What does Google actually say?
Less than the rumours. In February 2023 Google Search Central published its guidance on AI-generated content, and it has not changed in substance: “appropriate use of AI or automation is not against our guidelines”, while using automation, including AI, to generate content primarily to manipulate rankings is a violation of its spam policies. Google says its systems reward quality “regardless of how content is produced”, and asks publishers to focus on accuracy, quality and relevance, and to consider telling readers how AI was used 1.
So the question in the headline is the wrong question. Google does not rank content by its author’s species. It ranks content people find useful, and demotes content made to game it.
What does rank?
The same things that ranked before the tools existed, plus one new one. A page that answers the query directly, near the top. Headings that are the questions people ask. Facts with a source somebody can check. A named author with a reason to know. A date, and a revision date that moves. A page that loads fast on a phone: Google’s own research found 53% of mobile visits abandoned above three seconds 2. The new one is being cited: AI search answers lift passages, and they lift from pages structured to be lifted.
| Signal | Made by a person | Made with AI | What decides |
|---|---|---|---|
| Answers the query | Sometimes | Sometimes | Editing |
| Cites real sources | Often not | Often invented | Checking every link |
| Named author, dated | Usually | Usually not | Publishing discipline |
| Loads fast on a phone | Varies | Varies | The site, not the copy |
| Made to manipulate rankings | Spam | Spam | Intent, not tool 1 |
What gets a page cited by AI search?
Structure that a machine can lift. A short answer at the top of the page. Question-form headings with the answer in the paragraph beneath. A table, because a row can be quoted where a paragraph cannot. Numbers that carry a source with a link, because a fabricated statistic is now a checkable one. This article is built that way on purpose, and so is the rest of this blog since its rewrite.
Where does AI content go wrong?
Invented facts and invented sources. A model asked for statistics will produce plausible ones and plausible reports to attribute them to. Published, those become a claim under your name that anyone can falsify in a minute, and a search engine or an assistant that catches one will trust the rest of the site less. The rule that survives is simple: every number has a link, and every link has been opened. This blog once carried quotes attributed to consultancies that never said them; they are gone, and the guard that keeps them out is part of the build.
What should a South African business do?
Use the tools to draft and a person to check, cite and sign. Put the answer first and the evidence beside it. Make the page fast. Then leave the question of what made the words to Google, which has said it does not care.
Questions people ask
Do I have to label AI-written content?
Google does not require a label; it suggests telling readers how AI was used where that helps them. Accuracy and a named author matter more than a disclaimer.
Will AI content get my site penalised?
Not for being AI-made. Content made primarily to manipulate rankings is spam whatever made it; useful content is not, whatever made it 1.
Does DoubleDown AI write blog content?
Our website design product includes the copy, structure and search markup for the site itself; ongoing content is scoped separately under Custom AI.