Yes. AI-generated websites rank on Google every day, and Google has said flat out that it doesn’t care how a page got made. Google ranks pages that help the person searching, no matter who or what typed them, and it buries thin, mass-produced pages by the same rule. The sites that got wrecked in March 2024 weren’t wrecked for using AI. They were wrecked because nobody would have missed them.
We sit on both sides of this question. We give away the prompts people use to vibe code their own sites, and we build and rescue AI-built sites for money, so we watch what Google actually does to these sites, not just what it says. Here’s the official policy, the data, and the exact things to change in AI output before you hit publish.
Does Google penalize AI content? The official answer
Google put its position in writing back in February 2023 and hasn’t walked it back since. Google’s official guidance on AI-generated content says Search rewards high-quality content “however it is produced,” and that appropriate use of AI or automation doesn’t break any guideline. The company frames this as nothing new: it has always judged the quality of the content, not the production method behind it.
People quote that first half and skip the second. The same document says that using automation, AI included, to generate content whose primary purpose is manipulating search rankings violates Google’s spam policies. So the line was never human versus machine. The line is helpful versus manipulative, and it runs straight through both camps.
Underneath the policy sits E-E-A-T: experience, expertise, authoritativeness, trust. Those are qualities of a page and the business behind it, not of the typing. A services page drafted by Claude and edited by a plumber with 20 years in the trade can show all four. A page hand-written by a content mill can show none. The tool doesn’t earn rankings or lose them. The page does.
Plenty of people still assume machine-written equals spam, an idea left over from the article-spinner era. Google’s systems moved past that years ago. What replaced it is blunter and harder to game: does anyone actually need this page?
For anyone vibe coding their own site, that’s a little freeing. You’re not sneaking past a censor. Google is fine with your workflow. What it grades is whether the finished page reads like it came from someone who knows the business, charges real prices, and answers the phone. AI gets you 80 percent of the way there in an afternoon. The 20 percent it can’t do is the part that ranks.
The March 2024 lesson: scale with nothing inside gets buried
March 5, 2024 is the day the policy grew teeth. Google shipped a core update it said was more complex than its usual ones, folded the helpful content system into its core ranking systems, and introduced a spam policy called scaled content abuse: publishing lots of pages primarily to rank rather than to help, “no matter how it’s created,” per Google’s announcement. That last phrase is doing the work. The policy is method-agnostic on purpose. Ten thousand junk pages from an offshore article farm violate it exactly the same way ten thousand junk pages from a language model do.
Google predicted the changes would cut low-quality, unoriginal content in search results by 40 percent. When the rollout wrapped on April 19, the company said the real figure came in at 45 percent, per Search Engine Land.
The manual actions were the loud part. Within the first week, one tracker following 49,345 sites counted 837 knocked entirely out of Google’s index, sites that had been pulling a combined 20.7 million organic visits and roughly $446,000 in display ad revenue per month, Search Engine Journal reported. Originality.ai ran those deindexed sites through its detector and found that every single one showed signs of AI-generated content, and half of them were 90 to 100 percent AI posts.
Put those two findings side by side and the lesson writes itself. Google didn’t hunt AI. It hunted volume with nothing inside, and pure-AI-at-scale sites happened to be exactly that. If the plan is 400 lookalike pages by Sunday night, the plan is the problem, not the model that typed them. Volume was never the strategy. It was the tell.
So do AI-generated websites rank on Google right now?
Almost two years past that update, the data says AI-assisted pages aren’t just surviving. They’re most of the leaderboard. In mid-2025, Ahrefs analyzed 600,000 pages ranking in the top 20 for 100,000 random keywords. Only 13.5 percent were purely human-written. 81.9 percent mixed human and AI work. 4.6 percent were entirely AI-generated.
Here’s the number that settles arguments: the correlation between how much AI a page contained and where it ranked was 0.011. That’s statistical zero. No reward for using AI. No penalty for using AI. Google grades the page, not the author.
One detail from that study belongs on a sticky note over your desk. Fully AI-generated pages made the top 20 constantly but rarely took the number one spot. As of 2026, the gap between “ranks somewhere on page one” and “ranks first” is still mostly filled by a human who edited the thing, added something real, and made it theirs.
For a small business, that’s genuinely good news. You don’t need to out-write the internet. You need one page per topic that’s more specific and more true than the AI sludge it competes with, and the sludge sets a low bar. Most of your competitors ran the prompt, published whatever came out, and moved on. Editing is the entire moat now.
How AI content affects Google ranking: why some sites sink
Rescue work gives us a front-row seat on the flop side. When a broken AI-built site lands on our bench (enough of them do that cleanup is a service line now), the ranking problem is almost never “Google smelled the AI.” The pattern we keep seeing is three misses stacked together: copy that could belong to any business in America, no proof a real company exists, and technical basics the model skipped while it was busy writing hover animations.
| What Google weighs | Sites that rank | Sites that sink |
|---|---|---|
| The copy | Prices, process, neighborhood names, opinions only the owner holds | “Premier solutions tailored to your needs,” forty different ways |
| Business signals | Street address, phone number, owner’s name, real photos, reviews | Stock photos and a contact form floating in space |
| Technical basics | One h1 per page, descriptive title tags, a sitemap, fast load times | Five h1s, every title reading “Home”, no sitemap anywhere |
| Publishing pattern | A page ships when there’s something to say | 300 near-identical pages generated over a weekend |
That middle row is the cheapest fix on the board and the one owners skip most. Google has spent two decades learning to tell the difference between a website and a business that has a website. An about page with your actual face, a footer address that matches your Google Business Profile, a phone number someone picks up, a few reviews with your name in them. None of that is SEO wizardry. All of it tells Google, and every visitor, that a human stands behind the page, which is the one trust signal mass-generated sites can’t produce at scale.
