AI Fake News Claim: How to Check the Claim, Image, or Screenshot
About this check
Start with the exact claim
If you have just seen an AI-generated image, a video, or a screenshot attached to a news claim, begin by separating the thing you saw from what it is supposed to prove.
Write down the claim in one sentence. Include the person, place, event, date, number, or quote that matters. Then ask what a finished check should contain: a verdict, the evidence behind it, and the sources that support that verdict. A useful answer may be true, false, or inconclusive. “Inconclusive” is a valid result when the available evidence cannot settle the claim.
An AI-looking image is not, by itself, proof that the accompanying claim is false. It is one part of the material to examine. The claim still needs to be checked against evidence.
The manual method
1. Preserve the original material
Keep the image, video, or screenshot as you received it. Record the post’s wording, account, date, link, and any caption. If the material was forwarded, note that as well. Do not rely on memory or on a shortened version of the claim.
For a screenshot, copy the visible text exactly. Include headlines, names, dates, logos, and unusual wording. The text may be more useful for checking the claim than the image itself.
2. Break the claim into checkable parts
A single post can contain several claims. Separate them. For example, “This image shows an event that happened yesterday” includes a claim about the event, a claim about the date, and a claim about what the image shows.
Check each part rather than letting a plausible detail settle the entire post. Also distinguish a statement of fact from commentary, prediction, satire, or a question. They require different kinds of support.
3. Search the wording and key details
Search the exact quote or distinctive phrase in quotation marks, then search the important names, places, and dates separately. Look for the earliest available version of the claim and for reporting that identifies where the information came from.
Compare the wording, publication date, and context. A result that repeats the same post is not automatically independent support. If the claim concerns a developing event, note which parts are directly reported and which parts are interpretation.
4. Check the image or video in context
Look for the original upload, an earlier appearance, a caption, and a date. Compare the visible details with the claim: location, weather, clothing, signs, landmarks, and what the post says is happening.
For a screenshot, search text from the screenshot and inspect the page or account it supposedly came from. A screenshot can show what someone captured, but it does not by itself establish the larger story attached to it.
For an AI-generated image or video, inspect it as evidence rather than treating its appearance as the verdict. The question is still whether the claim is supported. A visual can be synthetic and paired with a true statement, or look ordinary and be used with a false one.
5. Compare sources, then stop where the evidence stops
Prefer sources that provide specific, checkable information. Compare independent accounts when they are available, and pay attention to corrections, missing context, and changed dates. Keep a record of the sources that support each part of your conclusion.
Manual checking has a boundary: search results can repeat an unsupported claim, the original source may be unavailable, and an image or video may not answer the question being asked. When the evidence does not settle the claim, the accurate result is not a forced yes or no. It is inconclusive.
What VOM checks
VOM checks claims using web search, can reuse a closely matching earlier check when a claim is not time-sensitive, and shows the sources it found. Its answer includes a verdict, a confidence score, and those sources. The verdict can be true, false, or inconclusive.
You can submit a claim by voice. VOM also reads text inside a screenshot, so you can submit a forwarded image without retyping its contents. When an uploaded image appears AI-generated, VOM warns you. Separately, VOM independently reviews an AI-generated image or video before approving it for publication.
For article sharing, VOM’s pipeline still publishes if its AI image or video generation is blocked, fails, or is exhausted. Without an account, VOM allows five AI actions per UTC day, shared across questions, verifications, and image generation.