Blog / AI & search /
AI-written content and Google: what Google has actually said
Since ChatGPT arrived, I’ve seen two confident claims doing the rounds. One: “Google can detect AI content and will penalise your site.” Two: “Google said AI content is fine, so publish as much as you can.”
Both are wrong, and both cause real damage. The first scares people off a useful tool. The second fills the web with pages nobody needed.
Google has actually been fairly clear about this. It’s worth reading what it said, rather than what people say it said.
The short version
In February 2023, Google published a post on its Search Central blog called Google Search’s guidance about AI-generated content. The core message:
- Google rewards high-quality content, however it’s produced.
- What it’s against is content made primarily to manipulate search rankings, whether a person or a machine made it.
- Using automation, including AI, to generate content mainly to game rankings is against its spam policies.
So the question Google asks isn’t “was this written by AI?”. It’s “is this helpful, and why was it made?”.
What the guidance actually says
Quality over method
Google pointed out that it has dealt with mass-produced content for years, long before AI writing tools: spun articles, scraped pages, templated junk. Its systems are built to judge the result, not the keyboard.
It also said directly that appropriate use of AI or automation isn’t against its guidelines. It gave examples where automation has long been useful, like sports scores, weather forecasts and transcripts.
Spam is still spam
The other side is just as clear. Google’s spam policies already covered automatically generated content intended to manipulate rankings. AI doesn’t create an exception. Generating hundreds of pages from a keyword list and publishing them unchecked is exactly what the policy describes.
E-E-A-T still applies
The guidance ties back to E-E-A-T: experience, expertise, authoritativeness and trust. AI can write a smooth paragraph about a hotel, but it hasn’t stayed there. It can explain a WordPress setting, but it hasn’t broken a site by changing it. That first-hand experience is exactly what Google added to its quality rater guidelines in December 2022. (More in E-E-A-T on a personal blog.)
Who, how and why
Google suggests thinking about three questions for any content:
- Who created it? Is it clear to readers?
- How was it created? Where automation played a big part, would readers expect to be told?
- Why was it created? Mainly to help people, or mainly to attract search visits?
It also said giving AI an author byline is probably not the best way to go about disclosure. The byline should be the person or organisation responsible for the content.
What the guidance doesn’t say
- It doesn’t say Google can reliably detect AI text. It talks about judging quality, not about detectors. I’d be very wary of anyone selling an “AI detection score” as an SEO metric.
- It doesn’t say AI content will rank. “Not against the rules” is a long way from “good enough to beat everything else on page one”.
- It doesn’t give a safe percentage. There’s no “up to 30% AI is fine”. The method isn’t the measure.
What I think it means in practice
AI makes it easier to produce more, but producing more isn’t automatically progress.
The risk I see isn’t a penalty for using ChatGPT. It’s that AI makes it effortless to publish the same average answer everyone else can publish. If your post says what the top ten results already say, in slightly different words, there’s no reason for it to rank, however it was written.
So the useful split is between what AI is good at and what only you can add:
Reasonable uses:
- researching and outlining a topic
- drafting a first version you then rewrite properly
- summarising notes or transcripts
- suggesting titles and meta descriptions to check and edit (I covered this in drafting meta descriptions with ChatGPT)
- grouping keywords
What it can’t supply:
- your examples, results and mistakes
- your screenshots and data
- your opinion about what actually works
- accountability for whether the facts are right
And accuracy matters a lot here. ChatGPT states wrong things with total confidence. Every fact, number and claim in an AI-assisted draft needs checking by someone who knows the subject.
A simple test before publishing
Ask: if a reader found out how this post was made, would they feel misled? And would they still think it was worth their time?
If the honest answers are “no” and “yes”, the method probably isn’t your problem.
Related reading
- ChatGPT for SEO: what it can and can’t do yet
- GPT-4 is out: what changes for SEO content
- Prompting ChatGPT for SEO briefs that aren’t generic
- E-E-A-T on a personal blog: showing real experience
Read Google’s guidance yourself. It’s short, and it’s on the Search Central blog. Then look at your last AI-assisted post and ask the who, how and why questions honestly. What would you change?