Skip to content
ChatGPTDetect
English

GPT-6 detector: checking text from current ChatGPT versions

Yes. Text from the current ChatGPT models can be checked with ChatGPTDetect. Veread works at the family level, so GPT-6 Astra, Sol and Luna and GPT-6.1 Sol and Terra all go through the same check. Of fully AI-written OpenAI-family texts, 100% were detected.

Current versions, as of 10 November 2026

Which model a given person sees depends on plan, workspace and region, and OpenAI publishes no single picker list. Older GPT-5.2 to GPT-5.6 models are still in the API catalog.

How well the detector handles the family

Veread, Fileverity's text model, detected 94.9% of text from AI models that were never used in its development, which is the closest test of how it treats a release it has not seen. Human-written text was wrongly flagged 0.5% of the time. The facts page has the sources.

How to check text from each version

The watermark OpenAI added is a separate matter and applies to EU text only. See the ChatGPT watermark status.

Questions and answers

Can GPT-6 text be detected?

Yes, within limits. Veread reads writing style at the family level, so text from GPT-6 Astra, Sol and Luna and GPT-6.1 Sol and Terra goes through the same check. Of fully AI-written OpenAI-family texts, 100% were detected.

Which ChatGPT model do free users get?

From 8 October 2026, Free and Go plans use GPT-6 Luna in the Chat tab, according to OpenAI's release notes.

Is there a separate detector for GPT-5 text?

No. Older GPT-5.2 to GPT-5.6 models remain in OpenAI's API catalog, and the same check applies to them.

Can the detector say which version wrote a text?

No. It reports similarity to a model family, not a version. Style similarity is a signal, not proof of authorship.