What’s New With Copyleaks

Copyleaks added an optional Object Detection field to the AI Image Detector API endpoint, so you can check specific objects in the image for AI manipulation. Supply a list of up to three objects to look for, such as a face, a necklace, or a car, and the API returns a pixel-level segmentation mask for each object found in a single synchronous call.

Each returned mask can be decoded to highlight, crop, or blur the object’s region.

Learn more: Object Detection API guide

Copyleaks Web Platform: September 8, 2026 (graduated rollout)

LMS and API: September 22, 2026

We’ve released version 11 of the AI Text Detection model. This release improves detection accuracy and extends coverage to recently released large language models.

Frequently Asked Questions

What new models can AI Detector now identify with V11?
V11 adds support for the most recent and widely used AI models, including:

• OpenAI: GPT-5, GPT-5.1, GPT-5.2 Thinking, GPT-5.3 (Codex and Instant), GPT-5.4, GPT-5.5, GPT-5.6
• Google: Gemini 3.x Pro and Flash (up to 3.6)
• Anthropic: Claude Opus 4.5, 4.6, and 4.8; Claude Sonnet 4.5, 4.6, and 5; Claude Haiku 4.5
• xAI: Grok 4.1
• DeepSeek: deepseek-v4-flash

Will my reports look different?
No. The report and workflow are unchanged.

Learn more on our AI Detector Testing Methodology page.

Every Copyleaks screen in your LMS now shares one navigation bar. It persists across the Copyleaks course hub, the assignment page, and the report.

• One menu: open the course hub, Analytics, and Quick scan from the menu on any screen. 
• Clickable breadcrumb path: see where you are, such as Course › Assignment › Similarity report, and navigate back to any step.
• Copyleaks course hub: the Assessment Overview is now the Copyleaks course hub, an area to access every Copyleaks tool. Each assignment row has quick links to its settings and analytics.
• Quick scan: the Teacher Scanning Tool is now Quick scan. It no longer needs a separate integration, and admins turn it on or off in Admin scan settings.

AI doesn’t just write text. It cites sources too, and some of those sources don’t exist. Citation Checker checks every reference in a document against a real source and tells you which ones it couldn’t confirm. A perfectly formatted citation still has to point to something real.

How It Works

Copyleaks pulls each reference out of the submitted text and checks the author, title, and date. It then compares each reference to trusted sources:

• Academic references (papers, journal articles) are verified against the Copyleaks academic citation index.
• Non-academic references (web pages, documentation, encyclopedias) are verified by opening the cited URL and comparing the live page to the citation.

Each citation comes back with detailed results:

• Found: it points to a real paper or live page.
• Suggestions: the source is real, but a detail is off. You get a suggested fix, such as the right year or author spelling, along with its source.
• Not Found: Copyleaks could not confirm the source.

Citations Cross-Check matches every in-text citation to the reference list and returns one of four results for each:

• Uncited Reference: In the reference list, but never cited in the text.
• Matched: The citation and reference agree.
• Mismatch: A detail differs, such as the year.
• Missing Reference: Cited, but not in the reference list.

Who It’s For

• Education/EdTech: check a student’s citations as part of grading, right next to the matched text and AI scores.
• Editors, fact-checkers, and analysts: confirm that a sourced claim points to a real article before it is published.
• Journal and peer-review editors: check a manuscript’s reference list before acceptance.

Frequently Asked Questions

Does it tell me if the text itself was written by AI?

No. Citation Checker verifies references only. To check whether text is AI-generated, use the AI Detector.

Is this the same as plagiarism checking? 

No. Plagiarism checking shows whether text was copied. Citation Checker shows whether the cited sources are real. A document can pass one and fail the other.

Learn more: Citation Checker API guide

Our latest update improves accuracy on blended images, where generative AI edits a real photo. The new v1.3 also improves accuracy in fine details and across image qualities, helping it stay current with the newest generative models.

The new Image Plagiarism Detection API is now available. Submit an image and get back a sorted list of matches, with the web pages where each one appears. Full matches are exact copies, while partial matches are modified (cropped, resized, recolored, or edited).

Who It’s For

• Photographers and artists: find unauthorized use of your work.
• Brand teams: track where logos, product photos and campaign images show up.
• Stock photo platforms, news agencies and UGC moderators: check whether a submitted image was copied from somewhere else.
• Educators: check whether images in student work came from public sources.

Frequently Asked Questions

Which image formats can I submit?

JPG, JPEG, PNG, GIF, BMP, WebP and ICO.

