← Framework

Stage 7 · Apply

Synthetic Media Evidence

How to preserve evidence around synthetic, AI-generated, edited, or manipulated media, including source, alteration, identity, context, disclosure, and claim boundaries.

creatorsbusinessesmedia teamsadvisersorganisationsinstitutionsAI teams

Synthetic Media Evidence

Synthetic media evidence helps show how media was created, generated, edited, altered, disclosed, and used.

It applies to AI-generated images, video, audio, text, avatars, voice clones, altered photographs, edited footage, synthetic training outputs, deepfake-like material, generated advertising, entertainment assets, political media, educational content, product imagery, and mixed human-AI media workflows.

Synthetic media creates a special evidence problem.

People may need to prove that media is synthetic. They may need to prove it is not synthetic. They may need to show consent, source, alteration history, disclosure, identity, creation process, or context. They may need to show that a media item was responsibly labelled, that it was not represented as real, or that a public claim about it stayed within the evidence.

The purpose of this guide is to help users preserve synthetic media evidence before confusion, misuse, reputational harm, platform action, regulatory review, or dispute occurs.

Quick Read

  • Synthetic media evidence should preserve how media was created, altered, generated, labelled, reviewed, and used.
  • Strong records keep source files, prompts, model or tool context, edits, versions, metadata, disclosure records, approval notes, and claim boundaries.
  • Synthetic media evidence supports explanation and verification, but does not automatically prove authenticity, consent, legality, authorship, originality, or absence of harm.

What this means

Synthetic media evidence is the evidence position around media whose creation or alteration may matter.

It asks whether someone can explain the media’s source, production path, human involvement, AI involvement, editing history, identity context, disclosure status, and intended use.

For an AI-generated image, this may include prompts, model or tool context, generation settings, source references, output versions, edits, and publication records.

For altered video or audio, it may include original captures, edit logs, project files, export history, consent records, review notes, and disclosure wording.

For public-facing media, it may include approval records, labels, content credentials, publication context, and takedown or correction history.

The evidence should make the media’s status understandable later.

Why it matters

Synthetic media can collapse trust quickly.

A realistic image may be mistaken for a real photograph. A voice clone may be treated as evidence of speech. Edited footage may be presented without context. A generated product image may mislead customers. A political or reputational claim may spread faster than correction. A media file may be labelled as AI-generated without evidence. A person may deny or challenge consent, source, identity, alteration, or publication context.

The evidence problem is not only whether media is fake.

The deeper problem is whether the media’s creation, alteration, disclosure, and use can be explained when questioned.

Synthetic media evidence reduces that risk by preserving the records needed to show what happened, what was claimed, what was disclosed, and what remains uncertain.

What strong synthetic media evidence should include

A stronger synthetic media evidence position usually includes:

  • The media item — the image, video, audio, text, avatar, voice output, project file, export, or mixed-media asset.
  • The media claim — what is being said about the media’s source, authenticity, synthetic status, alteration, consent, or use.
  • Source material — original captures, reference files, scripts, recordings, datasets, project files, prompts, or supplied materials.
  • Creation process — how the media was generated, edited, composited, transformed, rendered, or exported.
  • AI-use context — prompts, model or tool context, settings, seeds, versions, account or workspace context, where relevant and available.
  • Human contribution context — selections, edits, approvals, reviews, corrections, and creative decisions.
  • Alteration history — changes made to the media, including edits, retouches, cuts, voice changes, overlays, substitutions, or composites.
  • Identity context — whether real people, voices, likenesses, organisations, brands, locations, or events are represented.
  • Consent or authority context — permissions, releases, licences, approvals, or lawful basis relied on where relevant.
  • Metadata and provenance context — content credentials, embedded metadata, file history, export records, or external references.
  • Disclosure context — whether the media was labelled, described, watermarked, disclosed, or restricted.
  • Publication context — when, where, and how the media was shared, delivered, uploaded, or used.
  • Custody and retention context — where source files and supporting evidence are preserved.
  • Verification route — how someone could check the media evidence later.
  • Claim boundaries — what the evidence supports and what it does not support.

The more realistic, public, sensitive, or high-impact the media is, the stronger the evidence record should be.

Common weak points

Synthetic media evidence is usually weak when:

  • only the final output is preserved
  • source files or original captures are missing
  • prompts or generation context are lost
  • edits are made without version history
  • model or tool context is unknown
  • disclosure wording is missing or vague
  • consent or authority records are not preserved
  • content credentials are assumed to prove more than they show
  • metadata is stripped during export or publication
  • a realistic synthetic item is published without context
  • a real item is labelled synthetic without supporting evidence
  • publication records are separated from the media file
  • project files are overwritten
  • public claims say “verified”, “authentic”, “real”, “fake”, “AI-generated”, or “not AI-generated” without a clear evidence basis
  • private or sensitive media is shared unnecessarily during verification
  • EviWrite verification is implied where none exists

These weaknesses make later explanation harder and can increase reputational, legal, platform, or public-trust risk.

How to apply this yourself

For each important synthetic or potentially synthetic media item, create a synthetic media evidence note.

Ask:

  • What media item is being evidenced?
  • What claim are we making about it?
  • Is the claim about source, authenticity, alteration, synthetic status, consent, disclosure, or use?
  • What source files, original captures, prompts, project files, recordings, or references exist?
  • What AI tools, editing tools, models, settings, or workflows were used, where relevant?
  • What human edits, selections, reviews, approvals, or corrections occurred?
  • Are real people, voices, likenesses, brands, places, or events represented?
  • What permission, authority, release, licence, or disclosure context matters?
  • What metadata, content credentials, or provenance records exist?
  • Where and when was the media published, delivered, uploaded, or used?
  • Can the evidence be checked later without unnecessary exposure?
  • What does the evidence not prove?

Then preserve the media, source material, creation records, disclosure records, publication context, custody notes, verification route, and claim boundary together.

Do not rely on final exports alone.

What this does not prove

Synthetic media evidence does not automatically prove:

  • authenticity
  • consent
  • legality
  • authorship
  • ownership
  • originality
  • non-infringement
  • absence of manipulation
  • absence of AI use
  • presence of AI use unless the evidence supports that claim
  • absence of reputational harm
  • accuracy of what the media depicts
  • identity of every person or system involved
  • that content credentials are complete or unbroken
  • that EviWrite has verified or backed the record

Synthetic media evidence helps explain creation, alteration, disclosure, and use. It does not settle every legal, factual, or ethical issue.

Framework-aligned claim boundary

A person or organisation may use this guide as part of EviWrite Framework alignment if they apply the guidance honestly and avoid implying EviWrite involvement.

Acceptable wording may include:

“We use the EviWrite Framework to preserve evidence around synthetic and altered media.”

It must not be used to imply:

  • EviWrite has verified the media
  • EviWrite has confirmed authenticity
  • EviWrite has confirmed consent
  • EviWrite has confirmed legality
  • EviWrite has confirmed whether the media is or is not synthetic
  • EviWrite has approved the publication or use
  • the record is EviWrite-backed
  • the record is EviWrite-certified
  • the record carries the controlled ⓔ mark

Framework-aligned means public guidance was followed.

EviWrite-backed means the record was created through EviWrite or an authorised evidencing channel.

Related checklist

Use the Synthetic Media Evidence Checklist to check whether source files, prompts, edits, tool context, consent records, disclosure wording, publication context, verification routes, and claim boundaries have been preserved clearly.

This guide is public evidence-readiness guidance. It does not mean EviWrite has verified, certified, approved, anchored, or backed any record.

Return to Framework map