How to decide whether content may be AI-generated

Learn how to investigate suspicious text, pictures, video and audio without relying on a single clue or detection tool.

Easy tutorial · About 15 minutes

Important limitation

You usually cannot prove that something was made with AI just by looking at it.

Human-made content can look artificial, and AI-generated content can look convincing. Modern AI models are so good at generating convincing content that it is often hard to tell the difference.

Step 1Check where it came from

Start with the source rather than the appearance. Find the earliest version you can and ask:

• Who first published it?
• Is it on that person or organisation's official website?
• Does the account have a trustworthy history?
• Is the date and surrounding context clear?
• Could the post be a copy with its original explanation removed?

A familiar logo, realistic username or large follower count is not proof that an account is genuine.

Step 2Investigate suspicious details

Look for details that can be checked. Do not use a writing style, unusual hand or strange background as your only evidence.

For text

• check names, dates, quotations and links
• look for sources that do not exist or do not support the claim
• notice contradictions between different parts of the text

For audio

• listen for abrupt changes or words that do not fit the context
• confirm surprising requests using a known phone number
• do not trust a voice merely because it sounds familiar

For pictures and video

• zoom in on signs, labels and other details you can verify
• check whether shadows, reflections and movement remain consistent
• remember that editing and compression can also create odd details

Example: AI-generated video changes quickly

These two AI-generated videos use a similar prompt but were made with different generations of technology. The earlier example has obvious distortions. The newer example is more convincing, showing why yesterday's visual warning signs may not work today.

An earlier AI-generated example with more visible distortions. Source: Wikimedia Commons.

A newer AI-generated example made with Google Veo 3. Source: Wikimedia Commons.

Even in the newer example, there are some strange details that suggest it may not be genuine. For example, the audio makes the spaghetti sound "crunchy". This is an example of an inconsistency that might be a sign of AI generated content.

The first video above is from 2023. The newer video is from 2025. This shows how quickly AI technology can improve.

Step 3Find independent evidence

Search for the event, quotation or image using sources that did not simply copy one another.

• check official records or the organisation's own website
• look for reporting from more than one established source
• use reverse-image search to look for earlier versions
• compare the claim with the complete video, speech or document

File history and content credentials can provide useful evidence when present. Their absence does not prove that something is fake.

Step 4Treat AI detectors cautiously

Tools that claim to detect AI-generated writing, images or audio can produce false positives and false negatives. Their result should never be the only basis for an accusation or important decision.

Step 5Decide what to do

You do not always need a final verdict. If content asks for money, personal details or urgent action, pause and verify the request through a separate, trusted route.

• do not share uncertain content as fact
• contact the person using a number or address you already trust
• report impersonation or fraud to the relevant service
• describe uncertainty instead of making an unsupported accusation

Quick checklist

1. Check the original source.
2. Verify specific details.
3. Find independent evidence.
4. Do not rely on an AI detector alone.
5. Pause before sharing, paying or providing information.

Back to all tutorials