How it works
Upload content. Get a clear, evidence-based assessment. Make confident editorial decisions.
Image verification
Upload an image file or provide a direct image URL. The image is stored securely and submitted to external detection services that analyse visual patterns, compression artefacts, and statistical signals associated with AI-generated or manipulated media.
You receive a verdict (e.g. likely AI-generated, likely authentic), a confidence score, and — when relevant — a heatmap overlay highlighting the specific regions of the image flagged for concern.
Results are intended to support editorial judgement, not replace it. No detection system offers guaranteed accuracy; this tool helps your team flag risk and prioritise review.
Typical workflow
- 1Upload or paste an image URL
- 2Image is analysed for AI generation and manipulation signals
- 3Review verdict, confidence score, and optional heatmap
Text analysis
Paste article text into the analysis tool. AI detection models analyse linguistic, structural, and semantic patterns across the submitted text to assess the likelihood it was machine-generated.
You receive an overall AI-probability score and a sentence-level breakdown highlighting which sections are most likely AI-written. An optional pre-processing step can clean up formatting and typos before analysis.
These results help editorial teams identify submissions that warrant closer review. Detection is probabilistic — the tool flags risk rather than delivering certainty.
Typical workflow
- 1Paste article text into the editor
- 2Optionally run AI cleanup to normalise formatting
- 3Review overall score and sentence-level highlights
Video deepfake verification
Submit a video URL for classification. The service analyses frame-level patterns and temporal consistency to detect AI-generated or manipulated video content — deepfakes, face-swaps, and synthetic footage.
Audio verification
Verify audio recordings for synthetic or cloned voices. Essential for checking interview recordings, podcast clips, and phone call audio submitted by sources — voice cloning is now cheap and convincing.
Editorial review
An AI-powered forensic reviewer that reads submitted articles and returns a structured verification brief. This is not a rewriter — it's a pre-publication risk assessment tool.
It identifies unverified claims that need sourcing, flags statistics without attribution, highlights logical inconsistencies, and suggests specific questions the editor should ask the author. The output is a checklist, not a judgement.
What it returns
- • Risk level assessment (low / medium / high)
- • Unverified claims with suggested verification steps
- • Unsourced statistics and suspicious figures
- • Logical contradictions or consistency issues
- • Passages that read as AI-generated boilerplate
- • Specific verification questions for the author
Authorship analysis
Goes beyond “is this AI?” to estimate which model likely generated the text. Analyses vocabulary diversity, sentence patterns, and transition word usage to distinguish between ChatGPT, Claude, Gemini, and human writing. Detects mixed human/AI authorship within a single piece.
Metadata review
Extracts EXIF metadata from images to verify provenance. Identifies camera make/model, editing software, creation dates, and GPS data. Detects AI generation tools in metadata, checks for C2PA/Content Credentials, and flags stripped metadata — a common indicator of manipulation.
Watchdog monitoring
Add RSS feeds for continuous monitoring. Scan your own publication, competitor sites, or specific contributor bylines. Get flagged when articles with high AI probability appear — reactive detection becomes proactive editorial intelligence.
Bulk archive scanning
Submit a sitemap URL or list of article URLs. The system extracts text from each page and runs AI detection. Audit your entire back-catalogue to answer: “How much AI content have we already published?”
Compliance & trust certification
Generate structured compliance reports covering any date range — total checks, detection results, verification badges issued, and team member activity. Exportable as JSON or CSV for regulatory submissions.
The Trust Certification programme automatically grades publishers (Bronze → Silver → Gold → Platinum) based on verification consistency. Embed a public trust badge on your site to signal content authenticity commitments to readers and advertisers.
Made By Humans is designed to support editorial decision-making. Detection results are probabilistic and should be used alongside your team's professional judgement. We do not claim guaranteed accuracy or infallible results.
Ready to try it?
Create a free account and start verifying content in minutes.