Fewer People Post on Social Media as AI Use Rises

Logging on and posting are no longer the same thing

A new survey out of the UK found that fewer social media users are actually posting, sharing, or commenting — even as the share who use AI tools has gone up. It’s a result worth sitting with, because it says something about how people are spending their time online these days.

According to a report published on April 2, 2026 by Ofcom, the UK’s communications regulator, the share of adult social media users who mainly post, share, or comment fell from 61% in 2024 to 49% in 2025. Ofcom cautions, though, that because the survey methodology and question order changed between years, this year-over-year gap should be read as a directional signal rather than a precise measure of change.

That doesn’t mean people are leaving social media. 89% of UK adult internet users still use at least one social media platform. Among social media users, 25% said they mostly hit “like,” and 24% said they read without reacting at all. It’s worth separating the question of whether people are logging on from whether they’re speaking up publicly.

The English draft looks accurate and complete. No corrections needed.

The Burden of Leaving Something Public

This survey ran from September to November 2025, covering 7,533 people aged 16 and older in the UK. The share of respondents who said they’d tried a lot of new sites also fell, from 24% in 2024 to 14% in 2025. It’s possible this reflects a habit of handling more tasks within familiar apps, but the number alone doesn’t let us say online activity overall has shrunk.

As for why people are posting less on social media, two explanations come to mind: a shift toward video-centric consumption, and the growing weight of leaving something public. Both are plausible, but this survey doesn’t isolate the effect of either one.

On scrollable video feeds like Instagram Reels or TikTok, there’s plenty to do without ever writing a post yourself. In Ofcom’s in-depth interviews, one respondent recalled posting even what they ate for dinner in the past — but rarely posting anything now. Cases like this suggest people can keep using a platform constantly while posting far less often.

There’s also the burden of knowing a public post sticks around. The share of UK adult internet users who worry that something they said online could come back to haunt them rose from 43% in 2024 to 49% in 2025. Once you start thinking about who might read it, or in what context it might resurface later, even writing about your everyday life starts to feel like something to be careful about.

Instagram Stories, which vanish from the default view after 24 hours, and messages exchanged between close friends carry a different kind of burden than public posts. Still, this survey alone doesn’t tell us that users have all shifted toward those formats. And Stories, too, can be saved or screenshotted by someone else.

Readers and writers aren’t the same crowd

There have always been users who read content without ever leaving a trace of a reaction. We call them lurkers¹. A 2025 study involving Professor Anis Bakir of Northeastern University and others analyzed roughly 17 million tweets about military support for Ukraine. Actions like likes or reposts were rare compared to views, but this study didn’t count exactly 90% of all X users as lurkers.

Gallup Korea’s data also shows a gap between reading and writing. A 2018 survey, reintroduced in a 2025 report, found that among people in their 20s to 40s, about 75% read other people’s opinions, while no more than 20% wrote down their own thoughts. That’s a figure describing the participation gap of that particular moment—it can’t be used as evidence that writing has recently declined in Korea.

The “90-9-1 rule”² is another well-known way of describing how participation in online communities concentrates among a small slice of users. It’s a rule of thumb—the majority reads, a minority writes—not a fixed ratio that applies uniformly across every platform. Nor should we compare the ratio from 2006 with some different survey figure from today and conclude that participation has dropped further.

Borrowing sociologist Erving Goffman’s concept of “impression management”³, people also think about how they’ll appear to others when they post. Choosing not to voice an opinion publicly can stem not just from a lack of interest but from concern over reputation. Which means that if we take the opinions visible in comments as the views of the entire readership, we end up missing someone.

More AI usage doesn’t mean everyone’s looking for a conversation partner

Ofcom’s report found that 31% of UK adults said they used AI tools in 2024, rising to 54% in 2025. The purposes ranged widely—work or study assistance, information lookup, and more. Among AI users, only 12% said they used it for conversation, rising to 19% among 25-34 year-old users. That 54% figure doesn’t mean everyone in it treats AI like a friend.

Alongside the survey, Ofcom also ran a separate study involving long-form interviews with a small number of participants. Here, cases emerged of people talking to AI to ease loneliness while working from home, or to sort through their feelings after a breakup. A participant who used AI for breakup counseling explained that it wasn’t so much trusting AI more than human advice—it was closer to confirming their own thinking.

