Changelog
- Updated
- Updated
People type “does AI content rank in Google” into a search box, Google’s answer is a plain yes with one condition attached, and the more useful question is what happens to the page after Google says yes. I run Content Cucumber, a writing company that uses AI in its workflow with a human over each post, so I have a stake in the answer, and I also have a Search Console account that shows how our own post on this subject performs, which is a less flattering read than Google’s policy pages.
Google Answers the Question in Its Own Words
Google’s February 2023 post on Search Central, written by Danny Sullivan, carries the subheading “Rewarding high-quality content, however it is produced,” and the sentence under it that matters to a marketer reads “Appropriate use of AI or automation is not against our guidelines.” The condition lives on the helpful content page, updated December 2025, where Google writes “If you use automation, including AI-generation, to produce content for the primary purpose of manipulating search rankings, that’s a violation of our spam policies,” and the spam policies page names the pattern scaled content abuse, with “Using generative AI tools or other similar tools to generate many pages without adding value for users” as its first example. The AI features page adds that “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary,” so a page needs to be indexed, eligible, and useful, and Google says it grades the page rather than the typist, though the same guidance asks whether the use of automation is self-evident to visitors. Content Cucumber’s 2023 post on how Google views AI content walked through that policy shift in 2023, so the yes has been settled since then, and the ranking question is the wrong one to keep asking.
What Search Console Shows for Our Own Post on This Question
That 2023 post, refreshed in July 2026, is the page Google serves when contentcucumber.com appears for the does-AI-content-rank cluster, and the Search Console read for September 2026 puts it at an average position of 69 across about 133 impressions, with “does ai content rank in google” accounting for 58 of those impressions and “does google rank ai content” for 38. Position 69 is page seven, so Google indexed the page, judged it eligible, and shows it to almost nobody, which is the measured answer to the question. When I saw that number I asked what I could change in the post to increase its awareness, and I put the question to AI, which is a great analytical tool for reading a page against a query, before I decided what a person would rewrite. A page that restates Google’s policy in the same words as two thousand other pages gets ranked and then gets buried, and Kelsey Knight, a commercial executive in consumer and B2B SaaS, named the reason in our reader study when she called AI “derivative by construction,” which describes the 2023 post as well as it describes a model.
The First Failure Mode Is Ranking Without Being Read
The first failure sits inside the yes. A page can rank and earn nothing, because AI Overviews and AI Mode answer the question on the results page and the click does not arrive, a shift I traced on Content Cucumber’s own traffic in the post on why broad search queries stopped sending clicks. Rishabh Chatterjee, co-founder and CTO at Passionfruit, described the other version of the same failure in our study, companies that “push out thousands of AI-generated pages” and find “It helps for two weeks and then everything starts tanking,” which is the scaled content abuse pattern Google describes. I have watched the same curve on my own site, a cricket project I started where the pages ranked and then fell off. What I took from it is that niche content works but humans have to be behind it. The pages in Chatterjee’s account ranked and then dropped, and what pulled them down was the absence of anything a reader or a ranking system could use, so the pages held a ranking for two weeks and no reader used them.
The Second Failure Mode Is Being Read and Recognized as Machine Written
The second failure happens after the click. At eTail Boston and over LinkedIn I asked marketers and executives one question, should companies use AI to create content, and why, and the answers, gathered in what hundreds of people said about companies using AI to create content and first reported in the post on readers being able to tell when a machine wrote it, came back with the word “slop” unprompted from more than one person. William Jowett of Bidnamic told me “I can really tell when it’s been procedurally generated by a computer,” and Julian V Taylor warned of “a world full of perfectly written content that nobody remembers.” The tells the respondents named are small and repeatable, headings with each first letter capitalized, opening paragraphs that restate the question, corporate phrasing with no person behind it, and a reader who spots them stops trusting the brand that published the page, and the brand pays that cost with the reader, where no ranking system measures it.
