How Google Judges AI-Generated Content Now: Effort, Originality and Who Wrote It
Google judges AI-generated content by how much human effort went into it, whether it adds something that isn't already online, and whether the person it says wrote it is real. That's now in writing, in the helpful content guidance Google updated in early October, and it includes a new line calling fabricated author profiles (AI headshots, invented names, made-up credentials) "a form of deception."
None of that bans AI. It does put a name on the thing a lot of content programs have been doing for two years, which is publishing large amounts of AI text that nobody checked very hard, under a byline that may or may not belong to a person.
The short version
Google's guidance now lists four things its quality raters judge in a page's main content: effort, originality, talent or skill, and accuracy.
Using AI to produce lots of text "without manual oversight or curation" is Google's own example of little to no effort.
Fake author profiles are now called out by name as deception and a low-quality signal.
The same systems feed AI Overviews and AI Mode, so this affects whether you show up in AI answers too.
What Google changed in the helpful content guidance
The page is called "Creating helpful, reliable, people-first content," and if you've done SEO for any length of time you've probably read some version of it. Search Engine Journal noticed the changes in the version dated October 1 (they compared it to a saved copy from June 8 that didn't have them), and Google has kept editing the page since; the live version I read says it was last updated October 5 (Search Engine Journal).
Three things are new. There's a definition of main content, which Google describes as any part of the page that directly helps it do its job, including the text, images, interactive tools, comments and headings. There's a section on the four attributes raters use to judge that main content. Then there's a paragraph, sitting under the existing advice about accurate bylines, that tells site owners to "avoid using deceptive authorship information" (Search Engine Roundtable).
Barry Schwartz pointed out that Google had previously told him it wasn't worried about faked author information. So that's a change of tone, at the very least.
How Google judges AI-generated content, in its own examples
Google describes effort as the extent to which human work went into creating the content or the systems behind it, and then gives examples on both ends.
High effort, per Google, looks like writing original analysis, translating a poem by hand, or building a custom interactive map for a game. Little to no effort looks like automatically generating pages from feeds, or "using generative AI to produce large amounts of text without manual oversight or curation" (Google Search Central). There's also a line saying that attribution or credit to other sources doesn't replace the need for original effort. So a roundup that cites twelve studies and adds nothing of its own is still a roundup.
Google's helpful content guidance, Oct 2026
Four things raters judge in your main content
Source: paraphrased from Google Search Central, Creating helpful, reliable, people-first content (last updated Oct 5, 2026). Google says rater data isn't used directly in ranking. Columns not labeled as Google's examples, and the "What it means" notes, are Jarred Smith's reading.
Look at what Google's low-effort examples have in common, which is volume and a lack of oversight, and note that "the systems powering it" is in the definition on purpose. If somebody on your team built a calculator with AI's help, that's still a person deciding what the calculator should do.
One caveat, because I don't want to oversell this. The page frames these as things Search Quality raters are trained to evaluate, and it says elsewhere that rater data isn't used directly in ranking. Raters are how Google checks whether its systems are working. So I'd read this list as a description of what the systems are being tuned toward, which is close to the same thing for planning purposes, though not identical.
What the ranking data says about AI content
If you're wondering whether Google punishes AI content outright, the data I know of says no, it mostly doesn't care how the words got made.
Ahrefs ran its AI detector over the top 20 results for 100,000 random keywords, about 600,000 pages, and found 86.5% of them had some AI-generated content in them. Only 4.6% were purely AI and 13.5% were purely human. The correlation between how much AI a page used and where it ranked was 0.011, which is about as close to zero as you get (Ahrefs).
Graphite looked at it from the other direction. Using Surfer's detector on a sample of 65,000 articles from Common Crawl, they estimated that by November 2024 about half of new articles on the web were mostly AI-written, 50.3% to be exact (Graphite). When they checked what ranks and what gets cited, the share was much lower. In Google results, 14% of articles were classified as AI-generated. In ChatGPT and Perplexity citations, 18% (Graphite).
