Focus: AI content that survives (SEO/AEO)
🔴 Impact: High
🟢 Difficulty: Low
Hey {{First Name | there}},
What if we’ve got AI content depreciation completely backwards?
I ran a 900-article experiment and found something neither of this year’s two biggest AI content studies reported, and so far I haven’t seen anyone else report it either.
AI content does not necessarily have to depreciate.
There is a point on the editing ladder where AI pages stop losing ground and start finishing the year ahead of where they began.
And here’s where it gets interesting: whether you can reach that point depends on what trade you’re in.
A plumber gets most of the way there on first-hand detail alone.
A bookkeeper doesn’t get close.
And neither of them actually crosses until they do the boring structural work.
Then there’s the fix everyone seems to sell: slap a credentialed author on the page, add the schema, and call it trustworthy.
That works.
Just not nearly as well as people imply.
In my data, it pushed sixteen-month survival from roughly 11% to about 24%, and then hit a ceiling.
Meaning three pages out of four were still dead at the end.
That matters because the two big 2026 studies appear to tell completely different stories.
In March, SE Ranking followed 2,000 unedited AI articles and watched them collapse from 28% to 3% at the three-month mark.
Then, in July, Ahrefs looked at 331,000 pages and reported that AI content does not decay at all.
So which one is right?
Both.
They just measured different things.
I published 900 articles across six brand-new sites at five levels of human input, and followed every site from launch through month twelve.
And that’s where the picture gets a lot more interesting.
What I ran
I called it the editing ladder, and it ran for twelve months.
Nine hundred AI-written articles, six brand-new websites, and the same piece of content published 5 times over, with one more bit of human work added at every step.
Level 1 was raw AI. Level 2 added the copyediting most people mean when they say editing.