If you’re using an AI tool to write your Instagram captions or Facebook posts, you’ve probably wondered: is this actually working, or am I quietly losing engagement?
You’ve likely seen articles claiming AI content gets way more engagement than content written by a person. You’ve also probably seen the opposite: stories about “AI slop” and audiences turning against anything that looks AI-made. Both of those things are true at the same time, just in different situations.
This post walks through what the research actually found, in plain terms, so you can figure out what it means for your own posts.
A quick example of AI working well, and why it’s not a perfect comparison
A couple of well-known brands have used AI tools and seen strong results. Heinz, the ketchup company, asked the public to submit AI-generated images of “what ketchup looks like,” then turned the best ones into real ads and posts. The campaign got way more attention than their usual posts, with engagement 38% higher than past campaigns.
That sounds like a strong argument for AI. But here’s the catch: Heinz didn’t just type a prompt and hit publish. A professional creative team picked the idea, refined it and built an entire campaign around it. The AI was one tool among many, not the whole plan.
That’s the part that often gets left out when people share these success stories. AI worked well there because skilled people were steering it, not because AI on its own is a guaranteed boost to engagement.
What happens when researchers test AI content directly against human content
Case studies like Heinz tell you what happened once, for one company, with a lot of help. To get a fairer answer, researchers have run actual tests, posting AI-made content and human-made content side by side on the same accounts and comparing results.
One study posted 24 food photos on the same Instagram account, half made by a person and half made with AI, and tracked what happened. The result: AI posts received more profile visits and reached more new people, while human posts received more comments. In plain terms: AI was just as good at getting seen, but people talked back more when a human made the post.
A second study had people look at content created by AI, by a person, or by both together. It found that AI-only content got less engagement than human-made content, but content made by a person and AI working together performed just as well as fully human content.
The pattern shows up over and over. Pure AI content tends to fall a little short on connection. A person plus AI, working together, often closes that gap completely.
Be careful with the “AI traffic is way worse” stat you’ve probably seen
You may have seen a stat claiming human-written content gets 5.44 times more traffic than AI content. It gets repeated a lot, but it’s weaker than it sounds.
That number comes from one study that’s now two to three years old, and AI tools have improved a lot since then. It also measured blog website traffic, not social media engagement, which is a different thing entirely. Using it to say AI hurts your Facebook or Instagram numbers today is a stretch; the original research never actually claimed.
This is worth remembering any time you see a big, scary AI statistic: check what it actually measured and when, before you let it change how you post.
The real risk isn’t “AI.” Its content that looks cheap
While the engagement numbers are mixed, one thing isn’t: people get annoyed fast when content looks obviously fake or low-effort. McDonald’s and Valentino both pulled AI-made ads this year after people complained. Valentino’s AI ad was criticized almost immediately on Instagram, with viewers saying the imagery looked cheap and off-brand. McDonald’s AI holiday ad got similar complaints, with people saying it felt sloppy and emotionally flat.
This reaction is widespread, not just a few loud critics. About half of US adults say they’d use social media less if AI content in their feed kept increasing, and 85% say obviously fake-looking AI content pulls them right out of whatever they’re watching.
Here’s the useful part: every one of those backlash examples involved content that looked visibly fake or felt rushed. That’s a much narrower problem than “people hate AI.” It’s closer to “people hate content that looks like nobody bothered checking it before it went out.”
We’ve seen this exact pattern before
If this sounds familiar, it should. Back around 2010, a wave of websites figured out they could pump out cheap, low-quality articles and still rank well on Google, just by stuffing in the right keywords. Sites like Demand Media and eHow built huge teams of low-paid writers to churn out shallow articles that still drove traffic and ad money, at least for a while. Then Google changed its search rules to penalize that kind of content, and a lot of those sites lost more than half their traffic overnight, with some shutting down completely.
Something similar is happening again with AI. Once tools like GPT-4 came out in 2023, the same playbook of cheap, high-volume, low-effort content returned, only with AI doing the writing instead of underpaid humans.
Social platforms haven’t cracked down the same way Google did yet. But the lesson is the same either way: shortcuts that work today because nobody’s checking quality tend to stop working once people, or platforms, start checking.
The one thing that holds up no matter how you slice the data
Every angle in this research, the brand success stories, the controlled studies, the backlash stories and the history, points to the same conclusion: a person working with AI consistently beats AI working alone.
That’s not a hedge or a “well, it depends” answer. It’s the single most consistent finding across all sources, even those that disagree with each other on everything else.
What this actually means for your posts
You don’t need a content team or a budget to apply this. A few things you can do right now:
- Use AI to get unstuck, not to finish the job. Let it write you a first draft or a few caption options. Then read it out loud and change anything that doesn’t sound like you. Don’t post the first thing it gives you.
- Read your own captions before you publish. If a caption sounds stiff, generic or like it could belong to any business, that’s the kind of thing people are reacting to when they say content feels fake. Rewrite it until it sounds like a person actually wrote it, because that person should be you.
- Pay attention to comments, not just likes. Likes and views can look fine on AI-heavy posts while comments quietly drop off. Comments are a better sign of whether people actually connected with what you posted.
- Don’t assume today’s AI advice will still be true next year. AI tools and audience patience are both changing fast. Check back in on how your posts are doing every few months, rather than setting a strategy once and forgetting about it.
The data is pretty clear on this: AI by itself is hit or miss; content that looks obviously fake turns people off fast, and the combination of AI plus your own voice and judgment is the safest bet you can make. Try this on your next five posts: write one set fully with AI, edit a second set after AI drafts them, and compare how each group actually performs. The difference will probably tell you everything you need to know.