随着AI内容激增，社交媒体的轻松曝光正在消失，创作者需调整策略以降低对平台算法推荐的依赖。
I saw a bold prediction from my friend Kallway this week:
We have ~24 months until all social feeds go pay to play.
Free organic reach will drop to <10% of what it is now.
This will reset attention economy and break most business models.
Most are not positioned accordingly.
Get it while you can.
— Kallaway (@kanekallaway) July 18, 2025
If you're not familiar with Kallaway's stuff, he's really impressive. He obsesses over and wins social algorithms better than just about anyone I know. So when he talks about anything related to social, I listen.
He expanded in the replies:
1. AI content becomes indistinguishable 2. AI content flood feeds 3. Human:AI goes from 100:0 to 10:90 4. Human stuff can’t cut through 5. Human creators protest 6. Social platforms offer paid tier to cut through
Two weeks ago, I saw a (now-deleted) video on YouTube from Jake Fellman, a creator with ~26M subscribers via YouTube Shorts and ~14M on TikTok. He's an animator specializing in short-form vertical video – really high-quality, attention-grabbing stuff. But this video he published on YouTube caught my attention because it was long-form vlog style (he rarely shows his face). In this video, titled something like "AI Took My Job," he explained how AI-generated short-form video has out-competed him in different short-form feeds.
He said the video itself was a bit of an experiment. His animations typically take ~2 weeks to create, but AI-generated animations can be created in a matter of minutes. He was so frustrated with his loss of views that he was trying anything he could to publish more frequently – hence this long-form video.
The video was hard to watch. Jake was visibly distressed. Again, paraphrasing, he said something to the effect of, "I don't earn a living as an artist. I'm a content creator. And when my content doesn't get views, I don't make any money."
Back to Kallaway...
His prediction seems wild, but the more I've sat with it, the more I tend to agree with it. We're already seeing a ton of new, AI-generated content in our feeds. So much so that YouTube changed their guidelines to give itself the flexibility to de-monetize AI-generated content.
The ability to generate bottomless content, made available to anyone on the planet, is radically tilting the supply side of supply and demand on every platform. For those who don't care about building a brand and simply want to maximize views, why wouldn't they just throw as much AI-generated spaghetti against the wall from as many accounts as possible? Most of it will fail, but at some point, this becomes a numbers game: enough total content means that there will be enough positive outliers competing with human-made content in the feed.
On one hand, it's easy to be skeptical and think that social platforms will embrace AI-generated content. After all, more content is good, right? They may not even need to maintain monetization programs for creators if they can generate content with AI. Algorithms + AI: a match made in heaven.
Maybe. They're certainly going to experiment. These companies may generate their OWN content using AI to see if they can fill human users' feeds themselves.
But content alone isn't enough.
The incentive of any social platform (including YouTube) is to maximize human consumption time. Human consumption means more ad impressions and (crucially) more purchases from ads. So, they will do whatever is best for ad revenue (which is almost certainly related to human consumption time). If AI-generated slop expands human consumption time (and that trend seems like it will continue), then they will prioritize that content alongside any other human content that is getting engagement.
If, however, ads placed in/around AI-generated content convert so much worse that it threatens their ads business, they'll find a way to promote human content.
So what's that mean for humans like us who are putting care into our content and trying to make a living from it?
Algorithm-proofing
I've said for a long time that discovery platforms (social media) should be used to help you reduce your dependency on discovery platforms. That means quickly de-platforming social platforms to your relationship platforms: email, podcasting, private communities, and/or SMS.
I call these relationship platforms because they truly facilitate deeper relationships. The emphasis isn't on hooks or new viewers. It's not attention attraction, it's attention retention.
If you've built your platform on social media, that ground is increasingly shaky. I would prioritize investing in some platform(s) you have more control over for the long term. My recommendation for most people is email, and for email, I use and recommend Kit (affiliate).
If you have the stomach for it, consider starting a podcast. A major contributor to trust is time spent, and if people want to spend more time with you, nothing beats podcasting.
Let me be clear: podcasting is the most difficult platform to grow. But that difficulty is a feature, not a bug. If you succeed in podcasting, you'll have a resilient content business.
I've also never been more bullish on community. The Lab is built on Circle (affiliate), which I highly recommend.
AI will take a huge chunk out of today's self-paced learning materials like courses, but community-powered experiences will grow. But another word of caution: having a community or membership isn't enough. As the number of communities grows, people will get stretched thin and pull back to one or two they truly love. Your aim should be to make the cut.
Conclusion
The last 15 years have been an audience-building gold rush fueled by social platforms. It's an aberration, not a permanent state. For the last 15 years, there's been more demand than supply for great content. As AI flips that equation, that gold rush will become much more competitive (if not run dry).
But have faith! Strong, audience-driven businesses existed before this gold rush and will exist after it's over, too. Good old-fashioned word-of-mouth will continue to reign.
That new world won't be won with attention. It'll be won with trust.