You have a stack of student pastes, not one. Split drafts with --- or upload .txt and .md files. OmniKit scores ChatGPT-style tells in the browser with the same engine as the AI Writing Detector. A high score is a style flag, not proof of cheating and not a publisher detector. Copy CSV for notes, then talk about process with the writer.
Required fields on the left
Same style engine as the AI Writing Detector. A high score is a style flag — not proof a student used a model.
Appears after you run
Batch scan for teachers
Up to 12 drafts. Local scoring — no credits. Open “How to read these results” if Score, Label, or Flags is unclear.
This page batch-scans drafts in your browser with the same style engine as the AI Writing Detector. It does not spend credits and it does not name the model that wrote the text.
Use a high score to start a conversation about drafts, notes, and citations. Do not treat it as an honor-code verdict. Quoted methods sections can look formulaic.
Still unclear? Read the results guide on this page. For one paste with full signal detail, use the detector linked under Inputs.
A classroom AI checker batch-scores student drafts for ChatGPT-style writing tells. OmniKit uses the same local engine as the AI Writing Detector. A high score is not proof a student used a model, not a publisher detector, and not an honor-code verdict.
Teachers use it for a queue. Students use it to see flags before they submit. Keep the talk about drafts, notes, and citations.
OmniKit’s Classroom AI Checker batch-scores up to twelve drafts in the browser. Each paste is capped at 8,000 characters. You can split pastes with --- or upload .txt and .md files. CSV exports name, score, label, words, and flag hits. A high score is a style flag, not proof a student used a model.
Paste drafts separated by --- or upload .txt and .md files, up to twelve. Each paste is capped at 8,000 characters. Scoring runs in the browser. Copy CSV for name, score, label, words, and flag hits. No credits.
The engine matches phrase lists, rhythm, and structure. It does not identify GPT-4, GPT-5, Gemini, or Claude as the author.
Paste drafts separated by ---, or upload up to twelve .txt or .md files.
Scan in the browser. No AI credits.
Read score, label, words, and flag hits. Copy CSV for your notes.
Read the flagged passage. Ask for earlier drafts or notes. Quoted methods sections can look formulaic. Use the number as a prompt for a conversation, not as a penalty by itself.
Score is 0–100 for ChatGPT-style tells, not a percent chance the student used a model. Label is a band: 0–50 low, 51–75 medium, 76–100 high — so 45 can still read low. Flags count stock-phrase hits such as synergistic or landscape. Words is the chunk length. Name is Draft 1, 2, … or the filename.
The scan runs in the browser with the same engine as the AI Writing Detector. It does not fingerprint GPT-5, Gemini, or Claude. Short samples under-score. Copy CSV for private notes, not as a public grade.
No. Vendor-named guides reuse this scanner. Fluent assistant prose overlaps. OmniKit does not fingerprint a lab. Open the Gemini or Claude guides when that is the search query.
It will not watermark-decode a file. It will not replace a plagiarism index. It will not excuse skipping citation. Keep bibliography work on the citation generator.
A classroom score does not decide whether a public page is helpful. Google still judges useful, reliable content. For the open web, add experience and sources. For classwork, judge process.
Google Search Central, Using generative AI to create content: https://developers.google.com/search/docs/fundamentals/using-gen-ai-content . Creating helpful, reliable, people-first content: https://developers.google.com/search/docs/fundamentals/creating-helpful-content . SEO Starter Guide: https://developers.google.com/search/docs/fundamentals/seo-starter-guide
Cut stock openers. Add a detail only they know. Count length if the prompt has a cap. Fix grammar if the prose is messy. Do not hunt for a bypass.
A classroom AI checker batch-scores student drafts for ChatGPT-style writing tells. OmniKit uses the same local engine as the AI Writing Detector. A high score is not proof a student used a model, not a publisher detector, and not an honor-code verdict.
Up to twelve files or --- separated pastes. Each paste is capped at 8,000 characters.
No. Scoring is local, like the single-draft AI Writing Detector.
No. Style overlap is not authorship. Use the score as a conversation starter.
Label is a band, not a second score. 0–50 is low, 51–75 is medium, 76–100 is high. Score 45 sits in the low band. Flags count stock-phrase hits, not models.
Those guides reuse this engine. They do not fingerprint a vendor. Open them when the search query names a model.
Keep measuring in the same cluster — or jump to the next decision.
Google’s guidance on AI-generated content still puts responsibility on the publisher of a page, not on a classroom score. For student work, the responsible move is process: drafts, notes, and citations you can see.
Scan one paste on the AI Writing Detector. Scan a queue here. Polish scholarly cadence on Researcher Humanizer. Do not market any of those tools as a cheating verdict.
