ProvenanceGuard
← Go back

ProvenanceGuard guides

Long-form guides on content provenance for AI-generated images: inspection, sealing, fingerprints, platform survival, lineage, conformance evidence and the disclosure rules, written from what the tool actually measures. Start with the free seal, or look a travelled copy up in Recovery.

What Is Content Provenance for AI-Generated Images?

Content provenance explained for AI images: the record of origin, registration and copies, and why a registry survives where metadata does not.

Published November 14, 2026 · 8 min read

What Is C2PA and What Does a C2PA Manifest Actually Contain?

What C2PA is, what a manifest holds, how JUMBF carries it inside a file, and why distribution, not creation, is where manifests are lost.

Published November 14, 2026 · 8 min read

C2PA Manifest vs a Signed Provenance Assertion: What Is the Difference?

The honest comparison: a standards-based C2PA manifest inside the file versus ProvenanceGuard's own HMAC-signed XMP assertion plus registry record.

Published November 14, 2026 · 8 min read

Why Do Social Platforms Strip Image Metadata on Upload?

Why uploads lose EXIF, XMP and manifests: re-encoding, resizing, privacy and cost, and what that means for any marking strategy built on metadata.

Published November 14, 2026 · 8 min read

How to Check If an Image Has C2PA or Other Provenance Metadata

A practical byte-level check: C2PA and JUMBF markers, XMP, EXIF, PNG text chunks and prior assertions, read from the file rather than its name.

Published November 14, 2026 · 8 min read

How to Label AI-Generated Images Before You Publish Them

Labelling AI images so disclosure survives: a visible label in the pixels, an accurate caption, and a registry record behind both.

Published November 14, 2026 · 8 min read

EU AI Act Article 50: What AI Image Labelling Requires in Practice

Article 50's transparency duties as the product's evidence wording states them: applied 2 August 2026, marking grace ending 2 December 2026, and the records that support them.

Published November 14, 2026 · 8 min read

California AI Transparency Act: What It Means for AI Images

SB 942 as the product's evidence wording states it: operative 2 August 2026, provenance information registered and available on request, and what to keep per asset.

Published November 14, 2026 · 8 min read

New York's Synthetic Performer Law: Disclosure for AI People in Ads

GBL section 396-b, effective 9 June 2026, as the product records it: commercial advertising featuring a synthetic performer needs disclosure, and imagery needs a record.

Published November 14, 2026 · 8 min read

What Is a SHA-256 Fingerprint for an Image and How Does Verification Work?

SHA-256 explained through image files: exact matching, why one changed byte breaks it, and why a registry pairs it with perceptual fingerprints.

Published November 14, 2026 · 8 min read

Perceptual Hashing Explained: How Similar Images Are Matched After Edits

How perceptual fingerprints are computed from pixels, why they survive resizing and re-encoding, and where their matching limits honestly sit.

Published November 14, 2026 · 8 min read

aHash, dHash and pHash: How the Three Perceptual Fingerprints Differ

The three fingerprints ProvenanceGuard computes: average, gradient and DCT-based, what each notices, and why the registry combines all three over 192 bits.

Published November 14, 2026 · 8 min read

How to Verify an Image's Provenance With a Hash or Exhibit ID

The verification paths: public exhibit page, single-hash API lookup, check-a-copy upload, and how to read each part of the result.

Published November 14, 2026 · 8 min read

What Happens to Provenance When Metadata Is Stripped From an Image?

A strip, traced layer by layer: what EXIF, XMP and manifest loss does to the file, the label, the fingerprint and the registry record.

Published November 14, 2026 · 8 min read

Does Instagram Strip Provenance Metadata? A Survival Guide for AI Images

Instagram's simulated profile in the product: resize to 1080px wide, JPEG re-encode at quality 85, metadata stripped, and what survives it.

Published November 14, 2026 · 8 min read

TikTok and AI Content: Labels, Metadata and Provenance That Survives

TikTok's simulated profile, 1080px and quality 80 re-encode, plus the label-first workflow for a platform built on reposting.

Published November 14, 2026 · 8 min read

YouTube Thumbnails and Provenance: What Survives Upload and Re-Encode

The YouTube thumbnail simulation: 1280px wide, quality 90 re-encode, metadata stripped, and sealing the thumbnail as an asset in its own right.

Published November 14, 2026 · 8 min read

Posting AI Images on X: What Happens to Metadata and Labels

X's simulated profile, 1600px wide and quality 85, and why fast-moving reposting makes the registry and label the working layers on X.

Published November 14, 2026 · 8 min read

AI Images on LinkedIn: Disclosure, Metadata and Professional Trust

LinkedIn's simulated profile, 1200px and quality 88, and the professional-context case for labelled, registered AI imagery.

Published November 14, 2026 · 8 min read

Visible AI Labels vs Embedded Metadata: Which One Survives Publishing?

The head-to-head the simulations answer: embedded metadata dies in transport, a pixel label survives what keeps it in frame, and the registry covers both.

