 ads had overlay text on the very first frame. What that text said:

| First-frame text | Share of all ads |
| --- | --- |
| No text | 46% |
| Brand or product name | 10% |
| Offer, price or discount | 7% |
| Subtitle of what someone says | 7% |
| A claim or benefit | 6% |
| A question | 4% |
| A problem or pain point | 2% |
| Other (labels, stickers, mixed) | 18% |

The “start with a question” hook is common advice, but only 4% of these ads opened with one. Brand names, offers and subtitles were each more common.

Country mattered more here than anywhere else: first-frame text appeared on 36% of US ads but 74% of Japanese ads, 60% of German ads and 58% of Korean ads. If you run the same video in several countries, a text-free first frame for the US and a text-led one for Japan is a reasonable pair to test.

## Finding 4: Product ads are hands-on demos, not talking heads

The most common product-ad format was a person showing or using the product (37%). Talking to camera made up only 11% of product ads but led among apps, services and entertainment at 26%.

| Format | Product ads (452) | Other ads (210) |
| --- | --- | --- |
| Person shows or uses the product | **37%** | 5% |
| Product footage, no presenter | 21% | 13% |
| Person talks to camera | 11% | **26%** |
| Slideshow or static images | 6% | 7% |
| Motion graphics or animation | 4% | 16% |
| Skit or acted story | 4% | 10% |
| Before and after | 2% | 0% |
| Dance or trend format | 1% | 1% |
| Screen recording | 0% | 6% |
| Other or mixed | 12% | 14% |

Talking to camera is the format for apps, finance, education and services, where there is nothing to hold up. For physical products the winning move is usually to put the product in someone’s hands.

Country patterns, with small samples: hands-on demos made up 46% of Indonesian ads, talking-to-camera 37% of Brazilian ads, and product-only footage 30% of German ads and 25% of US ads.

**Likes, with a caveat.** TikTok shows likes on each listed ad. Among product ads, the median ad with a face in the first second had 24 likes; the median ad without one had 10. By format, talking to camera had a median of 28, hands-on demos 24 and product-only footage 6. Likes on ads are small numbers, they depend on spend and reach, and these ads are all already on the list, so treat this as a direction to test, not a rule.

## Finding 5: Half of product ads cut within 2.5 seconds

A new shot appeared between 0.05 s and 2.5 s in 50% of product ads and 30% of other ads.

There is a nuance in TikTok’s own sort orders. Among product ads that appeared in the 6-second view rate list, 43% cut within 2.5 seconds (159 ads), against 54% of product ads that did not (293 ads). They also showed the product by 1 second slightly less often (72% against 79%). Ads that keep people watching past six seconds are not simply the fastest-cut ones; a single continuous shot that is interesting enough can hold just as well.

## Finding 6: Captions are short, concrete and often about the offer

The median caption was 80 characters, about one sentence, and 28% mentioned a discount, price, free item or sale. Across all 705 captions:

-   42% contain a number, 42% contain a hashtag, 25% an emoji, 14% a question mark.
-   28% mention a discount, price, free item or sale. In Japan it was 61%; in Brazil 12%.

The first sentence of product-ad captions was most often a benefit claim (29%) or an offer (21%); questions (10%), calls to action (7%) and stories (4%) were less common. Many captions simply named the product and its use.

## What should you copy from top TikTok ads?

For a physical product, put the product, or a face with the product, on screen in the first second, skip the logo card, and let the length follow the demo. The full checklist from the patterns above:

1.  **Get the product, or a face with the product, into the first second.** 89% of product ads did one or the other.
2.  **Open on hands if you don’t have a presenter.** A hand already holding, opening or applying the product was the second most common opener.
3.  **Skip the logo card.** Only 6% of product ads opened on a text card. If you need the brand, put it on the product or in a corner.
4.  **Decide on first-frame text per country.** Text-free openings were common in the US; text-led openings were the norm in Japan.
5.  **Use the caption for the concrete part:** the benefit, the offer or the price, in about one sentence.
6.  **Don’t force a 9-second edit.** Top product ads had a median of 23 seconds. Test length as a variant.
7.  **Test one variable at a time.** Change only the opening and keep the rest of the ad the same. Our [creative testing workflow](/blog/creative-testing-workflow) shows how.

For twelve concrete opening patterns with first frames and caption lines, see [Video Ad Hooks: 12 Openers for the First 3 Seconds](/blog/video-ad-hooks-first-3-seconds).

## Three of the data’s openings, made with AI

With image-to-video, the clip starts from the still you give it, so you can design the first frame on purpose. Below are three openings from our own [showcase](/showcase) that match the most common first frames in the data. Our team made them in Google Flow, outside the Cut Marketing workspace, with fictional unbranded products; the prompts are the ones we used. They are examples of the opening, not test results.

