Definitions

• Let $A$ be the set of all AI-generated content.

• Let $B$ be the set of all content identified as AI-generated and removed by the moderators of the SCP wiki, based on human intuition, account age, and other observable tells.

• $A \cap B$ is the set of AI-generated content successfully identified by the moderators.

• Let $C$ be the set of all content on the SCP wiki.

Assumption

Assume that the detection mechanism, which includes human intuition and factors like account age and other tells, is not 100% perfect. Given the human margin of error on any task, this is all but a certainty. It has been established, then, that our Turing test is flawed, this means it cannot reliably identify all AI-generated content as such. Therefore, $A \cap B \neq A$.

Proof

1. Assume AI-generated content exists on the SCP wiki: Assume there exists some AI-generated content $x$ such that $x \in A \cap C$. This means $x$ is both AI-generated and on the SCP wiki.

2. Imperfection in detection: Given that $A \cap B \neq A$, there are some elements in $A$ that are not in $B$. Let’s denote this set difference as $D = A \setminus B$, where $D$ is non-empty.

3. Existence of undetected AI content on the SCP wiki: Since $D$ is non-empty, and $D \subseteq A$, and we assume $x \in A \cap C$, it’s possible that $x \in D \cap C$, in a statistically significant number of cases. It then follows that $x$ is AI-generated content that is both on the SCP wiki and has not been identified as AI-generated despite having been scrutinized by moderators.

Conclusion

Since $D \cap C$ is non-empty, there necessarily exists some AI-generated content on the SCP wiki that has not been identified as such by the screening process, despite human intuition and other detection factors. Thus, some AI-generated content that passes the community’s enhanced Turing test is shown to exist.