The creator format is the most reliably effective thing in social advertising right now, and the most consistently misfiled. Part four of the ad creative series.
The lift is real
TikTok's own analysis, covering February 2024 to January 2025, compared creator-led ads against non-creator ads at the same cost per thousand impressions.
At the same media cost. That last qualifier is what makes the number interesting — this is not a story about creators being cheaper, it is a story about the same impression being worth more when a person is in it.
Posting from the creator's own account adds more: 59% higher engagement and 16% better six-second view-through rate than running the same content from the brand account. TikTok's separate Creative Center work on Spark Ads puts the conversion lift at 43%, with 30% better completion and 142% more engagement, from aggregated brand-lift studies.
I would treat the Spark Ads figures as slightly softer than the first set. They come from a JavaScript-heavy interactive page rather than a stated methodology document, and they are the platform's own reporting on its own format. Directionally consistent with everything else, but not the number to build a business case on.
The trade-off almost nobody makes explicitly
Then there is the CreativeX analysis, which looked at 1.6 million ads representing roughly $2 billion of spend across about 800 brands, and found something considerably less flattering.
Read those three bars together. Fewer than half of creator ads show the brand in the window where most of the audience still exists. Only one viewer in seven is there after three seconds. And 45% of these ads manage both failures at once — no early branding, and nobody watching to see the late branding either.
That is an enormous amount of money buying attention for a product nobody can name afterwards.
And here is where it argues with practice
The CreativeX conclusion is straightforward: brand early, get 16% better completion. But the thing I have actually watched work is close to the opposite.
The creator ads I have seen perform best often carry no brand reference at all — no logo, just a product or service placed inside a real use case. They read native, and native is the whole reason the format works. When testimonials get over-templated, over-optimised, logo-stamped and loaded with a call to action, they stop looking like content and start looking like an ad, and they typically stop performing like content too.
I am not going to pretend these reconcile neatly. They are measuring different things: CreativeX is measuring brand attribution and completion, and I am describing response. Both can be right, and the trade-off is real — every unit of visible branding you add buys attribution and costs native-ness. What is unhelpful is pretending the trade-off does not exist, which is what "always brand in the first three seconds" does.
My working position: decide which one you are buying before you brief the creator. If the objective is attribution and recall, brand early and accept a lower response rate. If the objective is response from a cold audience, let it look like content and solve attribution somewhere else in the funnel. Do not try to have both in one frame — that is how you get the 45%.
The format detail
Vertical 9:16 travels everywhere now. It is mobile-first, so it fits short-form video placements, LinkedIn and Instagram without a re-edit, which makes it the sensible default for creator briefs rather than a platform-specific choice. The exception worth knowing is the newest surface: an ad inside ChatGPT is a single image and two short lines of text, so nothing shot for 9:16 travels there without being rebuilt as a static.
What SMBs get wrong about it
The most expensive mistake is filing this in the wrong budget.
Most small businesses treat user-generated content as a production expense, and then compare its cost against a designer's or videographer's day rate. On that comparison it usually looks expensive.
It is the wrong comparison. If a designer or videographer produces a video for you, there is a direct cost and you own an asset — but that content on its own buys no audience exposure. You still have to pay, separately, for every person who sees it.
A creator collaboration is not that. It bundles the content with the creator's audience. You are buying distribution that happens to arrive with a video attached. Which means the line item belongs in the media budget, alongside the ad spend, not in the production budget alongside the design invoice. Businesses that make that reclassification start making sensible decisions about it almost immediately, because they finally have it next to the thing it actually competes with.
What to read next
- B2B ad creative — where the same emotional logic is least applied and most valuable.
- AI ad creative — including synthetic presenters in creator-style formats.
FAQ
How much better do creator ads actually perform?
TikTok's own analysis, covering February 2024 to January 2025, compared creator-led against non-creator ads at the same cost per thousand impressions and found 70% higher click-through, 159% more engagement and 43% better conversion on Spark Ads. Posting from the creator's own account rather than the brand account added 59% higher engagement and a 16% better six-second view-through rate. Treat the Spark Ads figures as slightly softer — they come from an interactive page rather than a stated methodology document, and they are the platform reporting on its own format.
Should the brand appear in the first three seconds?
That is a trade-off to decide before you brief the creator, not a rule. CreativeX's analysis of 1.6 million creator ads found that branding early buys 16% better completion, and that fewer than half of creator ads show the brand while most of the audience still exists. Against that, the creator ads I have seen perform best often carry no brand reference at all, because reading native is the whole reason the format works. Every unit of visible branding buys attribution and costs native-ness. If the objective is attribution and recall, brand early and accept a lower response rate; if it is response from a cold audience, let it look like content and solve attribution elsewhere in the funnel. What does not work is trying to have both in one frame.
What is the 45% figure on this page?
It is the share of creator ads in the CreativeX analysis that manage both failures at once — no branding in the first three seconds, and nobody still watching to see the late branding either. Only 14% of viewers are there after three seconds at all. That is an enormous amount of money buying attention for a product nobody can name afterwards.
Which aspect ratio should a creator brief specify?
Vertical 9:16. It is mobile-first and travels everywhere now, fitting short-form video placements, LinkedIn and Instagram without a re-edit, which makes it the sensible default rather than a platform-specific choice.
Is user-generated content a production cost or a media cost?
A media cost, and filing it in the wrong budget is the most expensive mistake small businesses make with it. If a designer or videographer produces a video, you own an asset that buys no audience exposure — you still pay separately for everyone who sees it. A creator collaboration bundles the content with the creator's audience, so you are buying distribution that happens to arrive with a video attached. The line item belongs beside the ad spend, next to the thing it actually competes with.
Sources on this page
- TikTok, "The Creator Advantage", 2025. TikTok internal data, February 2024 to January 2025, comparing creator-led against non-creator ads at equal cost per thousand impressions.
- TikTok Creative Center, "Return on Influence", 2024. Aggregated brand-lift studies. Platform self-reported, methodology not stated in detail.
- CreativeX, "Are Creators Really Effective?", 2025, and Adweek's coverage of it. Observational analysis of 1.6 million paid creator ads, roughly $2 billion of spend, about 800 brands.
- The "79% of people say user-generated content highly impacts purchase decisions" figure is frequently attached to this topic. It is from 2019 fieldwork and the publisher's site no longer exists — see the source notes on the hub.
