Verification guides

Last updated: October 15, 2026

Short answer: AFIP guides explain how to check whether text, images, audio or video were made or altered by AI, and how much weight each check can carry. Each guide starts with a short answer, sets out the methods and their limits, and ends with steps you can apply to a real case.

Where to start

Pick the row that matches the case in front of you. The first guide gives the method; the second covers the most common complication.

If you need to	Start with	Then read
Decide whether a document was written with AI	How to tell if something is written by AI	AI humanizer detection
Check a suspicious video or photo	How to spot a deepfake	AI image detection
Verify a voice recording	AI voice detection	Deepfake verification steps
Understand labels and Content Credentials	C2PA explained	Content provenance guide

Text and authorship

How to tell if something is written by AI: Manual signs, statistical markers and a five-step cross-check for AI writing, plus whether any AI checker is reliable enough to trust on its own. Last updated October 16, 2026.
AI humanizer detection: How rewriting tools alter AI text, and what a reviewer should check when a clean detector score is the only evidence offered. Last updated September 18, 2026.
Can AI detectors be wrong?: Why detectors produce false positives and false negatives, what a score can prove, and the AFIP evidence-weight scale for weighing a flag. Last updated October 15, 2026.
How managers can detect AI in workplace reports: When to check staff documents, which signals matter, how to run a detector responsibly and how to raise the question fairly. Last updated October 15, 2026.
Can employers detect AI resumes and cover letters?: What applicant tracking systems do and do not check, a recruiter screening rubric, and how to keep AI screening fair. Last updated October 15, 2026.
AI detector false positives: Why a low false positive rate still wrongly flags many people at scale, a worked base-rate example, and how to reduce the harm. Last updated October 15, 2026.
Stylometry and authorship analysis: What stylometry measures, the Federalist Papers and Robert Galbraith cases, and why style comparison cannot prove AI use alone. Last updated October 15, 2026.

Images and video

How to spot a deepfake: Visual, audio and contextual signs of deepfake video and images, with a step-by-step verification process. Last updated June 12, 2026.
AI image detection: How generated images are identified through generation artifacts, metadata and forensic analysis. Last updated June 18, 2026.

Audio and voice

AI voice detection: How cloned and synthetic speech is identified through spectral and timing analysis, and where those methods fall short. Last updated June 15, 2026.

Provenance and policy

C2PA and Content Credentials explained: What the C2PA standard records, how Content Credentials work, and the limits of labels attached by the creator. Last updated June 22, 2026.
Content provenance: the complete guide: How the origin and edit history of media is established, and how label-based and forensic approaches compare. Last updated June 8, 2026.
EU AI Act and provenance rules: What European transparency duties mean for labeling and verifying AI-generated content. Last updated May 29, 2026.

How to use these guides

Each guide treats a detection result as one piece of evidence, never a verdict. The AFIP methodology explains how separate kinds of evidence are weighed together, the glossary defines technical terms such as false positive rate and perplexity, and the research briefs cover the science behind each method in more depth.

Guides are reviewed when the tools or standards they describe change. The date beside each guide is the date its content last changed.