Watermark status Β· Updated Aug 2026

Which AI models watermark text?

As of August 2026, only Google Gemini watermarks text in production. Anthropic has announced watermarking for Claude's newest models; no current Claude model is marked yet. ChatGPT and open-weight models do not watermark text.

1/6mark text

1 of 6 major providers watermarks text today

1
Live watermark
Gemini, all models
1
Announced
Claude, models after Aug 2
4
No text watermark
ChatGPT Β· Copilot Β· Llama Β· DeepSeek

Status by provider

ModelStatusMethodCurrent modelsDetection
Claude
Anthropic
Announced
SynthID-Text
Statistical, in word choices
Opus 5, Sonnet 5, Fable 5, Opus 4.8
Not marked (pre-cutoff)
API announced, not shipped
Gemini
Google
Live
SynthID-Text
Active since 2024
All current models
SynthID detection available
ChatGPT
OpenAI
Planned
None deployed
Discussed, not shipped
GPT-5, o-series
No text detector
Copilot
Microsoft
None
No text watermarking
β€”
β€”
Llama
Meta
None
Open-weight, no mark
Self-hosted
β€”
DeepSeek
DeepSeek
None
Open-weight, no mark
All models
β€”

Per-model index β€” Claude

No public detector exists yet. Anthropic announced a detection API but has not shipped it, and no third-party tool can read the statistical watermark. Published text stays checkable the day it ships, so cleaning Claude-assisted text now closes the window. Clean your text β†’