🤖 Case study · 292,784 TikToks tagged by visual origin

AI videos get 2× the views of real footage. Here is the catch.

Name: AI-generated vs real footage: views, likes, shares and engagement tier by visual origin
Creator: Revid
License: https://creativecommons.org/licenses/by/4.0/

We tagged 292,784 TikToks posted between May and December 2025 by what is on screen: real footage, AI generated, animation or a mix. AI-generated videos average 1.4M views against 661.4k for real footage, and get the fewest likes per view of any group. Then we looked at 388.4k posts on X, Instagram and Threads and the picture flipped.

2.1×

AI generated vs real footage, average views per TikTok

1.4M vs 661.4k

20%

of AI-generated TikToks pass 1M views

vs 10% of real footage

3.9%

like rate on AI videos, the lowest of the four groups

vs 7.7% on real footage

1.3%

AI share of viral posts on X, Instagram and Threads

vs 6.7% of low-engagement posts

AI generated vs real footage on TikTok, May to December 2025

1.4M

AI generated, 9,940 videos

2.1×

average views per video

661.4k

Real footage, 252,107 videos

Part 1: TikTok

Synthetic beats real on every view metric

Every TikTok in the index posted between May and December 2025 was tagged by a vision model as real footage, AI generated, animation or mixed. 292,784 videos carry the tag. Violet is AI, mint is real footage.

AI generated vs real footage: the whole distribution

Share of tagged TikToks with more than X views, for each visual origin. The violet line sits above the mint line at every threshold: 20% of AI-generated videos pass 1M views against 10% of real footage.

AI generated (19% pass 1M)Real footage (10% pass 1M)Animation / cartoon (14% pass 1M)Mixed (11% pass 1M)

Average views per video

2.1× AI vs real

🤖AI generated
1.4M
🎨Animation / cartoon
797.6k
🧩Mixed
713.8k
🎥Real footage
661.4k

Estimated median views per video

2.3× AI vs real

🤖AI generated
151.7k
🎨Animation / cartoon
121.3k
🧩Mixed
87.3k
🎥Real footage
67.3k

Like rate (likes per view)

3.9% on AI, lowest of the four

Animation / cartoon
8.4%
Real footage
7.7%
Mixed
6.3%
AI generated
3.9%

Share rate (shares per view)

0.94% on AI vs 0.87% on real

Animation / cartoon
1.04%
Mixed
1.01%
AI generated
0.94%
Real footage
0.87%

AI share of tagged TikToks by month, 2025

AI-generated videos as a share of all tagged videos posted each month. Peak of 4.3% in Jun 2025, 0.5% by Dec 2025. Small text under each bar is AI videos out of tagged videos that month.

AI-generated share of tagged TikToks, by month posted

4.3% peak in Jun 2025

The share of tagged videos in this index that were AI generated peaked in Jun 2025 and fell to 0.5% by Dec 2025. Late 2025 months have far fewer videos in the index, so the tail is noisier.

May 2025
944 AI of 25,558 tagged
3.7%
Jun 2025
2,956 AI of 68,125 tagged
4.3%
Jul 2025
2,177 AI of 56,599 tagged
3.8%
Aug 2025
1,501 AI of 41,668 tagged
3.6%
Sep 2025
1,449 AI of 43,086 tagged
3.4%
Oct 2025
758 AI of 33,963 tagged
2.2%
Nov 2025
121 AI of 17,151 tagged
0.7%
Dec 2025
34 AI of 6,633 tagged
0.5%

2×

AI-generated TikToks average 2.1× the views of real footage, and 20% of them pass a million views against 10% for real footage.

Revid index, 9,940 AI-generated vs 252,107 real-footage TikToks, May to December 2025 · revid.ai/case-studies/ai-videos-vs-real

Save as image Post

Why

People watch AI videos and forward them. They do not like them.

Three mechanics explain the pattern: the views come from novelty, the missing likes come from distrust, and the difference between distribution and affection.

1. Novelty and the curiosity gap

An AI video that works is surreal, impossible or uncanny, and the viewer stays to figure out how it was made. That is a curiosity gap TikTok rewards with watch time and completion. Real footage has to earn attention with a story; AI footage often gets it from the first frame. In the index, the estimated median AI video does 151.7k views against 67.3k for real footage, so the gap is not only a handful of monsters.

