Home
Wednesday, 7 October 2026 WARSAW 15:33 NEW YORK 09:33
AI · Search · Performance · E-commerce
Just in

Meta tests an Ads Manager assistant that changes budgets and targeting itself

AI

OpenAI will watermark ChatGPT text in the EU. Light edits weaken detection

textGrain arrives in ChatGPT and Codex in the EU in the coming weeks. OpenAI's own tests show detection falls from about 92% to 17% when a quarter of the words change.

Grzegorz Kubicki 7 Oct 2026, 09:07 reported from 2 sourcesOpenAI (source document), ACS Information Age
Hands typing on a laptop keyboard on a wooden desk

Photo: Tima Miroshnichenko / Pexels

OpenAI said on 5 October that it will add an invisible watermark to text from ChatGPT and Codex in the European Union “over the coming weeks”. The system is called textGrain. API customers anywhere can switch it on, but there it is off by default.

The company links the move to the EU AI Act, which asks providers of generative AI to make machine-made text detectable by software. The transparency rules started to apply on 2 August.

textGrain detection drops to 17% after heavy edits

OpenAI published its own test results. On unedited text the detector found the watermark about 92% of the time, the company said. When 10% of the words were swapped, that fell to 66%. With a quarter of the words replaced, it fell to 17%.

AdSense in-article 336x280

The company also warned that a missing watermark “does not prove human authorship”. The detector will not be public at first. OpenAI will give it to approved researchers and expert groups, who can apply through a form.

For images, OpenAI already adds Content Credentials, a C2PA standard, plus a SynthID watermark. Anyone can check an image at openai.com/verify.

Anthropic made a similar pledge in August

OpenAI is not alone. In August, Anthropic said it would put hidden watermarks into text from Claude, starting with future models, ACS Information Age reported. Both moves follow the EU code of practice on transparency.

Platforms are tightening labels too. Instagram shows unlabelled AI profiles to fewer people, and Google recently asked sites to fact-check AI content by hand.

EU marketing teams using ChatGPT for copy

For brands, the change mostly touches copy written in ChatGPT and pasted into ads, product pages or emails. The watermark is invisible to readers, so nothing changes on the page. Teams that edit drafts heavily will leave a weaker signal.

The open question is who gets the detector. If OpenAI opens it to platforms or ad networks, a watermark could start to affect how ad copy is reviewed. OpenAI has not given a date for wider access.

Sources

  1. OpenAI (source document) — textGrain announcement, rollout, test results, 5 Oct 2026
  2. ACS Information Age — Anthropic text watermark pledge, 17 Aug 2026

Published by IMARKETING Sp. z o.o., Al. Jana Pawla II 150/127, 31-864 Krakow, Poland. Company register (KRS) no. 0001203282, VAT ID PL6751820741.  ·  Editorial office: Al. Jana Pawla II 150/127, 31-864 Krakow, phone +48 516 110 813, editor@marketingnewsroom.com.  ·  Editor-in-chief: Grzegorz Kubicki.  ·  Corrections: editor@marketingnewsroom.com.