If you are a publisher or content creator today, you have probably noticed the slow leak: search traffic slips, ad revenue thins, and AI widgets answer readers’ questions before they ever reach your site. Jason Calacanis calls this the “Strip-Mining Era of LLMs” because the models vacuum up work that journalists, editors, and filmmakers paid to create, then send little to no traffic back to sustain an ad-supported business model.
Big national outlets can cushion the blow with millions of subscribers, but smaller publishers, local newsrooms, and niche blogs don’t have that luxury. For them, every lost page-view hurts.
When a Stormtrooper Goes Viral
Last weekend, my eight-year-old daughter and I shot a one-minute parody using VEO 3. Picture a stormtrooper giving career advice, Darth Vader hovering in the background, Ewoks staging chaos. We never asked Disney for permission. The clip looked great and took almost no time to make.
If two people with a laptop can remix Star Wars IP in an afternoon, imagine what a billion-parameter model can do with your reporting, your photos, or your paywalled archives.
This is the same drama is unfolding in the music industry at the moment. AI “bands” rack up streams while labels scramble to prove that these models were trained on their artists’ work without permission.
The Game of Chicken
Publishers have started to fight back. The New York Times is suing OpenAI, but models have already trained on oceans of questionably sourced data. Even if courts demand a purge, few engineers believe the material can truly be “unlearned.” Meanwhile, other outlets leave their pages open to crawlers, hoping for reach. The result is a classic game of chicken: opt out and risk invisibility, opt in and lose leverage.
Reddit was one of the first to jump in, signing a licensing deal with Google. For months, Reddit threads (often frustratingly) ranked above brand sites and news pages, spiking the platform’s traffic, but forcing users to find an answer from the contribution of a rando on the internet in a thread 30-comments deep .Yet the advantage may be fleeting. As soon as AI Overviews summarize those threads, users have less reason to click through, and Reddit’s ad impressions slide.
Smaller publishers face the same trap. Block crawlers and vanish; allow them and get zero guarantee of money or audience.
Regulation might eventually help, but the clock has already run out. The EU’s AI Act and Australia’s News Media Bargaining Code still quarrel over news links on social platforms, while billions of scraped words are already baked into model weights that can’t simply be pulled back out. Lawmakers are inching forward; the models have sprinted past the finish line.
Two Paths to Leverage
So what do you do when you cannot out-lobby a trillion-dollar platform? You choose one of two lanes.
- Scale: If you have decades of archives, own them boldly. Large, trustworthy datasets give AI companies credibility, which means they will pay to license them. AP, News Corp, and Axel Springer have already signed eight- and nine-figure deals.
- Specificity: If scale is out of reach, become irreplaceable. I have roasted coffee at home for twelve years, yet even I was surprised when James Hoffmann (my favorite coffee creator) released a nineteen-minute deep dive on V60 paper filters and 313k+ people watched in five days. That level of obsessive detail cannot be faked by a generic model. Niche audiences crave it, and sponsors will pay for the attention.
Pick a lane and double down. Own the data or own the depth.
What a Fair Deal Looks Like
Even with a clear lane, small publishers still need a platform that shares the upside. YouTube pays creators because its ecosystem would collapse without them. AI needs a similar revenue-share model, or the well of quality content will dry up. Until that happens, your best defense is to build direct loyalty: email lists, paid communities, live events, Patreon tiers, or anything that moves the relationship off intermediaries and into your own hands.
Quality should win, but only if the economics catch up. Push for a better split, and in the meantime make your content either so vast or so specialized that the machines cannot run without it.