How AI Content Kills Your SEO (16-Month Experiment)
A 16-month SEO experiment across 20 websites and 2,000 AI-generated articles shows that raw AI content experiences rapid initial success followed by dramatic ranking collapse.
Within 36 days, 71% of pages were indexed and early metrics suggested the strategy worked, but by month three, rankings plummeted from 28% of pages in the top 100 down to just 3%.
Google does not penalize AI content for being AI—it simply doesn't rank it long-term without original insights, backlinks, expertise, and site authority.
However, a follow-up March 2026 experiment revealed AI content does work in a different way: adding new AI articles to dormant sites triggered a "freshness signal" that boosted impressions on old posts by up to 19-fold.
When AI content was published on an established domain with human editing and real expertise, six articles generated over 500,000 impressions and three ranked in the top 10 search results.
The core lesson is that AI must be treated as a tool within a strategy, not a content strategy itself—scaling with AI alone is a trap unless humans verify, fact-check, and add genuine expertise.
If you or your team is relying on AI alone to write website content— our 16-month experiment proves this strategy is killing your SEO. We launched 20 brand-new domains across 20 different niches and flooded them with 2,000 pieces of raw AI content. At first, it looked like the shortcut everyone's been hoping for. We saw massive indexing and traffic spikes that would make any SEO jealous. Then Google killed it. But what matters isn't that it failed — it's when and why it failed. Because Google follows a very specific pattern with AI content, and if you're publishing it right now, you're somewhere in that pattern. I'll also show you something we found in March 2026 that actually reversed the damage. Plus, we’ll look at what happened when we published AI content on an established blog with real authority. So we published everything. 20 sites across 20 niches — everything from Finance and Law to Pets and Hobbies. 100 AI articles each.
No editing, no backlinks, no human touch. Within just 36 days, 71% of our 2,000 pages were indexed. In fact, 11 out of the 20 sites had every single page live. Here's what's interesting. Google wasn't treating every niche the same. Broad, evergreen topics like Home & Garden got indexed fast. Ecommerce and specialized niches got throttled. But indexing is just the start. We got over 122,000 impressions in the first month in total across all niches. Just look at the charts here. 80% of the sites were already ranking for hundreds of keywords. Some for thousands. Now you might think — this is it. AI content works. No backlinks, no editing, just publish and go. But to add fuel to the fire — by month two and a half, it got even better. Impressions hit 526K and clicks tripled. Twelve sites crossing 1,000 keywords. At this point, most people would've gone all in. But the numbers going up and Google being
impressed are two different things. What looks like Google rewarding your content is actually Google testing it. By month three, our "AI Honeymoon" was officially over. Pages in the top 100 went from 28% to just 3%. Let that land. Twenty-eight percent down to three by month 6. Google didn't deindex the pages. It just stopped showing them to anyone. But here you might see the growth in the dashboard after this point. Well, 70 to 75% of all impressions and clicks happened in the first two and a half months. Everything after that is leftovers. What this means is that if you're measuring an AI content strategy based on those early months — you're looking at a mirage. So why did it happen? Honestly — it's not complicated. The content had no backlinks. No original data. No real author. No first-hand experience. No internal linking tying topics together. On top of that, it read like a slightly reworded
version of whatever was already ranking. Google had no reason to trust it over anyone else answering the same question. As a result, it didn't. Now you'd think the story ends here. But we kept watching for another year and something weird happened. In late August 2025, Google rolled out a spam update. Half our sites got a short burst of impressions. Pages in the top 100 climbed from 3% back to 20%. But most of them lost the boost within two weeks. Now, here's a detail that says a lot: our Finance site had 9 pages indexed out of 100. Health had 14. These are the niches where Google cares most about expertise. AI content with no credentials just got erased. As a result, after 16 months and across 2,000 articles we had just over 1 million impressions and more than 1,000 total clicks. That's less than one click per article per year. This raises the question: is AI content completely useless? No. But we ran one more experiment in March 2026 and the results
flipped everything — because it has nothing to do with ranking the AI content itself. So we added new AI-generated articles to 8 of these same sites. Same approach — fully AI, no editing. But this time, the result was completely different. The new articles weren't even fully indexed yet. But the old posts — the ones that had been dead for months — suddenly started getting impressions again. One site went from 458 impressions in February to 7,750 in March. That's a 17-fold jump. Another law-focused website had a 19-fold boost. A science-focused site also saw their impressions increase by 19 times. I want to highlight the fact that the traffic wasn't going to the new content. It was going to the old pages. Okay, what's exactly happening here? Publishing new content — even AI content — tells Google the site is alive. That freshness signal lifts the whole domain. Pages that were invisible for a year suddenly get another shot.
That's a completely different use case. The original experiment asked: can AI content rank on its own? No. Not long-term. But this follow-up shows that AI content can support pages that already have something going for them. The content isn't the asset. The activity is. Now — these are early results. They might be temporary. But for a site that's been flatlined for months, that boost could be exactly the window you need. So what does this actually mean and what did our blog experiment show? First things first, don't stop using AI. That's not the point. The point is — stop treating AI content like a strategy. It's a tool inside a strategy. Here's what 16 months of data actually tells you: Google doesn't penalize AI content for being AI. We see a 71% indexation on brand-new domains, which means the search giant is clearly willing to crawl and rank AI content. But those early rankings are just a trial. You get about three months. If the content
doesn't prove itself in that window, it drops and it doesn't come back. What "proves itself" means here is that the content has original insights, real expertise, backlinks, internal linking, and site structure. The stuff AI can't generate on its own. In order to test whether AI content can actually work long-term on an established blog, we ran one more experiment on our blog. We published six AI-generated articles written by our AI Writer on the SE Ranking blog. Same AI starting point. But this time with human editing, fact-checking, and real expertise behind the content. As a result, we got — wait for it — over half a million impressions. Over 2 thousand clicks in a year. Three articles were ranking among the top 10 organic search results. Four cited as sources in AI Overviews. Plus, the traffic kept growing month over month instead of collapsing at month three. Now — the SE Ranking blog already has domain authority. This is obviously a factor. But that's kind of the point. AI content
on a trusted domain with human expertise behind it performs. AI content on its own, without either of those — doesn't. So yeah — AI speeds up creation. But scaling content with AI is clearly a trap if you do it without ever including a human in the loop. Now, this is just one experiment. We're constantly running studies like this. If you want to stay ahead of how Google is actually treating AI content, SEO strategies, and overall search visibility — there's more on the channel. Make sure to subscribe and give us a like!