### 1\. A face and the product in the first frame (29% of product ads)

Selfie taste test. The creator's face and the dessert are both in frame 1. Omni 1.2 in Google Flow, 10 seconds.

```
Vertical 9:16 selfie video, a Korean woman in her 20s in a convenience-store aisle holds an unbranded new cream-filled dessert, peels the lid, takes a spoonful and reacts in Korean: "이 가격에 이 맛이면 오늘 저녁 이걸로 끝!" Bright store lighting, handheld, authentic creator energy, no readable brand labels.
```

### 2\. Hands on the product before any face (27% of product ads open on hands or a body)

A hand reaches for the gloss in the first second; the face only appears when the product is applied. Veo 3.1 in Google Flow, 8 seconds.

```
8-second vertical UGC-style film with native audio: a hand lifts the wine-red gloss tube from the marble vanity and pulls out the doe-foot applicator, then a woman applies the glossy wine-red color to her lips, handheld close-ups, natural light, no text, no logo.
```

### 3\. The product alone, already moving (19% of product ads)

Product-only opening: the case is in frame from the start and the lid begins to open. Veo 3.1 in Google Flow, 8 seconds.

```
8-second vertical product film with native audio: the case lid slowly opens to reveal the earbuds as a band of cool light sweeps across them, slow push-in, drifting haze, cinematic, no text, no logo.
```

**Making these in Cut Marketing.** The workspace catalog has [Veo 3.1 Fast](/models/veo-3.1-fast) (8 seconds at 720p: 32 credits, $0.32) and [Gemini Omni Flash 1.2](/models/gemini-omni-flash-1.2) (4 seconds at 720p: 48 credits, $0.48). These are not the same variants as the Veo 3.1 and Omni 1.2 we used in Flow, so results will differ. Start from a real photo of your product, describe one opening action, and compare every frame with the product before you use it. Spoken lines and on-screen text are not guaranteed word-for-word, so add them afterwards in Finish the Ad in the studio: it places your exact hook and CTA text, burns in captions in 40 styles, reads your script as AI narration and exports 9:16. Our [AI product video checklist](/blog/ai-product-video-production-checklist) covers the review step, and [AI-generated ad disclosure rules](/blog/ai-generated-ad-disclosure-rules-2026) covers labeling.

## What can’t this data tell you?

It can’t tell you what causes performance. Every ad here is already on TikTok’s list, the sample is one 30-day window, and the labels cover three frames without sound.

-   **Correlation, not cause.** Every ad here is already on TikTok’s list. We can say what they have in common, not that a face or a cut made them perform.
-   **TikTok chooses the list.** The Creative Center does not publish how ads are selected or ranked, and signed-out visitors see only the first page of each list.
-   **One 30-day window.** Collected on 12 November 2026. Formats and trends move; we plan to repeat the study.
-   **Small country samples.** Country figures rest on 17 to 102 ads each. Treat differences of a few points as noise.
-   **Three frames, no sound.** We labeled frames at 0.05, 1.0 and 2.5 seconds. We did not analyze audio, voiceover or the rest of the video.
-   **AI-assisted labels.** A fixed rubric and spot checks reduce errors but do not remove them. The CSV lets you check our labels against the ads yourself.

## Look at the ads yourself

The numbers describe hundreds of ads; your category is what matters. Cut Marketing’s [Ad Finder](/ad-finder) lists the same TikTok top ads by country and industry, plus Meta Ad Library results by keyword, and turns an ad’s structure into a brief without copying its words, people or brand. If you have a reference video you own or have rights to, [Video Cloning](https://app.cutmarketing.app/studio/video-cloning?ui_locale=en) rebuilds its shot structure with your product. Pricing starts at $9 a month for 900 credits, and new accounts get 30 trial credits without a card. See [pricing](/pricing), or read [how to find competitor ads on Meta and TikTok](/blog/find-competitor-ads-meta-tiktok).

## Keep reading

[Video Ad Hooks: 12 Openers for the First 3 Seconds](/blog/video-ad-hooks-first-3-seconds): Twelve video ad hook patterns for TikTok, Reels and Shorts, each with a first frame, a caption line and an AI prompt, plus how to draft three hooks for 210 credits.

[How to Find Competitor Ads on Meta and TikTok (2026 Guide)](/blog/find-competitor-ads-meta-tiktok): Find competitor ads on Facebook, Instagram and TikTok with free public tools, read days running and versions honestly, and log what is worth testing yourself.

[Remove Objects from Photos and Video with AI: 2026 Test](/blog/ai-object-remover-test-2026): How to remove an object, sticker or text from a photo or video with AI: describe it or paint over it. Real tests of five image models, masks vs words, and costs.

## Make the next ad in minutes.

Start with a product image or a short brief. The studio shows the credit cost before you generate.

[Open studio](https://app.cutmarketing.app/?ui_locale=en)