Are there size limits?

Files must be under 20 MB and no larger than 75 megapixels. Filenames can be up to 255 characters and must include the extension.

Can I check more than one image at a time?

Each request takes one image. For volume, the limit is 1,800 requests per 15 minutes per user.

Does it tell me if an image is AI-generated?

No. This tool looks for copies of an image. To check whether an image was made by AI, use the AI Image Detection API.

Learn more: Image Plagiarism Detection API guide

The AI Video Detection API detects fully and partially AI-generated video files, covering the visual track and embedded audio. It returns frame-level and millisecond-level timestamps showing exactly where AI manipulation begins and ends.

Who It’s For

• Insurance: screen video and voice evidence submitted with claims for AI manipulation.
• Government and public sector: identify deepfake videos of officials and candidates.
• Media and publishing: verify user-submitted and source footage before it airs or goes to print.
• Entertainment and film: confirm the authenticity of footage.
• Enterprise and API integrators: build video checks into content moderation, review, or compliance workflows.

Detection Categories

• Fully-AI: entirely generated from a prompt (e.g., text-to-video)
• Fully-Human: authentic recording without AI editing
• Mixed / Spliced: partial manipulation (e.g., deepfake face-swap, AI voice clone inserted into a real clip)

Supported Models

Google Veo, ByteDance Seedance, Hailuo, Vidu AI, Runway, Wan

API Overview

• Single async endpoint: submit a file URL, poll or webhook for results
• Response includes: AI scores for video and audio separately, one overall AI score, and timestamps of AI-detected segments
• File is passed by URL reference, not uploaded directly

File & Media Requirements

File size	Max 500 MB
Resolution	Min 360×360 px
Frame rate	Min 16 FPS
Duration	Min 2s (video and audio), max 1 hour
Filename length	≤ 255 characters, must include extension
Supported formats	.mp4, .avi, .mov, .mkv, .webm, .flv, .wmv, .mpg, .m4v, .3gp, .mxf

Learn more: AI Video Detection API guide

Check content for AI, plagiarism, and grammar issues without leaving WordPress. The Copyleaks plugin shows detection scores right in your post list and page list, so editors can see which content needs a closer look before it goes live. Access full Copyleaks scan reports in one click.

• AI detection and plagiarism scores: a score for each scanned post or page.
• Grammar suggestions: easily fix spelling and grammar as part of the review process.
• Automatic scanning: scans start when a post reaches a predesignated status: Pending Review or Published. You may also scan on demand.
• Role permissions: choose who can see scores, adjust scanner sensitivity, or bypass content rules.

Learn more: Get the Copyleaks Plugin

Teachers can now check student work for AI and text match without leaving Google Classroom. Review happens in the grading workflow teachers already use.

What’s In It

• AI content detection: Flags text generated by LLMs, right in the Student Work tab.
• Text match: Compares work against web pages, academic journals, and institutional repositories.
• Cross-language detection: Catches sources translated from another language.
• Grammar checks: Optional, per assignment.
• Interactive reports: Color-coded and viewable in Classroom.
• Templates: Upload the assignment prompt so common questions don’t count as matches.
• Scan controls: Teachers scan all submissions at once or one at a time.
• Settings per assignment: Detection sensitivity, cross-language scans, Shared Data Hub, student privacy, and report moderation.

You can now check images for AI right in the Copyleaks app, with no API integration needed. Upload an image to see whether it was AI-generated or AI-edited.

How It Works

Each image scan returns as AI Detected or AI Not Detected. The report also shows where in the image AI was detected.

There are two pixel-level overlay options:

• Standard AI view: highlights the specific regions detected as AI. AI-generated regions are red, and human-made regions are green.
• Heatmap: shows how confident the model is about every region. The deeper the overlay color, the more likely it is AI-generated, from yellow (low) to red (high).

The enhanced Copyleaks analytics dashboard gives a more detailed view of performance and integrity at every level of your organization. See insights on student progress, AI trends, and course performance. Identify courses or assignments where academic integrity may be at risk.

• View data by assignment, student, course, and instructor.
• Flexible time frames: select date ranges to better track changes over time.
• Easier navigation: a visual dashboard, with one-click analytics buttons built into LMS assignment pages.

Learn more: Analytics Dashboard reports

The new version offers accuracy level improvements on social media, small, low-quality, and professionally filtered images. It also adds support for new models and smartphone features (Apple Intelligence Photo Tools, Google Magic Eraser, and Samsung Generative Edit).

