AI Misuse, Fraud & Deepfakes
How to Identify AI-Generated Fake Photos, Videos and News
हिंदी में पढ़ेंTo check a suspicious photo, video or news claim, investigate where it came from and whether reliable sources confirm it. Visual clues alone cannot reliably tell you whether AI was used. A real photograph can also be old, cropped or attached to a false story.
A message claims a picture shows an incident near a local school today. The picture looks believable, but an image search finds the same scene in a much older report from another place. The problem was the caption, whether or not AI was involved.
What should we check first?
Read the whole claim. Look for the original publisher, date, location and evidence. A forwarded message, anonymous screenshot or “everyone is sharing this” is not an original source.
Open the official organisation's website or use a known contact. For a school announcement, use the school's published notice channel. Do not follow a new payment or login link merely because the message carries a familiar logo.
How can reverse image search help?
For an ordinary public image, tools such as Google image search can help find related images and pages. Compare publication dates and contexts. Google documents how to search using an image; results are clues, not a certificate of authenticity. AI-S18: Read the source
Do not upload private child photos, identity documents or nude or sexual imagery involving children to search tools. For harmful content, preserve safe details such as URLs and account names and seek appropriate reporting help.
Can an AI detector or another chatbot decide?
Do not treat a detector score as proof. Unusual fingers, text or shadows can raise questions, but authentic media can contain odd details too. Good-quality fabricated content may have no obvious flaw.
Another chatbot may repeat the same false claim or invent a reference. OpenAI's accuracy guidance warns that AI can give incorrect information and fabricated citations. Open and read sources rather than trusting the existence of a link. AI-S06: Read the source
What if we still cannot confirm it?
Say, “I could not verify this,” and avoid forwarding it as fact. Ask the sender for an original source without accusing them. For an urgent safety issue, contact the relevant official service rather than waiting for a group chat to decide.
Try a family exercise with a harmless public news picture: identify its source, date and claim. Ask what would change your mind. The aim is not to win an argument or label every unusual image fake; it is to make a careful decision before believing, paying or sharing.
Where can you check the details?
What can you read next?
Want to check the sources?
These notes explain what was checked for the article. The date is an information check, not a publication date.
- AI-S18: Search with an image on Google
Google
Where to look: Image search and finding related sources
How it was checked: Official text retrieved
Reverse search is a clue, not proof. Do not upload private child images or sexual content involving children to search tools.
- AI-S06: Does ChatGPT tell the truth?
OpenAI
Where to look: Errors, fabricated citations, tools and verification
How it was checked: Official text retrieved
Confidence and citations are not proof. Current tools can provide web access; avoid outdated claims that ChatGPT never searches.
Next planned check: