AI Content Soars Then Crashes: A 20-Domain Warning for Agencies

The pattern was swift and, in hindsight, predictable. SE Ranking seeded 20 brand-new domains with 2,000 AI-generated articles in November 2024, deliberately stripped of any human editorial involvement, backlinks, internal linking, or images. Within 2.5 months, those sites had accumulated roughly 70 to 75 percent of all the impressions they would ever receive, a front-loaded surge that looked compelling on any dashboard. By month three, it had effectively ended.

At the experiment's early peak, the 20 domains logged a 70.95 percent indexation rate and climbed to 122,000 impressions by December 2024. The numbers briefly told a story agencies scaling on AI content would want to hear. Then they told a very different one. By February 2025, the articles had largely vanished from search entirely. Only 3 percent of pages held any position in the top 100 by the three-month mark, a collapse that no amount of patience reversed over the months that followed.

Across the full 16-month run, the 20 domains accumulated just over 526,000 cumulative impressions and a mere 782 clicks. The sites were left completely untouched throughout, with no promotion, no link building, and no manual optimization, specifically to test whether pure AI content could earn traction on merit alone. It could not hold what it initially gained.

The August 2025 Google spam update offered a brief, tantalizing exception. Ten of the 20 domains saw a two-week impression spike beginning around August 26, 2025, as the algorithm briefly reassessed the pages. Rankings in the top 100 climbed back to 20 percent following that update, up from 3 percent at the six-month mark, before settling back into minimal visibility with no sustained recovery recorded.

The contrast with SE Ranking's parallel test on its own established domain makes the underlying dynamic explicit. Six AI-assisted articles published on the SE Ranking blog between June and September 2024 accumulated nearly 555,000 impressions and more than 2,300 clicks over a comparable window, with three of the six landing in the organic top 10. The differentiator was existing domain authority, human editorial involvement, and the E-E-A-T signals Google's algorithm consistently rewards.

For agencies currently selling AI-at-scale as an SEO growth strategy, the experiment defines the ceiling precisely: rapid early indexation, a front-loaded impression window, and then a cliff. Volume and speed did not substitute for uniqueness, authority, or oversight. The 16-month timeline simply gave that lesson more room to compound.