AI usage has spread widely in the US too. In Pew Research Center’s 2025 survey, 64% of teens aged 13-17 said they had experience using chatbots. The roughly 30% who said they used one daily is also a share of all teens surveyed, not just chatbot users.

In a separate 2025 survey by Common Sense Media, 72% of US 13-17 year-olds said they had used an AI companion service at least once. Having used such a service and intending to replace friends with one are two different things. Since the surveying organizations and questions differ, this figure can’t be read alongside the Pew numbers as an indicator of rising or falling usage.

In Gallup Korea’s 2025 survey, 34% of respondents nationwide aged 13 and older (excluding Jeju) said they had used ChatGPT within the past year. In 2023, that figure was 7%. Again, this number doesn’t specifically indicate emotional conversation.

A decline in social media engagement and a rise in AI usage were observed together, but we can’t tell whether people who wrote less started talking to AI instead. Confirming that relationship would require tracking changes in the same individuals’ usage purposes and time spent, separately.

Oswarld’s Lens

I want to call what’s happening here the “broadcast-ification of social media” — a small number of creators making content that the majority watches. Anyone can post, but in practice, the posts and videos that actually fill the screen tend to concentrate around a handful of creators. Catching up with friends and watching popular content end up mixed together inside a single app.

From a GTM strategy standpoint, I think you have to look at user numbers and engagement patterns together. Even if ordinary users’ posts decline, if professional creators’ videos increase, the opportunity to show ads can be preserved. But a service people visit to check on friends and a service people visit to watch recommended videos meet their customers in fundamentally different ways. You also have to be careful about concluding that interest has disappeared just by looking at comment counts.

In Ofcom’s survey, the share of UK adult internet users who felt the internet’s benefits outweighed its risks fell from 72% to 59%, and the share of SNS users who felt SNS was good for their mental health fell from 42% to 36%. Even accounting for changes in survey methodology, there’s reason to look more closely at user satisfaction and burden. That said, these are responses about self-perceived impact, not measurements of actual changes in mental health.

It’s too early to conclude that AI resolves that dissatisfaction or substitutes for relationships between people. Getting momentary comfort from talking with AI, sorting out a worry through conversation, and avoiding meeting people are all different things. A number showing that usage time has increased alone can’t tell us which of these is actually happening.

I use AI every day myself, so I understand the convenience of sorting out my thoughts through conversation. What concerns me is the possibility that, simply because AI is convenient, we end up reducing the experience of negotiating opinions with other people too. Listening to someone, clearing up a misunderstanding, accepting that you’re different from someone else — these things take time. For me, the more important standard is whether we can keep having those experiences even while using AI.

Closing

What this data shows is that logging onto social media and participating publicly aren’t the same thing. Even users who only read have interests and opinions of their own — which is exactly why the words of those who leave reactions can’t stand in for the whole user base.

The rise in AI usage also needs to be broken down by purpose. Whether someone’s using it for work, chatting for fun, or bringing up things they can’t say to another person changes what the number actually means. Between the figures showing declining social media engagement and those showing rising AI use, there’s still a lot that needs to be verified.

Next time you scroll your Instagram feed, take a look at how much of it is posts from people you know versus recommended content. It might also be worth thinking about which posts you read but didn’t react to.

References & Further Reading

• Ofcom, Adults’ Media Use and Attitudes Report 2026, April 2, 2026. Covers 2025 survey findings and an explanation of changes to survey methodology.
• Northeastern Global News, Research on social media users who don’t leave reactions, April 14, 2025. Explains the findings of Bakir’s research team, which analyzed tweets related to military support for Ukraine.
• Pew Research Center, Teens, Social Media and AI Chatbots 2025, December 9, 2025.
• Common Sense Media, Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions, July 16, 2025.
• Gallup Korea, Usage rates for 18 media, content, and social network services, October 2, 2025. The ChatGPT usage experience data is from a 2025 survey, while the difference between writing and reading is re-cited from a 2018 survey.

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

Footnotes

1. Lurker: an online community user who mostly reads or watches content while rarely leaving public reactions. The same person may write in some topics or services and only read in others. ↩

2. The 90-9-1 rule: a rule of thumb describing participation imbalance in online communities — most people read, some occasionally participate, and a tiny minority mainly create content. It doesn’t reflect the actual ratio in every community. ↩

3. Impression Management: a concept describing how people adjust their words and behavior based on how they think others will perceive them. ↩