Search Engine Land Releases 16-Month Study on AI-Generated Content Performance

I’ve been following the debate around AI content for months, and Search Engine Land just added one of the strongest data points yet. In a newly published 16-month study led by SE Ranking and covered by Bogdan Babiak, we finally get a longer-term look at what happens when websites rely on fully AI-generated content with no human editing, no backlinks, no internal linking, and no promotional support.

The main takeaway: AI content can get indexed quickly and generate early visibility, but by itself, it struggles to maintain rankings over time.

What the study actually tested

The experiment started around November 2024, when SE Ranking bought 20 brand-new domains across 20 different niches. Those niches ranged from hobbies and business to more sensitive categories like health and finance.

Each domain published 100 fully AI-generated how-to articles, creating a total of 2,000 pages. Every page targeted long-tail, low-competition informational keywords. The setup was intentionally barebones so the team could test raw AI output on its own:

• No human editing
• No rewritten sections
• No images
• No internal links
• No backlinks
• No promotion
• No authority signals beyond submission to Google Search Console

That setup matters because it stripped away nearly all of the usual ranking support. What was left was a clean test of whether AI content alone could perform in Google Search over time.

The early results were promising

In the first 36 days, Google indexed 70.95% of the pages, a strong start for brand-new domains with no established authority. The sites generated 122,102 impressions and 244 clicks, and 80% of the domains ranked for more than 100 keywords. At that point, 28% of indexed pages had reached Google’s top 100.

For anyone hoping AI could speed up SEO, those early numbers looked encouraging. The study showed that Google was not automatically rejecting AI-generated content. It was willing to crawl, index, and test it in search results.

That matters because it pushes back on the idea that AI content gets instantly buried or blocked. Based on this study, Google gave these pages a shot.

Then the decline set in

By months two and three, cumulative impressions climbed to 526,624 and clicks reached 782. On the surface, that still looked like momentum. But the longer view told a different story.

Between months three and six, impressions rose to 706,328 and clicks to 1,062, yet rankings weakened sharply. The share of pages in the top 100 dropped from 28% to just 3%. Indexation also slipped, with only 61.45% of pages still indexed.

By month 16, the sites had generated 1,092,079 cumulative impressions and 1,381 clicks. The most revealing detail was that roughly 70% to 75% of all performance happened in the first two and a half months. After that, growth mostly flattened.

So yes, the content got visibility. It just didn’t keep it.

Why this matters for SEO

This study reinforces what many marketers have suspected for a while: AI can help produce content at scale, but scale by itself is not a ranking strategy.

Google appears willing to test fresh AI pages, especially when they target low-competition topics. But without stronger trust signals, authority, originality, and deeper usefulness, those pages tend to lose momentum. That lines up with Google’s broader focus on helpful content and E-E-A-T.

Bogdan Babiak’s summary captured it well: AI content wasn’t useless, but AI alone wasn’t enough.

That distinction matters. This wasn’t a blanket rejection of AI-assisted publishing. It was a warning against publishing raw, untouched AI content and expecting durable SEO performance.

YMYL niches looked especially weak

One of the clearest findings was how poorly AI-only content performed in YMYL categories like finance and health.

By month 16, finance pages had just a 9% indexing rate, while health pages were at 14%. That’s a dramatic gap compared with less sensitive niches and a strong sign that Google applies stricter standards where trust and expertise matter most.

If you publish in a niche involving money, medical advice, legal risk, or personal wellbeing, this data makes the risk hard to ignore: relying on unedited AI content is a serious gamble.

A late twist: fresh content briefly lifted older pages

The study also included a bonus experiment in March 2026. New AI content was added to eight of the original sites, and early results suggested that publishing fresh pages could temporarily increase impressions on older content.

The examples were notable. One business site jumped from 458 impressions to 7,750. A law site climbed from 19 to 356. Those gains suggest that site activity itself may act as a signal, or at least trigger renewed crawling and ranking adjustments.

It’s too early to call that a lasting recovery pattern, but it raises an important question: can a steady publishing cadence help support aging content, even when that content is AI-generated? That’s worth watching.

What marketers should learn from this

If there’s one lesson here, it’s simple: AI is an amplifier, not a substitute for strategy.

The study makes several points clear:

• AI content can be indexed quickly
• It can generate early impressions and some clicks
• It may rank briefly for low-competition terms
• It is far less reliable without authority and editorial support
• It performs especially poorly in trust-sensitive niches
• Long-term SEO still depends on signals beyond raw content volume

That means the strongest use of AI is probably the one many experienced teams are already adopting: use it for drafting, outlining, research support, and scaling production workflows, while keeping human editing, subject expertise, structure, and SEO judgment in the loop.

In short, AI works best when it has direction.

The bigger industry message

This study arrives at a time when search is changing fast. AI Overviews, generative search experiences, and shifting user behavior are already making traditional SEO less predictable. In that environment, shortcuts become even less reliable.

What SE Ranking documented here is bigger than a verdict on AI content. It’s a reminder that search performance is built on trust, relevance, and reinforcement. Content without authority may get tested. Content with authority, clarity, and editorial care is far more likely to last.

That’s the difference between a short-term bump and a real content engine.

FAQ

Does Google automatically penalize AI-generated content?

No. This study suggests Google is willing to crawl, index, and initially rank AI-generated content. The bigger issue is long-term performance, not automatic suppression.

Can AI content rank in Google?

Yes, especially for long-tail, low-competition informational keywords. But rankings may fade over time if the content lacks authority, originality, and editorial support.

Why did the AI-only pages lose momentum?

The study points to missing trust and quality signals. With no human editing, backlinks, internal links, images, or promotion, the content had little reinforcement once Google finished its initial testing phase.

Is AI content riskier in health and finance niches?

Yes. The weakest results appeared in YMYL categories such as health and finance, where Google appears to apply higher standards for expertise and trust.

What’s the best way to use AI in content marketing?

Use it as a support tool for research, outlining, drafting, and workflow efficiency. Then add human editing, subject-matter expertise, SEO structure, and clear quality control before publishing.

Conclusion

This 16-month study offers a useful reality check for marketers. Pure AI content is not invisible to Google, and it isn’t instantly penalized. But if the goal is sustainable growth, raw automation isn’t enough. The smarter approach is to pair AI speed with human oversight, niche expertise, and solid SEO structure. If you want a more strategic way to scale content without falling into the trap this study exposed, it’s worth looking at AIuthority as part of a more durable, quality-first publishing approach.