Does Google penalise AI-written blog posts, and does it run a detector?

Before the contract, ask him which problem he is solving. "Google penalises AI" and "I do not want my brand voice sounding generic" are different goals and need different clauses. The first is a claim about a search engine. The second is a quality standard, and you could meet it with a review step instead of a detector score.

Pretty sure Google has to run an AI-text detector. Otherwise how would they enforce a policy against it? You cannot punish what you cannot identify, and they have the engineers to build one.

Google's published policy targets behaviour, not authorship. Its March 2024 update strengthened the scaled content abuse policy, covering content produced at scale to rank, "whether automation, humans or a combination are involved." Nothing I read describes a check for AI authorship, and your client's detector score is not a Google signal.

The key passage is in Google's announcement of the March 2024 core and spam updates. It says Google has long had a policy against using automation to generate low-quality or unoriginal content at scale to manipulate rankings, and that it was strengthened "to focus on this abusive behavior" because it is no longer always clear whether automation was involved. Its example is "pages that pretend to have answers to popular searches but fail to deliver helpful content." In other words, the test is value and scale, not who or what typed the words.

Google's "How Search works" page on spam adds how enforcement runs: automated systems catch most spam, a manual review team handles the rest, creators are alerted to manual actions, and they can submit a reconsideration request once fixed. It does say machine learning and large language models help its spam-fighting systems spot patterns at scale. That is spam detection across the web, a different task from labelling one post as AI-written, and the page does not claim it does the second.

Two caveats. I could not read Google's older February 2023 guidance on AI content in a usable form, and Google revises these policies, so check the current spam policies before quoting anything in a contract. And a flag from a third-party detector on your own writing is a tool quirk, not a ranking event. For the clause, I would promise a process (human drafting, source checks, a record of drafts) rather than a detector outcome. See how accurate AI text detectors are.

Sources:

• Google, "New ways we're tackling spammy, low-quality content on Search"
• Google, "Detecting spam" (How Search Works)