2. Low like rate means low trust

A like is a small endorsement. Viewers give AI videos the fewest of them: 3.9% likes per view against 7.7% for real footage and 8.4% for animation / cartoon, the best of the four. People watch synthetic content but they are less willing to put their name on it. The comment rate is nearly flat across the four groups, so the reaction is not more debate, it is less endorsement.

3. Distribution is not affection

Shares travel; likes stay. AI videos get shared 1.08× as often per view as real footage (0.94% vs 0.87%), which is what "look at this" content does, even though animation / cartoon shares best of all and AI ranks #3 of four. A video can be forwarded across the platform without anyone loving it. That is exactly the profile you would expect from content people find impressive but not endearing.

Like rate by visual origin

Likes per view, TikTok. AI generated comes last at 3.9%, 2× below real footage.

Share rate by visual origin

Shares per view, TikTok. AI generated ranks #3 of four at 0.94%, against 0.87% for real footage.

Part 2: the catch

On X, Instagram and Threads, AI visuals live in the bottom tier

Same vision model, different index: 155,470 posts with a visual tag out of 388.4k in the Revid cross-platform index. Each post carries an engagement tier from low to viral. Here is the share of posts in each tier that were AI generated.

Out of 100 viral posts on X, Instagram and Threads, how many use AI visuals?

10,546 viral-tier posts with a visual tag. 142 were AI generated, 1.3%.

1AI generated

4Animation

2Mixed

93Real shot

And out of 100 low-engagement posts?

24,257 low-tier posts with a visual tag. 1,623 were AI generated, 6.7%, 5× the viral-tier share.

7AI generated

11Animation

3Mixed

80Real shot

AI-generated share of posts, by engagement tier

5× more common in the low tier than the viral tier

Low engagement
1,623 AI of 24,257 tagged posts
6.7%
Medium engagement
2,075 AI of 72,735 tagged posts
2.9%
High engagement
748 AI of 47,932 tagged posts
1.6%
Viral
142 AI of 10,546 tagged posts
1.3%

AI-generated share of posts, by platform

4.6% on X, 0.3% on Instagram

X
4,030 AI of 87,863 tagged posts
4.6%
Threads
425 AI of 24,948 tagged posts
1.7%
Instagram
133 AI of 42,659 tagged posts
0.3%

Across the three platforms, 4,588 of 155,470 tagged posts (3%) were AI generated. Threads posts have no public view count, so its tier comes from likes and replies.

6.7%

AI share, low engagement posts

1,623 of 24,257 tagged

2.9%

AI share, medium engagement posts

2,075 of 72,735 tagged

1.6%

AI share, high engagement posts

748 of 47,932 tagged

1.3%

AI share, viral posts

142 of 10,546 tagged

5×

On X, Instagram and Threads, AI visuals are 5× more common in the lowest engagement tier than in the viral tier: 6.7% of low-engagement posts vs 1.3% of viral posts.

Revid cross-platform index, 155,470 tagged posts across 388.4k rows · revid.ai/case-studies/ai-videos-vs-real

Save as image Post

Reconciling the two

Both results are true. They describe different jobs.

On TikTok the 9,940 AI-generated videos that made it into a viral-skewed index outperform real footage on average because the format itself, surreal, novel, "how did they make this", gets watched to the end and forwarded. On the text-first platforms, AI visuals are mostly filler under a post, and filler sits in the low tier: 1,623 AI posts in the low tier against 142 in the viral tier. The sample sizes differ by an order of magnitude and so does the role the visual plays.

What this means if you make AI videos

We make AI videos for a living, so read this as advice, not a boast. AI wins when it is the point of the video: something a camera cannot shoot, a scene that makes people stop and forward it. It loses when it is filler, a generic stock-style clip under a caption that would have worked with any footage. On TikTok the upside is real (2.1× the average views, 2.1× the viral rate), but the like rate says viewers still hold synthetic content at arm's length, so lean into the surprise and keep the script honest. If you want to see what that looks like in practice, the tools at revid.ai/tools are built around exactly those formats.