Why the Watermark Question Now Follows the Ranking Question
A third thing changed in the summer of 2026 that the 2023 guidance could not anticipate. Anthropic signed the EU Code of Practice on Transparency of AI-Generated Content in July 2026 alongside about 190 other signatories, the requirement took effect on August 2, 2026, and Google DeepMind had published its SynthID-Text watermarking method in Nature in 2024, so text a model wrote from scratch now carries a signal a platform or a buyer can check, which I walked through in the post on the AI-generated content paper trail. The watermark cannot tell a reader whether a page is any good, and Google’s ranking guidance has not changed because of it, but a buyer of content can now ask a vendor how a post was produced and expect a checkable answer, which folds the ranking question back into Google’s own “How” question about whether the use of automation is self-evident to visitors. Some buyers are already there, we have clients who insist that no AI is used on their work, and that instruction is easy to honor when the person who wrote the post can say so.
What a Human Reviewing Each Post Changes
Before Content Cucumber I started Content Basis in 2023, an AI-forward writing company, and the lesson I took from it, written up in the post on why human writers still win and how AI makes them better, was that “When everyone uses the same tools with the same defaults, every company starts sounding identical.” The moment I knew the AI-forward model would not hold was when social platforms started adding AI slop buttons. Clients wanted the speed, then found the drafts needed the same editing pass a human draft needs, and readers picked out the patterns unprompted. A person over the post changes three things a model does not do on its own, the origination of a point of view from something the writer did or saw, the removal of the patterns readers named as tells, and the judgment call on which claims need a source and which need cutting, and those three are the difference between a page Google indexes and a page a reader finishes.
How Content Cucumber Runs AI in Its Own Workflow
Content Cucumber uses AI in its workflow, and I am not going to claim otherwise, because the paper trail makes the claim checkable and because the tools earn their place. A writer originates the draft and the models do research, fact-checking, and SEO scoring, the same line Clay drew in the AI writing policy I wrote about copying, where the rule is about who originates the sentence rather than which tool is open on the desk. A brand-voice prompt counts the adverbs in a draft and shows me each one in a grid, a database of overused AI terms catches words like “delve” before a draft goes out, and RequestDesk checks each post against the stored brand voice, so the person is over the post at the first step and the last step, with the model in between. When content is going to be read by a human, that is where a human needs to step in. The pages that get cited in AI search, per our post on what gets content cited in AI search, carry specific facts and named sources, and a model can fetch those, but a person has to decide which ones are true.
The AEO audit from Content Cucumber measures which of your pages Google ranks, which the AI answers cite, and which nobody reads, before a word gets rewritten.
Let us help you get found in AI search results
AI Search Optimization →Keep Reading
Frequently Asked Questions
Yes, with one condition. Google's February 2023 Search Central post says appropriate use of AI or automation is not against its guidelines and that it rewards high-quality content however it is produced. The condition, restated on the helpful content page updated December 2025, is that using automation, including AI generation, to produce content for the primary purpose of manipulating search rankings violates Google's spam policies.
Scaled content abuse. Google's spam policies page lists using generative AI tools or other similar tools to generate many pages without adding value for users as the first example of the pattern.
Content Cucumber's 2023 post on how Google views AI content, refreshed in July 2026, sat at an average position of 69 across about 133 impressions for the does-AI-content-rank query cluster in the September 2026 Search Console read, which means Google indexed it and shows it to almost nobody.
The first is ranking without being read, where AI Overviews and AI Mode answer the query on the results page or thin pages rank briefly and then drop. The second is being read and recognized as machine written, where readers spot tells such as headings with each first letter capitalized, opening paragraphs that restate the question, and corporate phrasing with no person behind it, and stop trusting the brand.
Content Cucumber uses AI in its workflow with a human over each post. A writer originates the draft and the models handle research, fact-checking, and SEO scoring, a brand-voice prompt counts adverbs, a database of overused AI terms catches words like delve, and RequestDesk checks each post against the stored brand voice.
