One in four social media posts longer than 250 words was flagged as fully AI-written in Pangram’s scan of 1,002,627 posts collected between 24 April and 9 July 2026. The precise figure is 25.72%. This post gathers the detection studies, platform disclosure counts and consumer surveys behind AI-generated content on social media, each with its sample, measurement and period stated.
AI-Generated Content on Social Media Statistics
How Much AI-Generated Content Is on Social Media?
Pangram scanned 1,002,627 posts across LinkedIn, Medium, Substack, X/Twitter and Reddit after launching its Chrome extension on 24 April 2026. Only items above 50 words were scanned, and each post counts once.
The average AI rate across every scanned item was 13.8%. Restrict the sample to items over 250 words and 25.72% flagged as fully AI-generated.
Length drove the split on four of the five platforms. Substack was the exception, where the fully AI-generated rate stayed roughly flat regardless of post length.
| Pangram metric | Figure |
|---|---|
| Posts scanned across five platforms | 1,002,627 |
| Average AI rate, all scanned items | 13.8% |
| Items over 250 words flagged fully AI-generated | 25.72% |
| LinkedIn long-form posts flagged fully AI-generated | Over 40% |
| LinkedIn share of all items scanned | One third |
| LinkedIn share of all AI-flagged content | 62% |
| Substack posts flagged AI-generated or AI-assisted | 21.9% |
Source: Pangram, AI in Your Feed, posts scanned 24 April to 9 July 2026
X/Twitter articles carried the most machine writing once AI-assisted work was counted. Pangram flagged 23.9% as fully AI-generated and a further 22.9% as AI-assisted or mixed, leaving 53.2% classified as fully human-authored. X runs its own in-platform assistant, tracked separately in these Grok AI statistics.
X/Twitter Long-Form Articles by Authorship
AI-Generated Content on Social Media by Platform
LinkedIn supplied one third of the posts Pangram scanned but 62% of everything the detector flagged as AI. Pangram notes that people use AI more readily on a platform tied to their real professional identity than on casual or anonymous ones.
A second detector reached a higher number using a different method. Originality.AI analysed 5,000 public LinkedIn posts of at least 100 words across nine topics in July 2026 and classified 4,061 of them, or 81.2%, as Likely AI.
The two figures measure different slices. Originality.AI draws its sample from topic-and-date searches with a 100-word floor; Pangram measures what opted-in users encountered in their feeds at a 250-word threshold. Originality.AI’s earlier reading, covering 3,368 long-form posts from 99 influential profiles between January and November 2025, put the Likely AI share at 53.7%.
LinkedIn is responding at the product level. The platform is rolling out a native “Seems like AI slop” option inside its three-dot menu, and Chief Product Officer Hari Srinivasan has said AI slop is a top priority.
How Much of Reddit Is AI-Generated?
Reddit produced the largest scan volume and one of the lowest AI shares. Pangram attributes the gap to composition rather than to Reddit users writing differently.
| Reddit metric | Figure |
|---|---|
| Share of all items Pangram scanned | 36.7% |
| Combined AI share, all Reddit items | 4.4% |
| Top-level posts flagged as AI-written | 11.6% |
| Replies flagged as human-authored | 98.1% |
| Replies as a share of Reddit items scanned | 72% |
Source: Pangram, AI in Your Feed, posts scanned 24 April to 9 July 2026
Controlling for length, a top-level Reddit post carried a 5.25 times greater chance of being AI-generated than a reply. Pangram links the clean reply pool to Reddit’s spam policy, which removes accounts that mass-produce automated replies but catches less of the lower-volume, higher-impact top-level content.
TikTok AI-Generated Content Labeling Statistics
TikTok said on 10 July 2026 that it had labeled over 3 billion videos as AI-generated content. Its previous disclosure, published 19 November 2025, put the figure at over 1.3 billion. Both are floors rather than exact counts.
Three mechanisms feed the count: C2PA Content Credentials, creator labeling tools, and TikTok’s own invisible watermarking. TikTok joined the C2PA Steering Committee alongside the July announcement.
TikTok Videos Labeled as AI-Generated Content
| TikTok metric | Figure | Disclosure date |
|---|---|---|
| Videos labeled as AI-generated content | Over 1.3 billion | 19 November 2025 |
| Videos labeled as AI-generated content | Over 3 billion | 10 July 2026 |
| Fake accounts removed, first three months of 2026 | More than 86 million | 10 July 2026 |
| Committed to AI literacy expert partnerships | More than $4 million | 10 July 2026 |
| Views generated by expert partner content | More than 200 million | 10 July 2026 |
Source: TikTok Newsroom, 19 November 2025 and 10 July 2026
The 86 million account removals cover the first quarter of 2026 and sit alongside the labeling work rather than inside it. TikTok said it would begin testing improved detection for accounts dedicated to AI spam in politics and current events, financial advice and medical content. Labels only cover what the pipeline reads, which is why deepfake incident tracking still records synthetic material reaching feeds unmarked.