Methodology

TikTok corpus: 2,377,477 public TikToks in the Revid index, collected continuously since January 2026 by searching hashtags and keywords that Revid users track, plus the channels they watch. Average views across the whole index: 1.7M; estimated median: 110.7k; 22% pass 1M views.
Visual origin tag: a vision model looked at sampled frames and labeled each video Real shot, AI generated, Mixed or Animation/Cartoon. The tag only exists for videos posted between May and December 2025, 292,784 videos (9,940 AI generated, 252,107 real footage, 16,534 animation, 14,203 mixed). Videos without a tag are excluded from every number on this page. The monthly share chart divides AI-tagged videos by all tagged videos posted that month.
"Average views" is the arithmetic mean of public play counts. "Median" is estimated from the share of videos above 12 fixed view thresholds (log-linear interpolation), accurate to within a few percent. "Pass 1M" is the share of videos with more than 1,000,000 plays at indexing time. Like, share and comment rates are total likes, shares or comments divided by total plays for the group.
Cross-platform index: 388,379 posts from X, Instagram, Threads, YouTube, Reddit and LinkedIn. The visual tag covers 155,470 X, Instagram and Threads posts. Each post carries an engagement tier (low, medium, high, viral) assigned relative to its platform. The chart shows the share of tagged posts in each tier that were AI generated, so it is a composition number, not a performance number: it says where AI visuals sit, not what they earn.
Bias, stated plainly: neither index is a random sample. Both are keyword-sampled, skew toward content that already performed and toward the interests of Revid users, and the classifier is imperfect (photoreal AI can pass as real, edited real footage can read as mixed). That inflates every absolute number. The comparisons inside each index are the point, and on TikTok they hold at the mean, the median and the viral rate.

Data pulled September 15, 2026. The index grows every day; numbers are refreshed on a schedule and the date above always tells you which snapshot you are reading.

Cite this study

Revid (2026). AI Videos Get 2x the Views of Real Footage. Here's the Catch. (TikTok Data 2026). https://www.revid.ai/case-studies/ai-videos-vs-real

Made with Revid

Make the AI videos people forward.

Revid turns a prompt, a link or a script into a finished short video with voice, visuals, captions and music. Search the same index this study came from, see which AI formats are getting watched right now, and remake them in one click.

Make a video free Search the viral index

Free to start · no credit card

More from the Revid index

All case studies

💘Case study

7×

more views for couples content than crypto

Love beats money

We measured the average views of 46 hashtag themes across 2.4 million TikToks. Couples content averages 7x the views of crypto. Here is the full ranking of what TikTok rewards, from romance to real estate.

Read the study

🫣Case study

2.2×

more views for handheld footage than selfie videos

Your face costs views

Selfie-style videos average half the views of handheld footage, faceless videos beat person-focused ones, and animals beat everyone. The data on whether you need to show your face on TikTok.

Read the study

🩺Case study

3,000

viral health TikToks audited from their transcripts

TikTok health claims

An audit of the 3,000 most-viewed English health TikToks in our index: how many make a health claim, how many back it with anything, how many sell a product, and which claims went most viral.

Read the study

FAQs

Frequently Asked Questions

Do AI-generated videos go viral on TikTok?

In the Revid index, yes, more often than real footage. 19.5% of the 9,940 TikToks tagged AI generated passed 1 million views, against 9.5% of the 252,107 tagged real footage. Average views were 1.4M vs 661.4k (2.1×) and the estimated median was 151.7k vs 67.3k. The index skews viral, so treat these as comparisons inside the index, not as odds for a random upload.

Why do AI videos get fewer likes?

AI-generated TikToks have the lowest like rate of the four visual origins: 3.9% likes per view, against 7.7% for real footage, 2× lower. Our reading: viewers watch AI content out of curiosity and forward it because it is strange or impressive, but a like is a small act of endorsement and people are less willing to endorse something that looks synthetic. The share rate points the same way: 0.94% for AI vs 0.87% for real footage.

Does TikTok suppress AI content?

Our data cannot show suppression or promotion, because we only see public counts on videos that made it into a keyword-sampled index. What it shows is that AI-generated videos in that index outperform real footage on average views (2.1×), on median views (2.3×) and on viral rate (20% vs 10%). If TikTok were suppressing AI content across the board, we would expect the opposite pattern among the videos we can see.

How do you know a video is AI-generated?

A vision model looked at sampled frames of each video and tagged the visual origin as Real shot, AI generated, Mixed or Animation/Cartoon. The tag exists for TikToks posted between May and December 2025, 292,784 videos in total, and for 155,470 X, Instagram and Threads posts in the cross-platform index. It is a classifier, not ground truth: photoreal AI can be tagged real, heavily edited real footage can be tagged mixed, and AI-generated voice over real footage is invisible to a frame-based model. The gaps we report are large enough to survive that noise, but the exact percentages are not precise to the decimal.

Still have questions?

Find detailed answers to 100+ questions about features, tools, and workflows

or check our markdown version optimized for LLMs