ChatGPT
Claude1. Paste the passage you want to review
Sign in and paste 20 to 5,000 characters into the text box, then select Analyze Text. Keep related sentences together. A very short excerpt provides little context for the model to assess.
AI Content Detector for Reviewing Writing Patterns
Paste a passage to get an estimated AI probability, a confidence label, and the writing signals behind the result. Use the flagged phrases as a starting point for review. The score is a probabilistic aid, not proof of authorship.
Check a passage of text
Paste 20 to 5,000 characters from a draft. The tool analyzes writing style; it does not need you to name the model or author you think wrote it.
Read the score with its confidence
The result includes an estimated AI percentage, a Likely Human, Mixed Signals, or Likely AI label, and low, medium, or high confidence.
Inspect the reasons and phrases
Read the explanation, writing signals, and any flagged excerpts. These are the model's observations about the passage, not evidence of where it came from.
Keep a person in the review
Use the analysis to inform an editorial conversation. False positives and false negatives are possible, so do not base grading, hiring, or disciplinary decisions on a score alone.
Understand why a draft received a score
The report describes patterns such as repetitive rhythm, generic phrasing, or hedged claims, and may list excerpts from your input. Read the whole passage before deciding whether any of those observations are useful.
Review a draft
Edit for your reader, beyond the score
If a passage is vague or repetitive, revise the wording and add context you can verify. The AI humanizer can suggest a rewrite. It cannot certify human authorship, and a changed detector score does not validate an edit.
Explore the humanizerHow to use an AI text detector responsibly
A writing pattern can suggest a question to investigate. It cannot establish who wrote a passage, whether its facts are true, or whether it was copied.
2. Read the explanation before acting
Review the probability estimate, confidence label, summary, signals, and flagged phrases together. A high confidence label is the model's assessment, not a measured accuracy guarantee. It can still be wrong.
3. Compare with evidence outside the tool
For an authorship question, discuss the writing process and look at drafts or source notes with the author. For an editorial task, use your audience, brief, and factual sources to decide what needs work.
- Writers: look for repetitive wording in your own drafts.
- Editors: use specific passages to start a feedback discussion.
- Tutors: discuss the observations without treating the score as a verdict on a student.
Build a clearer draft from the start
For a practical writing exercise, use the audience, context, and tone prompts in our AI email writing guide. Review whether the result communicates your intent; do not optimize an email just to reduce an AI score.
Pricing
One plan can run your whole AI workflow.
Credits work across chat, images, video, voice, music, avatars, editing, and upscaling. Upgrade only when you need more monthly volume.
Loading plans
Cancel anytime
Costs shown before generation
One credit balance across tools
Still have questions?
Picking a plan, custom credits, team pricing, billing — we usually reply within a few hours.
One studio, every tool
Your image generations plug straight into the rest of the Oakgen suite — all on one shared credit balance.
Explore more ways to create
AI content detector questions
Everything you need to know about the ai content detector.
Can this prove that text was written by AI?
No. The tool asks an AI model to estimate authorship from writing patterns. It has no access to an author's drafting history. Treat its probability score and confidence label as uncertain observations, not proof or a measured accuracy rate.
Can human writing be flagged as AI?
Yes. Human writing can share the patterns the detector looks for, and AI writing can avoid them. Both false positives and false negatives are possible. Short passages may be especially uncertain. Review the explanation and other evidence instead of relying on the score alone.
Is this a ChatGPT detector?
You can paste text from ChatGPT or another writing tool, but this detector examines general writing patterns. It does not identify the model that produced a passage, and there is no guaranteed detection rate for ChatGPT, Claude, Gemini, or other systems.
What do I receive after a check?
The report contains an estimated AI percentage, a Likely Human, Mixed Signals, or Likely AI label, a confidence label, a short explanation, and a list of signals. It may also list flagged phrases quoted from your text when the model identifies them.
Does it check plagiarism or factual accuracy?
No. It does not compare your passage against a database of sources, verify citations, or fact-check claims. An AI score says nothing conclusive about originality, truth, or the quality of the writing.
How much text can I check, and does it cost credits?
Each request accepts 20 to 5,000 characters and requires sign-in. Checks currently deduct no credits, but account eligibility, rate limits, and free-tier usage limits apply. An account with a negative credit balance must be topped up before it can make a request.
Can I upload a PDF or Word document?
The detector accepts pasted text only. Copy the passage you want to review from your document, keeping it within the character limit. It does not analyze document formatting, file metadata, or revision history.