Most social networks are making efforts to reverse the rapid rise of AI-generated content on their platforms, and while some appear to be making modest gains, others continue to struggle.
Here’s what we saw for September:
LinkedIn and X (formerly Twitter) each saw an increase in likely AI content in September, while Threads, Facebook and Reddit each saw declines.
LinkedIn remains the social platform with the highest proportion of likely AI content in our sample by a sizable margin, which increased by another 2% last month. This was also the first month in which longer LinkedIn posts were more likely to be AI-generated than shorter posts since our first report in July.
Reddit, meanwhile, has managed to remain consistently below average. Its recent 2% decline in likely AI content erases the 2% increase noted in our August report, returning the platform to its July levels. Reddit continues to feature almost half as much likely-AI content as the next lowest platform in our analysis.
X leapfrogged Facebook and is now tied with Threads in second place for the highest proportion of likely AI content following a sizable 5% increase, and after Facebook and Threads each saw similarly sized declines.
The recent spike in likely AI content on X all but erases the 6% drop highlighted in our previous study, and brings the platform back within range of its all-time peak of 68% from December of 2025.
Facebook is the only platform that has seen a meaningful decrease in likely AI content recently, dropping 11% over the last two months. This suggests the platform’s efforts to reduce AI-generated content are proving more impactful than its competitors.
The data below represents how much each platform’s long-form written content is more likely than not to be at least 15% AI-generated.

Originality.ai analyzed 109 randomly selected long-form English-language LinkedIn posts of at least 100 words published in September of 2026. The analysis determined that 85, or 78%, were Likely AI.

LinkedIn remains the social media platform with the most Likely AI content in September, following a 2.35% increase. This latest spike further widens the gap between LinkedIn and the other social media platforms included in our analysis, which is now at 16%, up from just 10% last month.
While LinkedIn has made attempts to rein in AI usage on the platform, including through its “Seems like AI Slop” reporting feature and a September 15 update intended to “Tackle AI Slop,” those efforts appear to be making little or no impact.

Longer LinkedIn posts are now more likely than shorter ones to be AI-generated, reversing a trend noted in previous reports.
That’s because the proportion of likely AI-generated posts of 100-149 words dropped by 13%, from 86% to 73%, while likely AI posts of 150-249 words jumped from 73% to 76%. Furthermore, 84% of sampled LinkedIn posts 250 words or longer were Likely AI in September.

Posts related to careers narrowly edged out leadership and data and analytics as the category with the highest proportions of likely AI content, after a significant leap from 33% in August to 87% in September.
Leadership and data and analytics trailed closely at 86% each, followed by AI at 75%. At the other end of the spectrum, just 64% of emerging technology content and 60% of climate-related content on the platform is Likely AI. All content categories, however, continue to see more likely AI content than not.

Originality.ai analyzed 199 randomly selected long-form English-language X posts of 100 words or more published in September of 2026 and determined that 123, or 61.8%, were Likely AI-generated. That represents a roughly 5% increase from August.

The increase erases much of the 6% reduction in likely AI content the platform achieved between July and August, while pushing it from fourth position in our study to now tied with Threads for second.
At the same time, AI-generated content on X remains below its December 2025 peak, when 68% of posts were likely AI-generated. The back-and-forth nature of AI content on X is perhaps the clearest demonstration of a platform actively trying to rein in AI-generated content but only achieving limited or temporary results.

As seen in previous months, longer X posts typically feature more AI content. While just 53% of posts of 100-149 words are likely AI-generated, so too are 68% of posts 150-199 words, and 71% of posts 200 words or longer.

Finance and crypto have been dethroned as the category with the highest proportion of likely AI content following a notable spike in likely AI education and self-improvement content. That category saw a massive spike in likely AI content, from 25% in August to 86% in September, moving it from the bottom of the ranking to the top.
Most other categories, meanwhile, either stayed the same or saw reductions in likely AI content. Finance and crypto-related posts, for example, saw a 7% month-over-month reduction in AI-generated content; health and wellness remained consistent at 71%; business, work and marketing dropped from 65% to 62%, while the proportion of likely AI-generated politics, news and public affairs content declined from 59% to just 44%.
In August, three content categories on X remained below the 50% threshold, but in September only one — politics, news and public affairs — remained below 50%.
Personal life and general commentary, meanwhile, saw a slight uptick of 3%; AI-generated sports, entertainment, and gaming content increased more than 14%; while faith and spirituality increased from 40% to 57% (though the data for the latter may be skewed by a limited sample size).

Originality.ai analyzed 128 randomly selected English-language Threads posts of 50-99 words published in September and determined that 79, or 61.7%, were Likely AI.

The latest data represents a 4% month-over-month decline from August’s 66% reading. Originality.ai’s analysis of likely AI content on Threads has remained relatively consistent since 2025, which averaged 62.5%. Thus far, the 2026 average is only 0.4% lower, at 62.1%.

Originality analyzed 193 randomly selected long-form English-language Facebook posts of 100 words or more published in September of 2026 and determined that 104, or 53.9%, were likely AI generated.
That represents a 5% month-over-month decline in likely AI content on the platform since August, and an 11% decline since July — the largest and most consistent decline in likely AI content across all platforms included in this report.

The proportion of likely AI content on Facebook was still higher in September than in January, when it began the year at 49%. After six months of back-to-back increases, which peaked at 65% in July, Facebook appears to have successfully reversed the trend.

As noted in previous reports, shorter Facebook posts continue to feature the highest proportion of likely AI content. Roughly 59% of those in the 100 to 199 word range were likely AI-generated, compared with about 41% of posts between 200 and 299 words, and just 17% of those 300 words or longer.
Posts of all lengths have seen significant month-over-month declines in likely AI content, but the proportion of likely AI-generated posts of 300 words or longer is less than half the level recorded in August. A relatively small sample size, however, may be skewing the results.

Within Facebook groups, those dedicated to technology and business appear to be dominated by likely AI content, though a small sample size may be skewing those results as well.
That category aside, nonprofits and civic groups remain the biggest hotbeds of likely AI content, as seen in our two previous reports. However, Likely AI content in that category also declined significantly month-over-month, from 85% in August to 68% in September.

In fact, Originality.ai tracked month-over-month declines in Likely AI content across many Facebook group categories. The public figures and education creators category, for example, saw a meaningful drop from 69% in August to 35% in September. AI-generated posts within health-related Facebook groups also declined from 63% in August to 57% in September.
No Facebook group category fell below the 50% threshold in August, but three came in under that bar in September — culture, museums and libraries, government science and environment, and public figures and education.

Originality.ai analyzed 548 randomly selected long-form English-language Reddit posts of 100 words or more published in August of 2026 and determined that 159, or 29%, were Likely AI.
That represents a nearly 2% month-over-month decline from July and reverts the level of likely AI content on Reddit to the level we reported in June. (Reddit data is on a one-month delay due to data collection challenges).
Reddit remains the social media platform with the lowest proportion of likely AI content in our study by a sizable margin.
Previous Social Media Tracker Reports:
