Get a content audit →
The AI-content question

Google does not penalise AI. It penalises lazy.

The watermark, the detectors, the plagiarism scare: all of it dissolves into the one thing Google actually grades. Not the tool that typed the draft. The research behind it, the writing, and the human accountable for it.

We will show you, in Google's own words, what gets penalised and what gets ranked. 30 minutes, no pitch.
Straight from the source

Do not take our word for it. Take Google's.

Every fear about AI content runs into the same wall: Google, and the people who watch it closest, keep saying the tool is not the point. Here it is, in their words, with the links.

Our focus on the quality of content, rather than how content is produced, is a useful guide that has helped us deliver reliable, high quality results to users for years.
Google Search Central · Danny Sullivan and the Search Quality team, official guidance on AI-generated content developers.google.com
A thousand pages on our site that are us versus every possible competitor and the pages are all 80% similar: that is scaled content abuse.
Lily Ray · Amsive, on what actually gets penalised ppc.land
Just rewriting AI content by a human won't change that, it won't make it authentic.
John Mueller · Google, via Search Engine Roundtable seroundtable.com
If your goal when using AI is to find creative ways to add further context and information to help your readers, that is great. If you are using AI to find ways to get more content indexed by Google, then that is not great.
Barry Schwartz · Search Engine Land searchengineland.com
Scaled content abuse, especially via AI-generated content, has surged recently.
Glenn Gabe · GSQI, on the Aug 2026 spam update gsqi.com
Fear 1 · the penalty

Google does not penalise AI. It penalises scaled slop.

The sites that got hit did not get hit for using AI. They got hit for what they published with it: thin, templated, near-identical pages built to flood the index. The line is quality, and it always has been.

What gets penalised

Scaled slop

Near-identical pages built to blanket every keyword: the textbook case of scaled content abuse.

Templated pages churned out to blanket every keyword
Thin rewrites of whatever already ranks, nothing added
Raw model output published with no human in the loop
Made to game the index, not to help a reader
What gets ranked

Real work

The same bar Google has always applied, whoever holds the pen.

Original research and data a model could not invent
A genuine point of view on the topic
Primary sources, cited, and real first-hand depth
A human accountable for the final piece

Same web, same bar. The tool is not the problem. Having nothing unique to say is.

Fear 2 · the watermark and the detectors

The watermark is real. The penalty is not.

In 2026 Anthropic began watermarking Claude's text to comply with the EU AI Act. It is a transparency label, not a ranking signal, and it does far less than the scare suggests.

2 yrs
Google's own Gemini has carried a text watermark, SynthID, for about two years, and has never used it as a ranking signal. A watermark is for disclosure, not for punishing pages.
Seer Interactive analysis, 2026 · seerinteractive.com
What the Claude watermark actually does

It shows Claude was involved. That is all.

Added in 2026 to comply with the EU AI Act, not any Google update.

By Anthropic's own account it only estimates the likelihood Claude was involved. It cannot tell "Claude wrote this" from "Claude heavily edited this," and it is not a ranking signal either way.

Google does not use it. AI detectors, meanwhile, are notoriously unreliable, and Google does not rank on them either.

Anthropic, "How Claude's text watermarking works," 2026 · anthropic.com
Fear 3 · plagiarism

AI content is only plagiarism if the process is lazy.

A model left to itself paraphrases what already ranks. That is a process failure, not a property of the tool. Three habits turn a draft into original work.

Original
We start from first-hand research and data, the things a model cannot pull from the training set, so the piece has something genuinely new in it.
Research before writing
Yours
A point of view, examples and a brand voice a human supplies. The draft is a starting point, not the published article.
A human's craft and judgement
Checked
Every piece is cited to its sources, edited by a senior human accountable for it, and run through plagiarism and quality checks before it goes live.
Sourced, edited, verified

Original research plus a human editor is not just how you avoid plagiarism. It is how you write the thing everyone else ends up quoting.

The real answer

It was never AI versus human. It is process.

