Engine Updates
v2.1 currentEvery update to the GetSolved AI Detector is documented here. Each version ships improvements to detection accuracy, supported AI models, and user-facing features.
What's Included Today
A summary of all capabilities available in the current version (v2.1) of the GetSolved AI Detector. These features have been built up across six releases since the initial launch in December 2025.
Each capability is labeled with the version it was first introduced. All listed features are available to every user on the current engine.
~92% Detection Accuracy
Since v1.0
GPT / Claude / Gemini Attribution
Since v1.3
Downloadable PDF AI Report
Since v1.2
Sentence-Level Heatmap
Since v1.0
Mixed-Text Segmentation
Since v1.0
~24,000 Word Capacity
Since v2.0
Three-Layer Detection Engine
Since v1.0
English Language Support
Since v1.0
Current Version — v2.1
v2.1Current
April 2026
Borderline Precision & Performance Tuning
Version 2.1 builds on the v2.0 engine overhaul with recalibrated borderline-zone thresholds and improved detection confidence for edge-case texts. This update focuses on reducing false positives on human text that was previously flagged incorrectly.
~92% overall accuracy
Balanced evaluation set
~24,000 word capacity
With automatic segmentation
Model attribution
GPT, Claude, Gemini detection
PDF AI Report
Downloadable scan reports
Sentence heatmap
Per-sentence AI highlighting
Mixed-text detection
Independent segment analysis
What changed in v2.1
- →Recalibrated borderline-zone thresholds for improved precision on edge-case texts
- →Reduced false positive rate on heavily-edited human text by ~2 percentage points
- →Improved confidence calibration for less common AI models
- →Better scan performance for documents between 5,000 and 24,000 words
- →Minor improvements to sentence-level heatmap accuracy at document boundaries
Previous Updates
v2.0Major
Mar 2026
Detection Engine v2 — Major Upgrade
- →Completely retrained detection engine on a significantly expanded dataset for higher accuracy
- →Improved handling of code-mixed and template-heavy texts
- →Raised maximum document size to ~24,000 words with automatic segmentation for long documents
- →Improved detection of Gemini-generated text (+4 percentage points vs previous version)
- →Full per-model confidence breakdown now available in scan results
v1.3Feature
Feb 2026
Model Attribution
- →Detector now identifies which AI model generated the text — GPT, Claude, or Gemini
- →Per-model confidence levels shown alongside the main AI probability
- →Training data expanded with Claude 3 and Gemini 1.5 samples for better per-model detection
v1.2Feature
Jan 2026
Downloadable AI Detection Report
- →Download a formatted PDF report of any AI detection scan
- →Report includes: AI probability, model attribution, segment analysis, and sentence-level breakdown
- →Stamped with scan date, word count, and detector version for documentation purposes
- →Suitable for submission as evidence in academic or professional review processes
v1.1Feature
Dec 2025
Extended Text Capacity
- →Maximum input length extended beyond the initial 5,000-word limit to support longer documents
- →Automatic text segmentation for documents longer than ~2,000 words
- →Improved accuracy on mixed-authorship texts with partial AI content
v1.0Initial
Dec 2025
Initial Release
- →First public release of the GetSolved AI Detector
- →Three-layer detection: predictability scoring, linguistic analysis, and neural classification
- →Sentence-level heatmap highlighting AI-generated sections
- →Change-point segmentation for mixed-authorship detection
Results from different engine versions are not directly comparable. When tracking a document over time, note which version was used for each scan.