SCAIO Learn · Primer 04

Spotting synthetic media.

A plain-language primer on deepfakes, voice cloning, and AI-generated content — for voters, journalists, civic groups, and anyone who shares what they see online.

SCAIO · scaio.org

What we're talking about

Three categories of AI-generated content worth knowing.

01 · Deepfake video

Faces swapped, mouths moved

AI-generated or AI-edited video that depicts a real person saying or doing something they did not. Quality has improved sharply in the last two years.

02 · Voice cloning

A few seconds of audio is enough

From a short sample of public audio (a speech, an interview, a voicemail), AI can produce convincing speech in that person's voice saying anything.

03 · AI-generated text and imagery

From scratch, at scale

Convincing news stories, social-media posts, fake images, fabricated documents, doctored screenshots. The volume problem more than the quality problem.

The honest disclaimer

You cannot reliably spot all synthetic media by eye anymore.

The traditional tells — strange hands, off lighting, robotic intonation, telltale phrasing — are disappearing fast. The technology that produces synthetic media improves faster than the human eye learns to detect it.

That does not mean you are helpless. It means the response shifts from can I tell? to can I verify?

Still useful (sometimes)

Visual and audio cues that can flag synthetic content.

Inconsistent lighting and reflections. Eye reflections, shadows that don't match, jewelry or glasses that warp between frames.
Mouth/voice mismatch. Lips that don't quite sync, or audio that loops or breathes oddly.
Hands and edges. Hands with the wrong number of fingers; hair, glasses, or jewelry that "melt" at the boundary.
Compression artifacts in suspicious places. A face that looks higher-resolution than the rest of the frame.
Voice that lacks emotional range. AI voices are improving but often still flatter than the real person's typical speech.

The verification posture

Three habits that work better than trying to spot fakes by eye.

Pause before sharing. Synthetic media spreads because the first response is emotional. The simple act of waiting a few minutes lets verification catch up.
Verify at the source. If the content depicts a public figure or institution, does the original outlet confirm it? If a video shows a candidate making a statement, is it on their own channels?
Triangulate. Three independent, reputable sources reporting the same event is the floor. One screenshot is not.

In a synthetic-media environment, the question stops being 'is this real?' and starts being 'who is reporting it, what is their source, and is the story holding up across independent outlets?'"

A working posture for civic life online

Where this hits South Carolina hardest

Three categories of risk worth planning for.

01

Election-cycle risk

Synthetic content depicting candidates or officials

Voice or video of a SC candidate saying something they did not. AI-fabricated content about voting procedures. The cleanest defense is a clear voter-information channel that voters already trust.

02

Wire fraud

Voice-cloning impersonation of an executive

A cloned voice of a school superintendent, hospital CFO, or municipal official requesting an urgent wire or beneficiary change. Public-sector loss reports have measurably risen.

03

Reputation and crisis

Fabricated content about institutions

A fake "internal memo," doctored screenshot, or AI-generated statement attributed to an SC institution. Prepared communications templates and a fast verification path matter.

If you see suspicious political content

Who to contact in South Carolina.

Elections

SC State Election Commission

The state-level agency for elections administration. Channel for content about voting procedures, polling-place rumors, or election-process misinformation.

Press

SC Press Association

The statewide member organization for SC news outlets. A useful contact path for journalists trying to verify content.

Law enforcement

SLED · FBI Columbia field office

For cases that may involve fraud, identity misappropriation, or coordinated activity beyond ordinary political speech.

Platforms

The platform where it was posted

Every major platform has a reporting flow for synthetic-media content. Use it. Reports are how platforms learn what is circulating.

What would help, statewide

Three constructive moves SC is positioned to make.

Plain-language voter-education material on synthetic media, distributed through county libraries, civic associations, faith communities, and the SC Press Association well in advance of major election cycles.
Voluntary candidate-disclosure standards on AI-generated content in campaign communications, developed collaboratively by the State Election Commission and the state's two major political parties.
A defined contact path for journalists and campaigns to flag suspected synthetic content for review — making the verification work faster and more legible.

The bigger pattern

Synthetic media is a verification challenge, not a perception challenge.

The trustworthiness of what circulates online has always depended on the trustworthiness of the outlets carrying it. Synthetic media raises the stakes; it does not change the underlying logic.

What protects South Carolina is the same thing that has always protected functioning civic life: credible local news, plain-language guidance from trusted institutions, civic habits of checking before sharing, and a healthy press ecosystem with the capacity to verify on a tight timeline.

SCAIO Learn

Public-interest AI research for South Carolina.

SCAIO tracks AI's impact on South Carolina — including AI-driven risks to public discourse, election integrity, and institutional trust. Browse more primers at scaio.org or read the cybersecurity Journal article on AI-driven public risk.

scaio.org · jimmy@scaio.org