Artificial intelligence intersects with elections in several ways at once: it generates and detects media, it shapes what people see through recommendation systems, and it assists both investigators and bad actors. This guide maps those intersections and sets out an investigative posture that holds up as the technology improves.
Where AI appears in elections
- Generative media — synthetic images, audio, and video, including deepfakes that fabricate statements by candidates.
- Recommendation and ranking — algorithms that decide which content is amplified and to whom.
- Scaled influence operations — cheap, tailored content produced at volume to flood the information space.
- Investigative assistance — clustering, translation, and triage that help analysts work at scale.
The risks
- Fabricated statements attributed to candidates or officials.
- The liar's dividend — dismissing authentic media as "just AI," which corrodes trust in all evidence.
- Volume and personalisation — influence operations that are cheaper and more targeted than ever.
- Procedural confusion — AI-generated content spreading false voting information.
These risks compound the existing threats described in information integrity.
An investigative posture that survives better models
Detector tools are useful but imperfect, and they will keep chasing generative advances. A durable approach prioritises provenance over detection:
- Establish where a piece of media first appeared and who published it.
- Corroborate with metadata, reverse image search, and contextual analysis.
- Preserve everything so authentic evidence can be defended against the liar's dividend.
- Report what is verified and what remains uncertain — without overclaiming "AI."
This is the same discipline set out in how to verify political videos.
Key takeaways
- AI touches elections through generation, ranking, and scale.
- Provenance is more robust than any single AI detector.
- The liar's dividend makes preservation essential.
- Report uncertainty rather than overclaiming fabrication.
Next steps
Practise the method in how to verify political videos, and see how AI-scaled operations are investigated in investigating disinformation campaigns.
Frequently asked questions
How is AI used to influence elections?
AI is used to generate synthetic images, audio, and video (including deepfakes), to power the recommendation systems that decide what content is amplified, and to scale influence operations by producing large volumes of tailored content cheaply. It is also used defensively, to help investigators cluster, translate, and triage content.
Are AI deepfake detectors reliable?
Not on their own. Detectors are probabilistic and can be evaded, so they should be treated as one signal. Establishing the provenance of a piece of media — where it first appeared and who published it — is more robust than any single detector.
What is the biggest AI risk to elections?
Beyond specific fabricated clips, the broader risk is the "liar's dividend" — the ability to dismiss authentic evidence as AI-generated. This erodes trust in all media, which is why preservation and provenance are so important.