Using AI Content for SEO - What the Data Actually Shows?

If you've spent any time in an SEO Slack group or Twitter thread this year, you've probably seen someone post a scary-looking traffic graph, a website that used AI to pump out articles, watched rankings spike for a few weeks, then fell off a cliff. It's become the go-to argument for why AI content for SEO is a bad idea.

But is that actually what's happening? Or is it something else, sites publishing low-effort content at scale, and AI just happens to be the tool they used to do it?

A recent study from Ahrefs pulled data from over a million ranking pages to find out, and the results don't match the panic you see online.

What the Study Actually Looked At?

The Ahrefs team analyzed pages pulled from 100,000 search results pages in June 2026, running each one through an AI content detector to estimate how much of the page was written by a machine. They also checked indexation rates and tracked organic impressions over several months to see whether AI-heavy pages lost traffic over time.

A few things stood out right away. AI-written pages weren't rare in the top 10, they showed up at every single ranking position, including the number one spot. About 9% of pages ranking in positions 1 through 3 had an AI content score of 80% or higher, and 5.3% were flagged as fully AI-written. So the idea that Google automatically blocks AI content from ranking well just doesn't hold up.

That said, the study also found a pattern worth paying attention to. As AI content levels went up, average ranking position, indexation rate, and search impressions all went down a bit. Pages with low AI content had an indexation rate near 49%, compared to about 40% for pages flagged as heavily AI-written. Pages written mostly by people also pulled in two to three times the organic impressions of the most AI-heavy pages.

None of this points to a hard cutoff or a penalty switch. It's a gradual slope, not a cliff, which fits with what most of us who work in SEO have suspected for a while: Google isn't grading pages by who or what wrote them. It's grading them on whether they're actually good.

Why This Distinction Matters?

Here's the part that gets lost in most of the discussion around AI content for SEO: a page written by AI and a page that's thin, repetitive, or unhelpful are two different things, they just happen to overlap a lot in practice.

Most AI-written content that struggles in search shares a handful of problems. It repeats things that are already well known instead of adding anything new. It skips internal links, real examples, and images. It reads in that flat, over-explained tone that AI tools default to when there's no editing pass. And it sometimes gets facts wrong in ways a human writer would have caught.

Those are quality problems, and they'd hurt a page's ranking whether a person or a machine typed the words. Our own take on this lines up with what we've written about whether AI content is actually bad for SEO, the tool isn't the issue, the process behind it is.

How to Actually Use AI Content for SEO Without Getting Burned?

If you're going to write with AI tools, a few habits make a real difference:

Start with real research, not just a prompt. AI models are good at organizing what already exists, not so good at telling you something new. Pull from your own data, customer conversations, or original examples before you write, then use the tool to help shape the draft.

Edit like you mean it. A first draft from any AI tool needs a real pass, cutting filler, fixing claims, adding specifics, and making sure it sounds like a person wrote it. If you skip this step, it usually shows.

Don't skip the technical side. Even great content needs to be set up correctly to rank. That means working keyword variations naturally into headers and body text, we cover this in more detail in our guide on how to structure SEO content that actually ranks. It also means keeping an eye on the ranking signals Google has talked about publicly, which we break down in our piece on Google's ranking factors.

Watch your numbers, not just your gut. If you're publishing AI-assisted content at any real scale, keep an eye on indexation and impressions the way the Ahrefs study did. Our breakdown of how AI is affecting search rankings in the US has more numbers worth watching if you want a fuller picture.

Know where the real risk sits. The risk isn't using AI, it's using AI to publish hundreds of pages a month without anyone checking them. We've written before about the common worries people have around using AI in SEO and content marketing, and scale without oversight is the one that consistently causes trouble.

The Bottom Line

AI content for SEO isn't the problem people assume it is. Fully AI-written pages rank in the top 3 of Google right now, and a meaningful share of them stay indexed and keep pulling in traffic for months. What tends to fail is content that's thin, generic, and published faster than anyone can check it, and that would have failed just as hard if a person wrote it.

If you use AI tools as a starting point and put real effort into research, editing, and structure, there's no reason your content should perform worse just because a model helped write the first draft. The sites getting burned aren't being punished for using AI. They're being punished for skipping the work that was always part of good SEO in the first place.

Source: Data referenced from Ahrefs' study, "Google Doesn't Punish AI Content (Here's the Data)".