Chapter 1 of 5 · AI and SEO — Using AI Well, and Understanding AI Search

What Google thinks of AI-written content

Writing with AI is allowed, and it neither helps nor hurts by itself. Google looks at purpose and value. This chapter covers the scaled content abuse policy, the Who, How and Why questions, and telling readers that AI was used.

This series follows SEO for Thai Websites and covers the two questions site owners ask most right now: how far AI can help with a website, and how AI search such as AI Overviews or ChatGPT chooses which sites to show. Every point cites Google’s or each vendor’s own documentation as checked on 25 September 2026, and several lessons come from this site, which drafts its articles with AI.

1.1 Google does not forbid AI

The first worry is usually “will AI-written pages be demoted?” Google’s 2023 post answers plainly:

“Appropriate use of AI or automation is not against our guidelines.”
“Using AI doesn’t give content any special gains. It’s just content.”

But using automation, including AI, to generate content with the primary purpose of manipulating rankings violates the spam policies. If you see AI as “an inexpensive, easy way to game search engine rankings”, Google’s answer is no.

In short, Google does not ask who or what wrote a page but why it was written and whether it helps readers.

Google’s current documentation adds that generative AI is particularly useful for researching a topic and adding structure to original content.

1.2 Scaled content abuse: the trap AI makes easy

The spam policy most relevant to AI is scaled content abuse. Google defines it as many pages generated for the primary purpose of manipulating search rankings and not helping users, “no matter how it’s created”.

Google’s examples:

This violates the policy	Why
Using generative AI to create many pages without adding value for users	More pages, nothing more for readers
Scraping other sites and rewording or machine-translating them	Someone else’s content with different words
Stitching content from several pages together without adding anything	Nothing new
Spreading such pages across several sites to hide the scale	Deliberately evading detection
Keyword-stuffed pages that make no sense	Written for machines, not people

In 2026 Google made two points explicit:

the spam policies also cover attempts to manipulate generative AI responses in Google Search (clarified in May 2026);
its AI optimization guide says making a separate page for every variation of a query, primarily to manipulate rankings or AI responses, counts as scaled content abuse.

If a site already hosts such pages, Google says to exclude them from Search, with noindex or by removing them.

1.3 Who, How, Why: three questions Google asks you to ask

Google’s page on helpful content suggests three questions:

Question	What Google recommends
Who created it	Accurate author bylines where readers would expect them
How it was created	If automation or AI substantially generated the content, consider telling readers where they would expect it: a disclosure, how it was used, and why
Why it was created	Google calls “Why” perhaps the most important question; the answer should be “to help readers”, not “to attract search visits”

Two common mistakes:

Naming AI as the author. Google says this is “probably not the best way” to follow its byline advice.
Not saying AI was used at all on content where readers might wonder how it was made; Google finds disclosures useful there.

Also worth knowing: Google’s list of warning signs for “search engine-first” content includes “Are you using extensive automation to produce content on many topics?“

1.4 Accuracy does not stop at the article body

Google asks for accuracy, quality and relevance, especially when content is generated automatically, and that covers:

page titles and meta descriptions;
structured data;
image alt text.

Shops using Google Merchant Center have specific rules: AI-generated images must carry IPTC metadata marking them as algorithmically created, and AI-written product titles and descriptions must be labelled separately.

1.5 A lesson from KoishiAI: say you use AI, and don’t expect volume to be the answer

This site drafts articles with AI from a fact sheet with sources. We say so in our llms.txt file and in each article’s structured data; chapter 3 describes the system.

The lesson: publishing more easily did not make Google read more. Our Search Console crawl stats showed almost all of Googlebot’s requests re-reading pages it already knew, with hardly any spent on new ones. Most news added after a quiet spell had never been fetched, while the tool pages built in the same period were not in that group.

We do not know the exact reason. Google says crawl demand depends on a site’s size, update frequency, page quality, and relevance compared with other sites. Our site had gone quiet for a while, which bears on update frequency, and the first series found some internal links leading the crawler to the wrong places. Most of our tutorial pages, which are not commodity news, had not been fetched either, so we cannot pin it on the type of content. What we are sure of: AI makes writing fast, but more pages do not by themselves make Google more interested in a site.

1.6 After this chapter

Go through pages written with AI help and ask of each: “if Google did not exist, would this page still help a reader?”
Where the answer is no, add real value, or use noindex, or remove the page.
Put the real author or responsible team in the byline, not an AI.
Write a short note, somewhere readers can find it, on how the site uses AI.
Check that AI-written titles, descriptions and alt text match the actual content.

The next chapter looks at where AI genuinely helps with SEO, and where it confidently makes things up.