Freshness and Third-Party Proof Beat Generic ‘AI Content’ in Citation Studies

When researchers re-run the same buyer prompts days apart, citation sets churn. One tracker noted Claude keeping only a minority of URLs across a two-day gap, with Perplexity relatively more stable. That volatility punishes thin pages that exist only to “be in the training data someday.”

What keeps getting cited

Patterns across September-era write-ups:

• Comparison and review destinations
• Primary documentation and pricing pages
• High-signal community threads and video explainers (especially for Perplexity-like systems)
• Clear entities: who you are, what category you occupy, what you cost

What to stop doing

Publishing interchangeable AI-generated posts that restate the SERP without adding a fact, dataset, or first-hand test. Answer engines already have the mush. They need sources that resolve uncertainty.

AEO Soup’s simmering pot this week: treat AI visibility like reputation plus retrieveability — not like a new keyword density trick.