Published November 14, 2026 · 8 min read

How to Recover an Image's Provenance After It Has Been Reposted

The recovery workflow: fingerprint the returned copy, read its distance honestly, and reconnect it to the exhibit, creator and registration time.

Published November 14, 2026 · 8 min read

Image Lineage: Building a Chain of Custody for an Edited AI Image

Lineage chains explained: origin hash, local transforms executed on real bytes, declared external steps, and signed links between stages.

Published November 14, 2026 · 8 min read

VERIFIED, UNKNOWN or MISMATCH: Reading an Image Lineage Record

The three lineage states defined strictly: recomputed and matching, declared and never upgraded, and held bytes that fail recomputation.

Published November 14, 2026 · 8 min read

What Is an Evidence Pack for AI-Generated Content?

Inside the pack: exhibit, hashes, fingerprint, tool, timestamps, inspection results and EU, California and New York disclosure wording.

Published November 14, 2026 · 8 min read

Article 50(2) Marking-Conformance Testing: Proving Marking Survives Your Pipeline

The vault battery: real transforms on real sample bytes, marking re-inspected after each, and a signed hash chain of runs as the evidence.

Published November 14, 2026 · 8 min read

Hash-Chained Evidence Packs: Why Each Run Seals the One Before It

How the vault's hash chain works: each pack hash covers the previous pack hash plus the canonical run, HMAC-signed, and tampering breaks recomputation.

Published November 14, 2026 · 8 min read

Bulk Verification: Checking a Library of Image Hashes at Once

The paid bulk-verify API: up to 100 SHA-256 hashes per call, returning registration, exhibit, verification URL and registration time per hash.

Published November 14, 2026 · 8 min read

Content Provenance for Newsrooms Publishing AI-Assisted Images

Newsroom use: sealing AI-assisted imagery at the desk, verification links in captions and credits, and records an editor can defend.

Published November 14, 2026 · 8 min read

Provenance for Brands: Protecting AI Product Images Across Campaigns

Brand use: sealing campaign imagery once, at source, so every market, partner and repost carries the same checkable record.

Published November 14, 2026 · 8 min read

Provenance for Creators: Sealing an AI Portfolio Before Clients Repost It

Creator use: sealing portfolio pieces at export, labelled copies for public posts, and recovery when work travels uncredited.

Published November 14, 2026 · 8 min read

Agencies and AI Imagery: Handing Clients Provenance, Not Just Files

Agency handover practice: exhibit IDs in the delivery note, evidence packs for regulated clients, and bulk verification as the acceptance check.

Published November 14, 2026 · 8 min read

AI Images in Ecommerce Listings: Disclosure and Provenance That Holds Up

Listing imagery: sealing product visuals, labelled copies where imagery is synthetic, and records that answer marketplace and regulator questions.

Published November 14, 2026 · 8 min read

AI-Enhanced Real Estate Images: Labels, Disclosure and Records

Property imagery: virtual staging and enhancement disclosed, sealed and recorded, so a listing's images carry their declarations with them.

Published November 14, 2026 · 8 min read

Submitting AI Images to Stock Libraries: Provenance Records That Help

Contributor practice: sealing before submission, declarations that match the library's, and records spanning every library the same asset reaches.

Published November 14, 2026 · 8 min read

Synthetic Performers in Advertising: Disclosure From Brief to Publish

End-to-end practice for ads featuring generated people: the New York duty, sealing the frames, and wording that survives the media plan.

Published November 14, 2026 · 8 min read

AI Influencer Content: Keeping Provenance When Every Post Is Synthetic

Virtual influencer operations: sealing every post's imagery, consistent labelling as identity, and records across brand deals.

Published November 14, 2026 · 8 min read

Provenance Is Not Deepfake Detection: What Each One Actually Does

The distinction: detection guesses at origin from pixels under uncertainty; provenance records origin at creation and verifies it. Why both exist.

Published November 14, 2026 · 8 min read

How to Prove You Created an AI Image When Someone Disputes It

The dispute playbook: a registration that predates the dispute, fingerprint matching on the contested copy, and an evidence pack instead of screenshots.

Published November 14, 2026 · 8 min read

Your AI Image Was Reposted Without Credit: What Provenance Can Prove

Repost response: match first, document second, message third, with the registry record and its date doing the persuading.

Published November 14, 2026 · 8 min read

Screenshots Destroy Provenance: How to Check a Copy Instead

Why a screenshot has no file provenance at all, what survives inside it visually, and how fingerprint recovery treats the pixels that remain.

Published November 14, 2026 · 8 min read

Cropping and Resizing: How Edits Change an Image Fingerprint

Edit distance intuition: re-encodes and resizes barely move a fingerprint, crops move it more, and composite edits can move it past matching.

Published November 14, 2026 · 8 min read

JPEG vs PNG for Provenance: Where Assertions and Metadata Live in Each Format

Format anatomy: JPEG application segments and XMP, PNG text chunks, how converters treat each, and why inspection reads every territory.

