If you’re running a content operation right now, you’ve probably had this argument in a Slack channel at least once: does AI-written content hurt your rankings, or is that just SEO folklore?

The short answer, straight from Google, is no. Google has said since February 2023 that its ranking systems “reward original, high-quality, people-first content, however it is produced.” Not “written by a human.” However it is produced. That line hasn’t changed since Google published it, and it’s still the clearest statement of policy anyone has to work from.

What Google does penalize is scaled content abuse: publishing large volumes of low-effort pages with the primary goal of manipulating search results. That’s a policy about intent and quality, not a policy about which tool typed the words. A human writer churning out 200 thin, templated pages a week to farm long-tail keywords is doing exactly what that policy targets. An AI-assisted page built on real research and reviewed by someone who knows the subject is not.

So why does so much AI content still tank?

Because most of it skips the work that used to be forced on writers by the sheer effort of writing. Before, if you wanted a page to rank, you had to research the topic yourself, which meant you accidentally learned something about what was already ranking. Now a tool can generate 1,500 words in ten seconds with none of that friction, and that’s exactly the problem. The content fails not because a model wrote it, but because five things are usually missing:

No one looked at what’s currently ranking for the target query before drafting. No one added anything the writer or brand actually knows first-hand, so there’s no data, no case study, no experience underneath the sentences. The stats and claims went in unverified, pulled from wherever the model’s training data landed rather than from a primary source. The piece was written with no sense of who the audience is or what the brand’s actual position is. And nobody with real knowledge of the topic read it before it went live.

Strip out research, verification, and expertise, and you get pages that read fine but say nothing that wasn’t already said better somewhere else. That’s the “unoriginal, low-quality” bucket Google’s systems are built to catch, and it would catch a human writer doing the same shortcut just as readily.

There’s some evidence Google’s systems are getting better at catching exactly this. When Google rolled out its March 2024 core and spam updates, it reported a 45% reduction in low-quality, unoriginal content showing up in search results, ahead of the 40% it had initially targeted. That update wasn’t aimed at AI content specifically; it was aimed at content with no original substance, wherever it came from. AI-generated pages made up a disproportionate share of what got hit, simply because a disproportionate share of that content skips the steps below.

The process that actually protects a ranking

Treat AI as a drafting tool inside a process, not a replacement for the process. In practice that looks like seven steps: pull and read what’s currently ranking for the target keyword, build an outline around the gaps and angles competitors missed, feed the model real context about the brand and the audience instead of a bare topic prompt, draft against that outline rather than asking for a finished article cold, verify every factual claim against a primary source before it goes near the page, add something the model couldn’t have known, and have a person with actual knowledge of the subject read it before publishing.

That splits the work in a specific way. AI is genuinely good at speeding up research, generating angles, building a first-pass outline, producing a draft, and catching typos. It’s not good at deciding what’s actually true, contributing a real opinion or a result nobody else has, knowing who the content is for, or being the last check before something goes public. Those stay with a person, every time, no exceptions.

What this means if you’re deciding whether to use AI at all

The question was never “AI or no AI.” It’s whether the five failure points above get fixed before publishing, regardless of what wrote the first draft. A brand that runs AI-assisted content through genuine research, verification, and expert review is doing exactly what Google says it rewards. A brand publishing unverified, unreviewed AI output at volume is doing exactly what scaled content abuse policies exist to catch. Same tool, opposite outcome, and the difference has nothing to do with the tool.