Every objection on this page lands in the same place. What decides your rankings is what you research, how you write it, and who is accountable for it. That is a process, and it is the one we run.

01

Research

Primary sources, original data and a real angle, gathered before a word is drafted. This is what makes the content worth ranking, and very hard to plagiarise.

The Research desk
02

Writing

A defined brand voice and a human's craft shape the draft into something with a point of view. AI drafts. It does not decide, and it does not publish.

The Writing desk
03

Review

A senior editor is accountable for every piece, and it clears plagiarism, fact and quality checks before it ships. A human owns the final word, literally.

The Quality desk
04

Measure

We track whether it actually ranks and gets cited, and refine. The test is not "did AI write it." It is "did it earn the position."

The scoreboard
This is the Editorial Engine. Four desks, one accountable human, on every piece.See the Editorial Engine →

The compounding system it feeds: the Compounding Engine.

Proof

Content built this way does not get penalised. It holds through a core update.

The receipts · two clients on the desk

Research-led content, ranked and rewarded by Google.

A wealth-management SaaS held its number-one ranking straight through a Google core update that reshuffled the whole category, the clearest signal there is that Google rewarded the work rather than penalising the method. And Varniya, from a weak domain, reached #1 for its head term over national media and is now cited in Google's AI Overview.

Clients on the desk, named under NDA in the room. Every piece: original research, senior-edited, sourced. See the case studies: Wealth-Management Platform → · Varniya →
Held #1
through a Google core update, whole category reshuffled
33
qualified leads a month from organic, through a core update
+275%
Varniya organic traffic in about four months
Cited
Varniya named as a source in Google's AI Overview
Common questions

The objections, head on.

Does Google penalise AI-generated content?

No. Google's published guidance is explicit that it rewards high-quality content however it is produced, and that appropriately using AI is not against its guidelines. What it penalises is content made primarily to manipulate rankings, which its spam policies call scaled content abuse. In Google's own words, the focus is on the quality of content, not how it is produced. The tool is not the risk. Thin, unoriginal, low-value content is.

Will the Claude watermark or an AI detector get me penalised?

There is no evidence Google uses AI detection or watermarks as a ranking signal, and Google's own Gemini has carried a text watermark for about two years without it affecting rankings. Anthropic added a watermark to Claude in 2026 to comply with the EU AI Act. By Anthropic's own account it can only estimate the likelihood Claude was involved, it cannot tell whether Claude wrote a piece or heavily edited it, and it is not a ranking signal either way. AI-text detectors are unreliable and are not a Google ranking factor. Quality is.

Isn't AI content just plagiarism?

Only if the process is lazy. Plagiarism and duplication come from publishing raw model output that paraphrases what already ranks. A real editorial process prevents it: original research and first-hand data, a point of view the model cannot invent, proper citation of sources, and human editing and plagiarism checks before anything is published. Done that way, the output is original work that happens to have been drafted with a tool, not a copy of someone else's.

So what actually decides whether content ranks?

Process. What you research, the original data and point of view you bring, the writing craft and brand voice, and the human who is accountable for the final piece. Google grades that, not the software that typed the first draft. This is exactly why we run a research-led editorial process rather than publishing raw AI output.

Do you disclose that AI was used, and is the work checked?

Every piece is built on original research, edited by a senior human who is accountable for it, and run through plagiarism and quality checks before it goes live. AI is a drafting tool inside that process, not the process. We are happy to walk you through exactly how a piece is made, desk by desk.

Can you guarantee I will never get a Google penalty?

No honest agency can guarantee rankings or immunity from an update. What we can do is remove the reasons sites get hit: no scaled or templated pages, no thin or duplicated content, original research and a human accountable for every piece. The proof it works is content built this way that has held its number-one ranking straight through a Google core update.

Stop asking if AI is safe. Ask if the work is good.

Book a free 30-minute content audit. We will show you, in Google's own words, what gets penalised, and show you the process that gets content ranked and cited instead. No pitch.

See case studies Book Strategy Session →