How Often Do People See AI-Generated Content on Social Media?
Sprout Social’s Q1 2026 Pulse Survey polled 2,250 social media users across the US, UK and Australia between 5 and 9 February 2026, fielded online by Glimpse. Just over half report frequent exposure.
Reported Frequency of Seeing AI Slop
56% of respondents see AI slop often or very often, and 83% see it at least sometimes. Baby Boomers report the least exposure, with 37% seeing it regularly and 19% saying they never do.
Boomers also react most sharply: 56% said they would be very unlikely to interact with AI slop. Gen Z was more mixed, with 34% at least somewhat likely to engage and 40% unlikely.
| Generation | Yes | No | Not sure |
|---|---|---|---|
| Gen Z | 50% | 40% | 10% |
| Millennials | 44% | 43% | 13% |
| Gen X | 38% | 48% | 14% |
| Baby Boomers | 29% | 54% | 17% |
Source: Sprout Social Q1 2026 Pulse Survey, “Have you unfollowed, muted or blocked a brand or creator because their content felt like AI slop”, fielded 5 to 9 February 2026
The generation most averse to interacting with AI content encounters the least of it. Unlabeled AI posts also topped the list of behaviours consumers want brands to drop, ahead of engagement bait.
What Consumers Want Brands to Stop Doing in 2026
Trust moved less than the exposure numbers suggest. Overall trust in social media rose for 16% of respondents over 12 months, stayed the same for 49% and fell for 35%. Among those reporting a decline, misinformation was cited by 30% and unregulated AI slop by 20%.
Change in Social Media Trust Over 12 Months
News is where the erosion is sharpest. 88% strongly or somewhat agreed that video generation AI tools have left them with less trust in the news they see on social media, with US respondents most likely to strongly agree at 46%.
Can People Spot AI-Generated Content on Social Media?
Two experiments published on arXiv reach different numbers because they test different material. In the December 2025 study, 165 users completed 233 sessions classifying 20 images each drawn from 120 deliberately difficult cases, and averaged 54% accuracy with a mean response time of 7.3 seconds.
The earlier “Real or Not Quiz” analysis covered roughly 287,000 image evaluations from over 12,500 participants and reported a 62% success rate. Participants there were most accurate on human portraits and least accurate on natural and urban landscapes.
Source: arXiv 2512.22236, December 2025; arXiv 2507.18640, “Real or Not Quiz” data collected August 2024
Both sit above a coin flip and well below reliable. Machine detection carries its own error rate: Pangram states a 0.01% false positive rate for the Pangram 3.3 model used in its social media scan.
Who Is Making AI-Generated Content on Social Media?
Production and reception move in opposite directions. Hootsuite’s Social Media Trends 2026 research put daily AI use among social media managers at 79%, while more than 30% of consumers say AI-generated ads make them less likely to choose a brand.
| Metric | Figure |
|---|---|
| Social media managers using AI daily | 79% |
| Marketers calling human involvement very important or critical for AI content | 91% |
| Consumers less likely to choose a brand with AI-generated ads | More than 30% |
Source: Hootsuite, Social Media Trends 2026
Most of this output starts in general assistants and editing suites rather than dedicated AI feeds, which tracks with wider generative AI adoption rates and with the install bases behind ChatGPT usage figures.
Pangram sets the social numbers in a wider frame, citing research that estimated 35% of newly published websites on the open internet were AI-generated or AI-assisted. Assistant-side demand also shows up in Meta AI install and usage data and these AI mode usage statistics.
FAQs
How much of social media content is AI-generated?
25.72% of scanned posts over 250 words flagged as fully AI-generated, against a 13.8% average AI rate across all scanned items. Pangram scanned 1,002,627 posts between 24 April and 9 July 2026.
Which social media platform has the most AI-generated content?
LinkedIn. Pangram flagged over 40% of its long-form posts as fully AI-generated in July 2026. Originality.AI classified 81.2% of a separate 5,000-post search sample as Likely AI that same month.
How much of Reddit is AI-generated?
4.4% of all Reddit items Pangram scanned carried an AI flag. Broken down, 11.6% of top-level Reddit posts flagged as AI-written, while 98.1% of replies flagged as human-authored.
Why does Reddit have less AI content than LinkedIn?
Composition. Replies made up 72% of the Reddit items scanned and were 98.1% human-authored. Pangram credits Reddit’s spam policy for removing accounts that mass-produce automated replies.
How many AI videos has TikTok labeled?
Over 3 billion videos as of 10 July 2026, using C2PA Content Credentials, creator labeling tools and invisible watermarking. TikTok’s previous disclosure, on 19 November 2025, put the count at over 1.3 billion.
Sources
https://www.pangram.com/blog/ai-in-your-feed
https://newsroom.tiktok.com/helping-people-spot-and-understand-aigc-on-tiktok
https://sproutsocial.com/insights/press/social-media-is-now-the-top-source-for-breaking-news-new-sprout-social-research-finds/
https://originality.ai/blog/ai-content-published-linkedin
