AI and Social Media: What the Research Shows

AI and social media are now the same product. AI decides which posts, videos and ads you see, a growing share of the content itself is AI-generated, and chatbots sit inside the apps. The research so far points one way: AI makes feeds better at keeping you scrolling, and not always better at giving you what you actually want.

This isn't a robots-are-coming piece. Most of what follows comes from peer-reviewed studies and one unusually candid earnings call. I'll stick to what was measured, say where the evidence is thin, and end with what you can do about it on your own phone.

What Is the Role of AI in Social Media?

AI does four jobs on a modern social app. It ranks your feed, choosing from thousands of possible posts the few you see first. It moderates, flagging spam and banned content before a human ever looks. It generates, from AI filters and captions to whole images and videos. And it chats, through assistants like Meta AI and Snapchat's My AI.

The first job matters most for your attention. Feeds used to be mostly posts from people you chose to follow. Now a large share is recommended by a model that has never met you but has watched every pause of your thumb. If you want the mechanics, our piece on social media algorithms goes deeper.

Does AI Make You Spend More Time on Social Media?

Yes, and the companies say so out loud. On Meta's Q3 2024 earnings call, Mark Zuckerberg said improvements to AI-driven feed and video recommendations had led to an 8% increase in time spent on Facebook and a 6% increase on Instagram that year alone. That was a selling point for investors, not a warning label.

Independent research backs up the direction. In a large experiment published in Science in 2023, Andrew Guess and colleagues moved consenting Facebook and Instagram users to a plain reverse-chronological feed for three months during the 2020 US election. Time on the platforms dropped substantially, and so did activity. People did not switch off, though. Many simply spent more time on TikTok and YouTube, which tells you how strong the pull of ranked video is.

Does the Algorithm Show You What You Want?

It shows you what you engage with, which is not the same thing. Smitha Milli and colleagues ran a preregistered audit of Twitter in 2023 with 806 users, comparing the posts the engagement-based algorithm would show against the latest posts from accounts people followed. The results were published in PNAS Nexus in 2025.

The algorithm boosted posts expressing anger, sadness and anxiety, and political posts that were more partisan and hostile to the other side. Readers felt angrier after reading them. The kicker: when asked, users said they did not prefer the political posts the algorithm picked. The system was good at predicting what they would react to and bad at predicting what they would choose.

That gap is the whole story of AI and social media in one finding. Engagement is easy to measure, so that is what gets optimized. If you have ever closed an app feeling worse than when you opened it, this is probably part of why. See the brain science of doomscrolling for what that loop does to you.

AI-Generated Content on Social Media

The second shift is what fills the feed. Renée DiResta of Stanford and Josh Goldstein of Georgetown studied 125 Facebook Pages that each posted at least 50 AI-generated images, for a 2024 Harvard Kennedy School Misinformation Review paper. Most were spam or scam operations. Their images got hundreds of millions of exposures, and one AI image post hit 40 million views.

Two details stood out. Facebook was recommending these images to people who did not follow the Pages, and comments suggested many viewers had no idea the pictures were fake. Some Pages pushed people toward content farms; others tried to sell products that did not exist.

Text is no easier to spot. In a University of Zurich study published in Science Advances, 697 people could not reliably tell tweets written by GPT-3 from tweets written by real users. AI-written false tweets were also a little harder to catch than human-written ones. That study used a 2020-era model. Today's are better.

AI Chatbots Inside Social Apps

The newest layer is conversation. Meta AI lives inside Instagram, Facebook and WhatsApp, and Snapchat pins My AI at the top of chats. Separate companion apps market AI friends outright.

Teens took to this fast. A 2025 Common Sense Media survey of 1,060 US teens found 72% had used an AI companion and over half used one at least a few times a month. About one in three found those conversations as satisfying as talking with real friends, or more. About one in three users had brought serious matters to an AI instead of a person.

There is a reassuring number in there too: 80% of teens said they still put real friendships first. Still, a chatbot that is always available, always agreeable and built into the app you already open 50 times a day is a new kind of pull. We cover the habit side in AI chatbot addiction.

The Good Side of AI on Social Media

It would be lazy to call all of this bad. AI moderation catches enormous volumes of spam and abuse that no human team could review. Auto-captions and image descriptions make apps more accessible. And recommendations can surface a niche community or a creator you would never have found by yourself.

The same Zurich study found that accurate AI-written tweets were easier to understand than human ones. The tool is not the problem. The goal it is pointed at is. For more on what social media does well, see positive effects of social media, and for the other column, negative effects of social media.

How to Take Back Control From AI-Driven Feeds

You cannot out-think a model trained on billions of sessions. You can make it work with less material. These are the moves I would start with:

Tools like Go Gray put grayscale one tap away in Control Center, give you a 30-minute color break when you need a photo or a map, and nudge you after 30 minutes and an hour on the apps you choose. Go Gray Pro adds app blocks past a daily limit and shows your pickups app by app.

The Bottom Line

AI and social media together mean a feed that learns faster than you do, more content nobody can verify, and a chatbot that never gets tired of you. The evidence is clearest on time spent and emotional content, and still early on long-term mental health.

You do not need to quit to push back. Pick your own feed, cut the triggers, and make the screen dull enough that the algorithm has less to work with. Get Go Gray on the App Store and see how much less interesting a perfectly ranked feed is in gray.