Skip to content
DarkCore Grove
Book a demo

AI content on affiliate sites: what Google allows

Google does not ban AI-written articles; it bans pages made only to rank. What that means for affiliate sites and the checks an AI article should pass.

Published
3 min read
3 min read
Русский
Українська
A mechanical writing automaton at a desk beside a tall inspection gauge with dials
On this page
  1. What Google actually said
  2. Where AI articles go wrong on offer sites
  3. The checks an AI article should pass
  4. Authors and images
  5. How Grove does it

Google does not penalise an article for being written with AI. Its guidance of 8 February 2023 says that appropriate use of AI or automation is not against its guidelines, and that it rewards helpful, original content however it is made. What Google treats as spam is automation used mainly to manipulate rankings. For an affiliate site the question is therefore not “AI or not” but “would this article be worth publishing if a person had written it”.

What Google actually said

The February 2023 post makes two points worth keeping:

  • How it is made does not matter, what it is does. Google judges quality, with expertise, experience, authoritativeness and trustworthiness (E-E-A-T) as the frame.
  • Automation to game rankings is spam. That was true before AI and stays true with it.

The spam policies add the concrete line: using generative AI to create many pages that add no value for users is an example of scaled content abuse. See scaled content abuse for affiliate sites.

A stack of brass sieves on a wooden stand, each finer than the one above

Where AI articles go wrong on offer sites

Generated text fails in predictable ways, and on affiliate sites each failure has a cost:

  • Forbidden claims. A health or finance offer usually lists what may not be promised. A model that never saw that list will promise it.
  • Nothing new. A model asked for “the best X” tends to rewrite the top 10, which is the definition of no added value.
  • Overlap with your own sites. Two sites of one portfolio get articles on the same search and end up competing. See keyword cannibalization.
  • Facts without sources. Numbers and health statements with nothing behind them.
  • Invented authors. A made-up expert with a stock photo is the fastest way to lose a reader’s trust.
  • Broken technical SEO. Bad slugs, missing structured data, links to the landing instead of through tracking.
A botanical press holding a single pressed fern

The checks an AI article should pass

Treat generation as the first draft and the checks as the editor. Before an article goes out, it should pass at least these:

  1. Claims. No wording the offer forbids, no paraphrases of it.
  2. New value. Something the current top 10 does not already say.
  3. Cannibalization. Not the same as another article of your portfolio.
  4. Similarity. Not a near copy of a ranking page.
  5. Facts with a source. Flag what has none; let a person decide on health and numbers.
  6. Structure and language. Headings, readability, the right language for the site.
  7. Technical SEO. Slug, structured data, links, image metadata.

The first four should block publishing. The last three can warn or fix themselves.

A small balance weighing a single seed

Authors and images

Sign articles with a real author when the site has one, and with an honest editorial line when it does not. Do not pass an AI image off as a photo. Both come down to trust, which is the T in E-E-A-T.

How Grove does it

In DarkCore Grove you pick topics and press Generate, for one article or a batch. Grove writes the text, makes the images and runs seven Gates, exactly the list above. Claims, new value, cannibalization and similarity block publishing. On Autopilot an article with health or number findings still waits for a person. Articles are signed by a real author or the Editorial team, and Grove never invents people.

Keep reading

All articles