Both studies lean on AI detectors, which make mistakes (Ahrefs says so, and Graphite reports a 4.2% false positive rate for the detector it used), and both are from 2025. Still, put them next to each other and the picture is pretty consistent with the new guidance. Lots of pages that rank have AI in them, and mostly-AI pages make up a much smaller share of what ranks and gets cited than of what gets published. My guess is that most of that gap comes from the effort and originality stuff Google just wrote down.
What this means for AI Overviews, AI Mode and other AI answers
In May, Google published a guide on optimizing for its generative AI features and said, more or less, that optimizing for AI search is "still SEO" (Search Engine Journal). I wrote about that back in June, in Google Just Called Most GEO Tactics Useless, and my take hasn't changed. For Google's own AI Overviews and AI Mode, the pages that get pulled into an answer come out of the same index and the same quality systems as everything else. A page Google's systems see as low effort is a weak candidate for a citation, too.
The other engines do their own thing, but they read the same web. In the AI Verdict Study, when I turned web search off, Claude told a buyer that Huntress needed in-house security expertise and rejected it both times. With search on, the same model found a description of the 24/7 human team and said yes six times out of six. The engines are building their answers out of whatever they can find written about you, and what's been written about you is the material they have to work with. I went into where that material comes from in Where AI Engines Go to Decide About Your Brand.
So the effort question reaches past Google's rankings. If what you publish is a thinner version of what's already out there, there's less reason for any engine to quote you, and I made the bigger version of that argument in The Click That Didn't Happen.
The fake author paragraph
Google's list of fabricated creator profiles includes AI-generated headshots, invented names and false credentials, and it says any form of deception makes a page untrustworthy to users and to Google's "automated quality systems."
A few years ago it got kind of common to publish under a persona, with a friendly name, a stock or generated photo, and a bio that says "content strategist" without saying for whom. Sometimes that was a stand-in for a real team, and sometimes it was a way to make a lot of AI copy look like it came from somebody. Either way, it now matches Google's written description of deception, and the quality rater guidelines already gave deceptive pages the lowest rating (Search Engine Journal).
Google doesn't say how its systems would detect a fake author, and I don't know either. I'd rather not find out the hard way. I've argued before, in The Coming Backlash, that the human behind the content is becoming the thing readers and engines both look for, and this is Google putting a version of that in its documentation.
What I'd do with your content program
Number one, find out where your effort is going. Pull your last 30 or so published pieces and ask, for each one, what's in it that wasn't online before you published it. That could be a data point, an opinion with a name on it, a tool, a photo of the real thing, or a comparison nobody else bothered to make. If the honest answer for most of them is "nothing," that's the problem to work on first, and AI has made it a lot cheaper to keep doing.
Two, fix your bylines. Every author on your site should be a real person who would recognize the piece. If your team uses AI heavily, Google's own "How" guidance asks creators to be open about it where readers would wonder how something was made, and I'd lean toward saying so plainly. (Generated people in your ads are a related issue with a legal side in New York, which I covered in What New York's AI Disclosure Law Means for AI Search Visibility.)
Three, spend your AI time on the high-effort side of Google's own list. The examples Google gave for high effort are original analysis and custom interactive tools, and AI is good at helping you build both. It's a lot less good at producing original thinking from nothing, which is the part you have to bring.
Then last, check what the engines say about you after you make changes, alongside where you rank. If you want a quick read on that, the free Quick Check will show you whether the AI engines mention you for the questions your buyers ask. There's a short self-check on the guidance below.
Self-check, 6 questions
Would your content pass Google's effort test?
Answer for the last ten or so pieces your team published. Nothing is stored or sent anywhere.
Questions based on Google Search Central, Creating helpful, reliable, people-first content (Oct 2026). Scoring is a rough guide by Jarred Smith, not a Google measure.
I'd love to hear what you're seeing, especially if your AI-heavy pages are doing fine and you think I'm wrong about where this goes. I write about how the engines pick brands in more depth in my book, Explainable, and a lot of it comes back to the same idea, which is that the engines can only recommend you for things somebody has written down.
Jarred Smith is the author of Explainable: Why AI Recommends Some Brands & Ignores Others, an Amazon bestseller on AEO, GEO, and SEO. He's a marketing leader with nearly 20 years of experience across healthcare, public media, retail, and environmental services. Find him at jarredsmith.com.