Check whether a circulating tweet, article or chat screenshot is real. Forging one takes seconds with browser dev tools.
Free with no sign-up, and the analysis runs inside your browser — files never leave your device. On our public gate (30 real + 83 AI images) it measured 94.7% at original quality and 87.6% on recompressed copies, with zero real photos confidently mislabelled as AI. Treat it as an indicator, not proof: the newest commercial generators evade every detector.
A tweet, article or chat "screenshot" is the cheapest forgery there is — edit the text in a browser, capture, done. The key check is not image analysis but whether the original actually exists.
Find the original on the real account or outlet; if deleted, web archives often keep a trace. Reverse-search the capture to see if it was used for a different event before. Then run the pixel check. This site's editing-trace (ELA) tool can show whether the text region was recompressed differently from the rest.
Number contradictions are great clues too — more shares than views, a post older than the account, timestamps in impossible time zones. Forgers rarely get the details right.
Screenshot forgery is common because it needs no skill. Open the browser developer tools and the text on screen can be edited in place; capture that and the image is indistinguishable from the original. It takes seconds and no special software.
Messenger conversations are easier still. Services exist purely to mock up chat windows, with names, timestamps and read receipts all configurable. Treat a conversation capture as proving nothing on its own.
The weight of judgement therefore sits on finding the original, not on analysing the image. The pixel check here offers supporting clues: a text region re-compressed differently from its surroundings, for instance. Confirm the original exists first; look at pixels second.
What forgers fail to keep straight is the surroundings rather than the content. Does the status bar clock match the claimed posting time? Are battery and carrier indicators consistent across the capture? Does the crop match the aspect ratio of the device it supposedly came from?
Interface versions are a clue too. A recently redesigned app appearing in its old layout while showing a new feature, or dark and light mode mixed inside a single capture, points to pieces stitched together.
Numbers contradict themselves often. Shares exceeding views, a post dated before the account existed, reply counts that disagree with what is displayed. Concentrating on rewriting the sentences leaves those values untouched.
How far a forged capture travels depends on the people forwarding it rather than the person who made it. Posts circulating as an image with no link to the original are the dangerous kind — no link means the means of checking was removed along the way.
One question covers it: where is the original? Ask whoever sent it for a link, and if there is none, search the account or outlet yourself. It takes seconds and most forgeries do not survive it.
If you already shared something that turns out to be fake, a correction beats quiet deletion. Deleting reaches nobody who already saw it. Posting what was wrong, in the same place, is what actually slows the spread.
A capture resembling an article from a well-known outlet circulated, with only the headline and opening paragraph carrying an inflammatory claim. Fonts and layout matched the outlet exactly. The pixel check finds nothing at all. It is a fresh capture of a manipulated screen, so there is no editing history. This type resolves in seconds by searching the outlet's own site for the article.
A screenshot of an inflammatory post circulated, but the text appeared nowhere on the account, in search, or in archives — and letter spacing and icon placement in the capture differed subtly from the real interface. The weight of the judgement sits on whether the original exists, not on image forensics. Editing text in browser dev tools leaves no pixel trace at all, so finding the source is the only conclusive check.
The screenshot came with a link to the original post, and an archive service held a snapshot from the same time. Here the capture is a citation, not evidence. Even if the original is deleted later, the archived snapshot preserves the fact that it existed.
Posts that were published and then removed often survive in neither search nor archives, producing the same "no original" result as a fabrication. Absence of the original is unverifiable, not fake. Rather than declaring it either way, look for independent sources — other witnesses, reporting, replies quoting it.
"Captured before deletion" is also the standard forgery excuse. Without independent secondary evidence — web archives, press quotes, other witnesses — it is safest to treat it as unconfirmed.
No — ELA is an observational tool that shows recompression differences and reacts to normal edits too. Its value grows when it stacks with a missing original and number contradictions.
No. Manipulate a screen and capture it fresh and there is no editing trace, because the image is a genuine capture of a forged screen. Nothing found means no information, not authenticity. Finding the original always comes first.
Yes. Cropping part of a genuine post to invert its meaning, or stitching separate posts into what looks like one conversation, leaves no editing trace. Once you find the original, compare it for trimmed sentences before and after, and check that the sequence holds.