Friends who run businesses ask me the same thing: "Can I use ChatGPT to write my blog without Google punishing me?" Yes, you can. Google has said since February 2023 that there is no policy against AI content, and its October 2026 guidance confirms that generative AI is a useful tool. What gets punished is using automation to manipulate rankings.
What scaled content abuse really means
Google's spam policies describe scaled content abuse as generating many pages primarily to manipulate rankings and not to help users, however they are made. There is no page count that makes something "scaled". It is judged by purpose and by the lack of value. Bing says something similar: large-scale content generated without oversight, quality control or editorial review often lacks usefulness, accuracy and originality, and may be excluded from its index.
My five-step workflow
- Proprietary input first: an interview, notes from a real project, your own numbers. AI never starts from a blank prompt.
- AI researches and drafts: structure, first version, alternatives for titles. This is where it saves hours.
- A named person fact-checks everything against its source, including title, meta description, structured data and alt text.
- That person adds what only they know: the decision, the mistake, the result.
- A short disclosure tells readers how the content was made, and then it is published.
Why the review is not optional: generative models predict the most likely next words, they do not retrieve facts. That is why Google asks for a manual fact-check before publishing.
How skills make the process stick
The hard part isn't knowing the rules, it's applying them every time. That's why I build Claude skills for teams: one skill holds the brand voice, another holds the review checklist, another checks that nothing is published without sources. Anyone on the team runs the same process in one step, and the quality stops depending on who wrote the prompt.
A note on the quality rater guidelines
Google points to sections 4.6.5 and 4.6.6 of its Search Quality Rater Guidelines, about scaled content and main content made with little effort or originality. Raters don't change rankings directly, they help Google check its systems, but reading those sections is a great way to see your content the way Google does.
Quality starts with the brief
Every piece I commission starts from something only the company has: its results, its cases, the mistakes it has seen, the real questions its customers ask, its own measurements. If there's no such input, I don't commission the piece, because the model would only reshuffle what's already online. A good brief fits on one page: who the reader is and what they are trying to solve, the proprietary input, the sources the model may use, and what it must never claim.
The instructions I give the model
- Do not invent figures, dates, names, studies or quotes. If a fact is not in the brief or the sources, leave it out.
- Write for the reader: normal paragraphs and headings that organize, no special formatting for machines.
- Put the proprietary input early, in the section where it belongs.
- Propose a title, a meta description and alt texts, marked as pending human review.
- If the topic already exists on the site with the same intent, say so and do not write a new piece.
Publish at a human pace
The practice I follow is that volume is set by how much a person can review, not by how much the model can write. I start with what one person can read in full, check that new pages are indexed ten to fifteen days later, and only increase the pace if indexing and performance keep up. If indexing of new content starts to drop, I slow down.
Two questions instead of a page count
Since Google doesn't give a number for "at scale", I use two questions. Does someone read every piece before it's published? Would each piece exist even if search engines didn't? If either answer is no, the volume is too high, whatever the number is.
Other spam policies AI can trigger
- Site reputation abuse: publishing third-party content on your domain so it ranks thanks to your site, not its own value.
- Expired domain abuse: buying an old domain to host low-value content and exploit its history.
- Hidden text or instructions aimed at AI models. Google treats attempts to manipulate its AI responses as spam.
- Scraping: copying or lightly rewriting other sites without adding value.
Images count too
Alt text is reviewed like the rest of the content, so write it for people who can't see the image. Images made with AI should carry the IPTC TrainedAlgorithmicMedia metadata. And remember that Google doesn't index images set as CSS backgrounds, so any image that matters belongs in the HTML.
My checklist before anything goes live
- It contains a fact, an experience or a judgment that is not on other sites.
- A named person has read the full text and the metadata and checked the facts.
- Title, meta description, structured data and alt text match what is visible on the page.
- There is a note about how it was made where a reader would expect it.
- No other page on the site covers the same intent.
- The page can be indexed and shown with a snippet, and the site is included in AI features.
- It is logged with a date to check its indexing ten to fifteen days later.
What Bing adds
Bing's guidelines point the same way. Large-scale content generated without oversight, quality control or editorial review often lacks usefulness, accuracy and originality, and may be left out of its index. Bing also turned keyword stuffing into keyword stuffing and artificially engineered language, which includes text designed only to trigger AI citations. The safe path is the same everywhere: write naturally, for a person.
If you want AI to multiply your team's output without putting your site at risk, let's design the workflow together.
Sources
- Google Search's guidance on using generative AI content on your website · Google Search Central, updated October 1, 2026
- Spam policies for Google web search · Google Search Central, updated August 28, 2026
- Google Search's guidance about AI-generated content · Google Search Central blog, February 2023
- Bing Webmaster Guidelines · Bing, updated February 27, 2026
How I made this article: I researched and drafted it with the help of Claude, then checked every fact myself against the official documentation listed in the sources.