Google has spent two years telling everyone that AI content is fine, as long as it's helpful. The reality on the ground tells a different story.
Across multiple core updates in 2024 and 2025, sites that scaled content production through LLMs without meaningful human input have lost serious traffic. Some have seen 60 to 80% of their organic visibility wiped out in a single update. The pattern shows up often enough that SEO practitioners now treat "mass AI content" as a known risk factor, no matter what Google says publicly.
Why? It looks like quality assessment is happening at both the page and site level. Google's systems are getting better at spotting content that doesn't add anything not already sitting on the web somewhere else. AI-generated articles, by their nature, tend to be sophisticated paraphrasing of what already exists. They feel comprehensive. They contain nothing new.
That's a problem for the entire content marketing industry. The dominant playbook for the last five years, produce more content, target more keywords, capture more organic traffic, was already showing diminishing returns before AI showed up. Now it's actively risky.
What still works? The same things that worked before, just more obviously now.
Original research with first-party data ranks because nobody else has the data. A B2B SaaS company publishing benchmarks from its own customer base creates something that doesn't exist anywhere else on the internet. Expert interviews and quotes from actual practitioners add information no LLM can synthesise. Strong, specific points of view, arguments that take a position and defend it, get cited and linked. AI summaries rarely earn that.
There's a counterargument worth taking seriously. AI tools genuinely can accelerate good content when they're drafting, structuring, or editing work that originates from a human expert. The distinction isn't AI versus no-AI. It's whether what you publish contains something the world didn't already have.
The uncomfortable implication for B2B marketers: the content team that was supposed to publish 40 posts a month with AI assistance is probably hurting the site. The team publishing four posts a month, each anchored in proprietary data or real expertise, is probably winning.
Volume was never the moat. It just felt like one when the bar was lower.