File & Media Requirements

Formats	PNG, JPG/JPEG, BMP, WebP, HEIC, HEIF
Minimum size	518×518 px
Maximum size	6000×4500 px (27 MP).
Max file size	32 MB

Learn more: AI Image Detection API reference

Copyleaks now works directly with native Canvas assignments through the Canvas Asset Processor (LTI 1.3). Instructors turn on Copyleaks scanning in the standard Canvas assignment setup, and similarity and AI reports appear in SpeedGrader and the submission view.

• Native Canvas assignments: no separate External Tool assignment needed.
• Reports where you grade: view similarity and AI reports in SpeedGrader and the submission view.
• The new Canvas standard: Instructure recommends the Asset Processor instead of the older Canvas Plagiarism Framework (CPF) for all new integrations.

How to get started

Ask your Canvas representative to turn on the Asset Processor feature flag for your institution. Then follow our guide to add it to your Canvas instance.

Learn more: Canvas Asset Processor help guide

Copyleaks Web Platform: October 26, 2025 (graduated rollout)

LMS and API: November 10, 2025

The V10 AI Text Detector improves detection of text from newer models, led by OpenAI GPT-5.

Learn more on our AI Detector Testing Methodology page.

Smarter Content Oversight. Powered by Contextual AI.

We’re excited to announce the launch of Text Moderation, a new feature designed to help organizations maintain safe, respectful, and inclusive digital environments. Unlike traditional keyword filters, Text Moderation uses contextual AI to evaluate how language is used, making it easier to distinguish between harmful or policy-violating content and harmless conversation.

Whether your team is moderating user-generated content, managing classroom discussions, or ensuring compliance across community platforms, Text Moderation delivers clarity, control, and confidence.

Key Features

• Context-Aware Flagging
  Goes beyond basic keyword filters by analyzing the way language is used, reducing false positives and unnecessary disruption.
• Robust Tagging System
  Flags content across categories like sexual, toxic, violence, profanity, self-harm, harassment, hate speech, drug use, firearms, cybersecurity, and more—helping teams align detection with policy requirements.
• Customizable Filters
  Select which categories to monitor so you can tailor detection to your platform’s unique guidelines and risk levels.
• Cultural & Regional Nuance Detection
  Recognizes slang, idioms, and tone across English dialects for more accurate, globally relevant moderation.
• In-Context Highlighting & Explanations
  Pinpoints flagged text within content and provides human-readable explanations for why it was flagged.

Who It’s For

• Content reviewers & trust and safety teams
• Publishers & media platforms
• Social media and UGC app moderators
• Review platforms and feedback tools
• Online forums and community managers
• Teams curating or moderating LLM training data

Why It Matters

For years, customers have asked for moderation tools that understand meaning, not just words. Keyword-only systems often over-police users and miss nuance. Text Moderation solves this by delivering contextual awareness that helps teams act with precision—protecting communities without silencing them.

For customers already using Copyleaks AI detection, Text Moderation adds another layer of confidence, ensuring that repeated or reused AI-generated patterns tied to harmful content don’t go unnoticed.

The new AI Image Detection API is now available, enabling you to detect if an image is AI-generated or AI-edited.

Key Features

• Pixel-level analysis: See exactly which parts of an image are AI-generated vs. human-created, with overlays to show if the part you care about was manipulated
• Metadata extraction: Automatically extract AI generation metadata when available (issuer, creation time, tool used)

Who It’s For

• Insurance: Screen claim photos (vehicle, property damage, medical) for manipulation.
• Financial services: Check documents, statements and IDs submitted at account opening (KYC).
• Government: Verify IDs, passports and official records, digital or scanned.
• Media and publishing: Journalists and editors validating images before publication.
• Online marketplaces: Confirm listing photos match the real product.
• Profile and UGC platforms: Flag AI-generated or manipulated profile photos (dating, social).
• Real estate: Catch altered or AI-staged listing photos, such as on short-term rental platforms.
• Enterprise and API integrators: Embed detection in content management or claims processing software.

File and media requirements

Formats	PNG, JPG/JPEG, BMP, WebP, HEIC, HEIF
Minimum size	512×512 px
Maximum size	6000×4500 px (27 MP).
Max file size	32 MB

Learn more: AI Image Detection API reference

Copyleaks Web Platform: June 24, 2025
LMS and API: July 8, 2025

Overview

AI Detector V9 is built to keep pace with the evolving landscape of Large Language Models (LLMs). This update delivers greater detection accuracy, enhanced contextual analysis, and improved performance in complex, high-stakes environments like education, publishing, and enterprise.

