AI content sometimes performs well right out of the gate, but that early strength doesn’t hold. A survey of 100 companies tracked performance of AI-generated text, images, and video against a human-generated content baseline over a full year, and the decline for text content is steep and sustained. Images tell a different story, dipping briefly before recovering back toward their starting level, while video settles into a low, flat performance well below the baseline for the rest of the year.

Essential Statistics

• AI-generated text starts at a performance score of 1.3, above the human-generated baseline of 1.0, but falls to 0.1 by the beyond-one-year mark.
• AI-generated images start at 0.8, dip to a low of 0.6 in months 7 to 9, and recover back to 0.8 by the one-year mark and beyond.
• AI-generated video starts at 0.6, drops to 0.3 by months 2 to 3, briefly rises to 0.4 in months 3 to 6, then settles at 0.2 from month 7 onward.
• Text is the only format that starts above the human-generated baseline of 1.0, and it also experiences the steepest overall decline.
• By the one-year mark, both text and video sit well below the baseline, while images are the only format that returns close to its starting performance.

Key Takeaways

• Text starting above the human baseline but ending far below it suggests any early performance advantage from AI-generated text tends to be temporary rather than durable.
• Images following a dip-then-recover pattern suggests this format may be more resilient over time, possibly because visual content ages differently than written content in search and engagement contexts.
• Video settling into a low, flat performance suggests this format struggles to gain traction at any point in its lifecycle, not just at launch.
• The steep and sustained decline in text performance points toward AI-written content losing relevance or ranking power faster than other formats over time.
• This pattern suggests format matters significantly when deciding how much long-term risk to take on with AI-generated content.

Actionable Insights

• If you’re publishing AI-generated text content, plan for a performance decline starting within the first few months rather than assuming early results will hold. Building in a refresh or human-rewrite cycle around that window is worth testing.
• Treat AI-generated images as a more durable content investment than AI-generated text or video, since this data shows images recovering back toward baseline performance over the course of a year.
• Reconsider heavy investment in AI-generated video given how consistently low its performance stays throughout the tracked period. This format shows the least recovery of the three.
• Set up a monitoring cadence to track your own AI content performance by format at the 3-month, 6-month, and 9-month marks, since this data suggests decline patterns vary significantly by format and timing.
• Use early strong performance from AI text content as a signal to invest more in human editing quickly, not as validation to scale that content further, since the data shows that advantage fades fast.

Text starts hot and fades the fastest. Images hold up best. If you’re building an AI content strategy, that alone should shape how you invest by format.’ – Neil Patel