Notice what’s missing from that table: which model wrote the first draft. Google can’t see your git history and wouldn’t care if it could. Vibe-coded static sites actually start with an edge on the technical row, because lean hand-rolled HTML without page-builder bloat tends to load fast, and speed is one of the few ranking inputs you control completely. The full checklist, from title tags to schema to Search Console, lives in our guide to SEO for vibe-coded websites.
Fix these things in AI content before you publish
Before any AI draft goes live, run it through the list below. Figure 20 to 30 minutes per page. That half hour is the difference between joining the 81.9 percent of ranking pages that blend AI with human judgment and joining the pile Google cut by 45 percent. None of it needs an SEO certification. It needs honesty and a delete key.
- Rewrite the first paragraph from scratch. AI intros summarize the topic; yours should answer the question or make the offer. If a stranger couldn’t quote your opening sentence and learn something, redo it.
- Swap every generic claim for a number you own. “Competitive pricing” becomes “$95 service call, credited if you hire us.” “Years of experience” becomes “since 2011.” A model can’t know these things, which is exactly why they read as real.
- Add one thing per page that only you could write. The job that went sideways, the brand of pipe you refuse to install, the street you always park on. Experience is the first E in E-E-A-T and the only one a model can’t fake.
- Fact-check every number, name, and claim the model produced. AI states wrong things with total confidence, and one invented stat costs you the one reader who knew better. That reader was probably your best lead.
- Delete the throat-clearing. Any paragraph announcing that your industry matters, or that the world moves fast, gets cut whole. Filler is the strongest AI smell there is, and readers bounce off it before Google ever has to weigh in.
- Write your own title tags and meta descriptions. One page, one search intent, one honest title built from the words your customer actually types.
- Replace placeholder images and alt text. A phone photo of your actual storefront beats a stock handshake in every way that matters, image search included.
Our 40 free prompts handle a chunk of this at the source. Each one runs around 12,000 characters and bakes the skeleton in: semantic headings, a single h1, real title tags and meta descriptions, tested on ChatGPT, Claude, and Gemini. What no prompt can do is know your prices or your stories. The structure comes out clean. The truth is still your job.
Where this leaves you (and where we come in)
Everything above now pulls double duty, because the qualities that rank on Google are the same ones that get pages quoted by ChatGPT, Perplexity, and AI Overviews: specific numbers, quotable standalone sentences, evidence that a real business stands behind the page. We broke down that side of things in our guide to getting cited by AI search, and the overlap with this post is not a coincidence.
If you’d rather hand the whole thing off, that’s the other half of our shop. The SEO and AI-search optimization service starts with a $1,200 one-time audit that tells you exactly why your site sits where it sits, on Google and inside the answer engines, and ongoing work runs from $950 a month. No discovery-call funnel, no 40-page PDF of jargon. You ask a question, you get an answer within a day.
Questions we get about AI content and Google
Does Google detect AI content?
Google has classifiers, but detection isn’t really the mechanism, because its written policy judges quality and purpose rather than authorship. The March 2024 deindexing wave shows what actually trips the wire: hundreds of same-shaped pages, no named author, no lived experience, published in bursts. Sites that behaved like publishers kept ranking. Sites that behaved like printers got removed.
Will Google penalize my site for using ChatGPT to write pages?
Not for using it. Google’s guidance explicitly allows AI-assisted content, and the scaled content abuse policy targets volume produced primarily to manipulate rankings. A business owner editing AI drafts for a 10-page site isn’t anywhere near that line. An 800-page programmatic blast with zero editing is standing right on it.
Do fully AI-written pages ever hit number one?
Rarely. In Ahrefs’ 600,000-page study from mid-2025, entirely AI-generated pages appeared throughout the top 20 but seldom held the top spot, while pages mixing AI and human work dominated the results. The draft gets you into the game. The editing is what wins position one.
Can Google tell my site was vibe-coded?
No, and it wouldn’t matter if it could. Google reads the HTML your server sends, not the process that produced it, and clean semantic markup is clean semantic markup. If anything, vibe-coded static sites hold an advantage here, because skipping page-builder bloat means faster loads, which feeds straight into Core Web Vitals.
How long before a new AI-built site starts ranking?
Same timeline as any new site: a few weeks to get indexed, a few months to rank for anything competitive, depending on your market. There’s no AI-specific waiting room. What shortens the wait is boring and proven: a Google Business Profile, consistent name, address, and phone everywhere, a handful of real local links, and pages that answer questions people in your town actually type.