What’s New in V9?

Improved Accuracy & Contextual Analysis
V9 strengthens Copyleaks’ detection engine with:

• Enhanced precision in identifying AI-generated content
• Smarter contextual understanding for nuanced writing
• Optimized performance across diverse use cases and formats

Frequently Asked Questions

Q: Why this update?
A: V9 is a crucial upgrade designed to keep pace with the latest advancements in generative AI, ensuring Copyleaks remains the most accurate and reliable detection solution available.

Q: What new models can AI Detector now identify with V9?
A: V9 expands support to include the most recent and widely used AI models, such as:

• OpenAI: GPT-4o, GPT-4.1, GPT-4o-mini, GPT-0.3
• Google: Gemini 2.5 Flash, Gemini 2.5 Pro (03-25)
• Anthropic: Claude 3.7 Sonnet (20250219)

These additions enable deeper coverage and more confident identification of content generated by today’s most powerful LLMs.

Q: Is V9 validated for accuracy?
A: Yes. V9 has been rigorously tested against real-world data and use cases by Copyleaks’ QA and Data Science teams. It maintains our industry-leading benchmark for accuracy.

Learn more on our AI Detector Testing Methodology page.

Overview
The Copyleaks Google Docs Add-on allows users to scan content directly within their Google Docs for plagiarism and AI-generated text. This seamless integration eliminates the need to upload files or switch platforms, enabling real-time content verification and editing.

Key Features

• Real-Time Detection: Instantly scan for plagiarism and AI-generated content while writing.
• Snapshot Results: View originality and AI insights in a sidebar without leaving Google Docs.
• Edit-as-You-Go: Make quick revisions to flagged content directly within the document.
• Integrated Access: Open detailed reports in the Copyleaks app if needed.
• Seamless Workflow: No more uploading or moving between tools—scan and edit in one place.
• Multi-Audience Support: Designed for individuals, students, educators, and enterprise teams.

Target Users

• Students & Educators: Uphold academic integrity effortlessly.
• Writers & Content Creators: Ensure originality and avoid accidental plagiarism.
• Enterprise Teams: Protect brand reputation and maintain compliance directly in the workflow.

Benefits

• Verify originality without disrupting your writing flow.
• Save time with instant, actionable insights.
• Enhance content integrity and compliance across educational, personal, and enterprise use cases.
• Easily scale across individual users or entire organizations using Google Workspace.

Overview:

We’re excited to introduce AI Logic, Copyleaks’ next-generation AI detection technology designed to provide deeper transparency and explainability in content analysis. AI Logic brings two powerful components:

• AI Source Match — Identifies if flagged AI-generated content has been published elsewhere, helping to detect potential AI plagiarism or content reuse.
• AI Phrases — Highlights patterns and phrasing statistically common in AI-generated text, providing additional insight into detection outcomes.

Key Benefits:

• Greater transparency into why content is flagged as AI-generated.
• Source-backed evidence for more confident decision-making.
• Multi-layered insights supporting authorship validation, originality assurance, and responsible AI use.

Availability:

AI Logic is now integrated into Copyleaks’ web platform and API. LMS will be available late summer 2025.

AI Insights is now available in English, Spanish, French, German, Portuguese, and Italian, with more languages currently in the works.

• Major improvement in accuracy for English language: We have made significant improvements in the AI detection capabilities compared to V6.1 in English. No changes have been made to the other languages.
  
• Humanizers and paraphrasers detection: Using the highest sensitivity results in an improved paraphrasers and humanizers detection. See definition below.
  
• AI Insights: a new version of the AI Insights in English is published, built on a much larger database for improved accuracy, as well as new AI Insights for humanizers and paraphrasers tools.
  
• Faster models: This version ensures faster and more efficient AI detection in English resulting in an even quicker prediction.

A first-of-its-kind, patent-pending technology that provides a comprehensive analysis of the factors leading to the identification of text as AI-generated. This approach enhances transparency and deepens the understanding of AI patterns, enabling more accurate assessments of authenticity and originality, thereby supporting informed decision-making.

• Show the frequency of the phrase and how often it is used in AI outputs compared to Human outputs in the form of a ratio.
• Differentiate between low, medium, and high-frequency phrases found in AI content in shades of purple.
• Provides additional resources for the content reviewer to understand our AI detection capabilities and possible next steps they can take with the student or writer.