Published November 14, 2026 · 8 min read

XMP Metadata Explained: The Packet Inside Your Image Files

XMP unpacked: an XML packet inside JPEG and PNG, what ProvenanceGuard embeds there, and why platforms' re-encodes erase it.

Published November 14, 2026 · 8 min read

EXIF Data in AI Images: What It Shows and What It Never Proved

EXIF realism: camera fields in generated files, platform stripping, and why EXIF is the weakest provenance territory of the three inspected.

Published November 14, 2026 · 8 min read

PNG Text Chunks: A Quiet Place Provenance Signals Hide

PNG chunk basics: tEXt and iTXt fields where generators write prompts and disclosures, why converters drop them, and checking them on inspection.

Published November 14, 2026 · 8 min read

HMAC-Signed Assertions Explained: A Signature You Can Verify

How the assertion is signed: the payload, HMAC-SHA256 under the product signing key, and what verification does and does not establish.

Published November 14, 2026 · 8 min read

How a Provenance Registry Works: Fingerprints, Exhibits and Public Lookup

Registry mechanics end to end: what is stored at sealing, what is not stored, and how exhibit IDs, hashes and APIs resolve a lookup.

Published November 14, 2026 · 8 min read

The Free Pre-Publish Provenance Check: Five Minutes Before You Post

The exact free routine: inspect the final export, seal it, download both copies, publish the labelled one, file the verification URL.

Published November 14, 2026 · 8 min read

A Provenance Workflow for Small Teams Publishing AI Images Daily

Small-team operating design: one owner, one sheet of exhibit IDs, sealing at export, and paid depth only where the volume justifies it.

Published November 14, 2026 · 8 min read

When Provenance Fails: Diagnosing a Copy That Will Not Verify

The diagnostic order: wrong file, post-seal transform, broken signature, wide fingerprint distance, or never registered, and the fix for each.

Published November 14, 2026 · 8 min read

AI Images in Press Releases and Media Kits: Records Editors Can Check

PR distribution practice: sealing kit imagery, verification links in the kit itself, and why syndicated copies make the registry the record.

Published November 14, 2026 · 8 min read

AI Images in Campaign and Advocacy Material: Disclosure That Travels

Campaign material practice: images designed to be shared by supporters, where captions vanish first, making in-pixel labels and registry records the disclosure.

Published November 14, 2026 · 8 min read

AI Images in Training and Education Material: Keeping Records Straight

Course and training use: generated diagrams and scenarios sealed, labelled for learners, and records that survive slide-deck circulation.

Published November 14, 2026 · 8 min read

Image Provenance as Legal Evidence: What Makes a Record Useful

Evidence qualities: contemporaneity, integrity checks, declared contents, chain discipline, and the line between evidence aid and legal conclusion.

Published November 14, 2026 · 8 min read

Video Thumbnails and Keyframes: Sealing the Still Image Around a Video

The stills around video: thumbnails, posters and keyframes are JPEG and PNG assets that travel as images, and can be sealed and registered as such.

Published November 14, 2026 · 8 min read

Provenance for Cover Art: Podcasts, Books and Releases

Cover art practice: one image, every storefront, endless resizing, and why covers are the archetypal case for registry-backed provenance.

Published November 14, 2026 · 8 min read

AI Edits to Event Photography: Provenance for Hybrid Images

Hybrid work: real photographs, AI retouching or extension, and layered records that say which half is which without diminishing either.

Published November 14, 2026 · 8 min read

Provenance for Social Media Managers: A Multi-Brand Routine That Scales

Manager workflow: many brands, one routine, exhibit IDs per brand sheet, and survival checks as the pre-flight for every platform.

Published November 14, 2026 · 8 min read

Running a Provenance Audit Before a Campaign Launch

The pre-launch audit: inventory, inspection sampling, sealing gaps, label checks at platform sizes, and a bulk verification to close.

Published November 14, 2026 · 8 min read

The 2 December 2026 Marking Grace Deadline: What Existing Systems Must Show

The grace deadline as the product records it: for systems on the market before 2 August 2026, marking duties bite in full on 2 December 2026, and conformance evidence is the preparation.

Published November 14, 2026 · 8 min read

For AI Image Providers: Testing Whether Your Outputs Stay Marked

Provider-side practice: registering system versions in the vault, running the battery on every release, and publishing a chain head customers can check.

Published November 14, 2026 · 8 min read

Provenance Glossary: The Key Terms Behind AI Content Marking

One reference page defining the working vocabulary: assertion, fingerprint, exhibit, survival, lineage, conformance, and the rest, as the product uses them.

Published November 14, 2026 · 8 min read

Provenance for Nonprofits and Public-Sector Communications

Public-trust publishing: AI imagery in public information, records that outlast staff changes, and disclosure as an institutional habit.

Published November 14, 2026 · 8 min read