Recommended Free Tools
OpenAI’s text watermarking rollout is phased, not a switch that instantly marks every ChatGPT response. As of October 7, 2026, API customers worldwide can opt in for select models, while OpenAI says it plans to begin adding invisible watermarks to eligible ChatGPT and Codex text output in the EU over the coming weeks. Its text detector is initially available only to approved researchers and expert organizations.
What OpenAI is rolling out, and where
OpenAI announced the plan on October 5, 2026. It has three distinct parts, with different audiences and availability:
As an Amazon Associate I earn from qualifying purchases.
| Feature | Who and where | Availability and default |
|---|---|---|
| API text watermarking | API customers worldwide, for select models | Customers can opt in; OpenAI says it is off by default. |
| Watermarks in ChatGPT and Codex text output | Eligible output in the European Union | Planned for rollout over the coming weeks. OpenAI has not specified the date for each account or listed all eligible models and accounts. |
| Text detector | Researchers and expert organizations | Access is application-based and initially restricted to approved applicants. |
These are not interchangeable settings: the API option is a global opt-in, the planned ChatGPT and Codex rollout is EU-focused, and detector access is restricted. OpenAI’s announcement does not say the rollout is complete or that watermarking is a global default. (OpenAI, “Our approach to EU text provenance rules,” October 5, 2026; OpenAI Help Center, “EU AI Act: OpenAI Resources and Customer Guidance.”)
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
How textGrain puts a signal in text
OpenAI calls its text watermarking method textGrain. As a model chooses among possible words or word pieces, the method subtly shifts those choices to create a statistical pattern that a detector can search for. The signal is part of the wording; it is not a visible label or a separate mark a reader can spot.
#1 Best Overall
OpenAI says textGrain does not add hidden characters, invisible spaces, or unusual punctuation. That makes it different in form from a label attached to a document or metadata embedded in a file: the signal depends on patterns in the generated text itself. (OpenAI Help Center, “Provenance signals in OpenAI-generated content.”)
How well OpenAI says its detector works
OpenAI’s reported evaluation results show that detection depends on the material being checked. In one evaluation, it reported detecting about 80% of 200-token psychology passages and about 95% of 400-token psychology passages at a target false-positive rate of 1%. It said detection was substantially lower for mathematics, where there is less flexibility in word choice.
In a separate evaluation of 400-token passages, OpenAI reported detection of about 92% before synonym changes, 66% after 10% of the words were replaced, and 17% after 25% were replaced. These are company-reported results for the evaluated passages, not guarantees for arbitrary documents or real-world detector performance. OpenAI also notes that shorter or constrained text is harder to detect and that results can include false positives and false negatives. (OpenAI, “Our approach to EU text provenance rules,” October 5, 2026.)
Can editing remove or weaken a text watermark?
Editing can weaken the statistical pattern, as OpenAI’s synonym-replacement results illustrate, but those figures do not establish a universal threshold at which a watermark disappears. Translation, extensive rewriting, or other transformations may also make detection less reliable; the cited evaluations do not quantify every kind of change. A detector’s failure to find a signal therefore does not establish that text was written by a person.
Rank #3
- Used Book in Good Condition
What a detector result can—and cannot—show
OpenAI says a detected watermark can indicate that an OpenAI system generated or processed part of a passage. It does not establish how much a person contributed, who used the system, who owns the text, or who is responsible for it. It also cannot determine whether the text is accurate or misleading.
- A positive result is a limited provenance signal, not proof of a specific person’s involvement, a plagiarism finding, or an authorship verdict.
- A negative result is inconclusive. The passage may be too short, edited, translated, generated by an unsupported model, created before watermarking, or produced by another provider.
Watermark detection is not a truth check. It cannot, by itself, settle the origin or reliability of a document. (OpenAI Help Center, “Provenance signals in OpenAI-generated content.”)
Rank #4
What Article 50 of the EU AI Act requires
Article 50 sets out two different transparency duties relevant to generated text. One applies to providers of generative systems; the other applies to deployers in a narrower public-facing situation. A machine-readable provider mark is not the same thing as a deployer’s public disclosure.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11| Duty | Who must act | What it requires |
|---|---|---|
| Article 50(2): marking and detectability | Providers of systems generating synthetic audio, images, video, or text | Ensure outputs are marked in a machine-readable format and detectable as artificially generated or manipulated. Technical solutions must be effective, interoperable, robust, and reliable as far as technically feasible, taking account of content-specific limitations, implementation costs, and the state of the art. |
| Article 50(4): public-facing disclosure | Deployers publishing AI-generated or manipulated text to inform the public on matters of public interest | Disclose that the text was artificially generated or manipulated. The text exception applies where it has been subject to human review or editorial control and a natural or legal person holds editorial responsibility. The statute also provides a law-enforcement exception. |
The European Commission says Article 50 applies from August 2, 2026. A limited grace period applies only to the provider-side marking and detection duty in Article 50(2): providers of systems placed on the market before August 2 must comply with that duty from December 2, 2026. The Commission also says content generated before August 2 does not need retroactive labeling. These dates do not turn the separate deployer disclosure duty into the same obligation or grace period. (European Commission AI Act Service Desk, Article 50, consolidated text as of July 27, 2026; European Commission, “Code of Practice on Transparency of AI-generated Content,” and linked Article 50 FAQ.)
Is the Commission’s Code of Practice mandatory?
No. The Commission describes the Code of Practice as a voluntary framework for demonstrating compliance, while Article 50’s transparency requirements are legal obligations. The Commission says signatories can rely on the code’s measures; providers choosing other means must demonstrate that their measures are adequate. The code’s voluntary status does not make the statutory duties optional. (European Commission, “Code of Practice on Transparency of AI-generated Content.”)
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How text watermarking fits with other provenance signals
OpenAI describes a broader, layered approach to provenance that includes Content Credentials (C2PA) and invisible watermarks for supported media. Those approaches should not be treated as one universal detector. textGrain is a statistical signal in text; C2PA involves content credentials, and media watermarking concerns supported image or audio outputs. OpenAI describes public verification tools for supported image and audio signals, while its text detector is initially restricted.
Each signal has different content coverage, access conditions, and sensitivity to editing or transformation. A detector or verification tool for one format should not be assumed to authenticate every kind of AI-generated content. (OpenAI Help Center, “EU AI Act: OpenAI Resources and Customer Guidance”; OpenAI Help Center, “Provenance signals in OpenAI-generated content.”)
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
