</> aicodedetectAI Code Humanizer

Codex Code Detector

Paste a snippet to check whether it looks like it was written by OpenAI Codex or by another AI assistant. Free, instant, no sign-up.

0 / 8,000 characters

What Codex is today

The name Codex has meant two things. In 2021 it was the OpenAI model behind the first version of GitHub Copilot; that model has since been retired. Today Codex is OpenAI's coding agent. It runs as a command-line tool on a developer's machine, inside editors such as VS Code, and as a cloud service that works on a copy of a repository and proposes a pull request. In every form it reads the project, edits files and can run tests, so its output looks more like a set of changes than a single pasted answer.

Check the commit history, but expect gaps

Codex is less consistent about signing its work than some other agents. The command-line tool has a commit_attribution setting that adds a Co-authored-by trailer naming Codex, and setting it to an empty value turns the trailer off. Public accounts disagree on whether it is on by default, and users who studied Codex-linked commits have reported finding no marker at all. So a Codex trailer is a direct statement, but most Codex-assisted code will not have one. An AGENTS.md file in the repository shows the project gives instructions to coding agents such as Codex, though not which lines they wrote.

Signals in Codex-written code

Without a trailer, you are back to reading the code. The detector looks for the habits AI assistants share:

  • Tests written alongside the change. Codex is often asked to run tests, and its changes tend to arrive with new or updated tests in the same style.
  • Edge cases nobody asked for. Empty inputs, missing keys and invalid types handled in code that was meant to cover one case.
  • Comments written for a stranger. Explanations of why a step is needed, aimed at a reader who has never seen the problem.
  • A finished shape. No dead code, no debug output and no half-done ideas, which is unusual for code someone worked through by hand.

Short snippets give the detector very little to read, so expect a less confident result for anything under a few dozen lines.

Read more

Frequently asked questions

Can this tool prove that Codex wrote the code?

No. Codex does not watermark the code it writes, so any detector reads style and gives a probability with reasons. Treat a high score as a reason to look closer, not as proof.

Can it tell Codex apart from ChatGPT, Claude or Copilot?

Not reliably, and we do not claim to. Codex and ChatGPT are built on related OpenAI models, and all the major assistants share most of their coding habits. A high score means the code looks AI-written in general.

Does Codex sign its commits?

Not dependably. The Codex CLI has a commit_attribution setting that can add a Co-authored-by trailer, but public reports disagree on whether it is on by default, and users can set it to empty. Many Codex commits carry no marker at all.

Is this the old Codex model from 2021?

No. Codex today is OpenAI's coding agent, available as a command-line tool, an IDE extension and a cloud service. The detector reads any code you paste, whichever version produced it.

Is my code stored?

No. The snippet goes to a third-party AI model provider (Groq, with Google Gemini as a fallback) for the analysis and the result comes straight back to you. We do not save the code or write it to our logs.