How the analysis works
Every meme is evaluated with the same structured method, adapted from the Adversary Amplification framework and its “Meme as a Weapon” cognitive-exploitation model. The AI first reads the image and extracts the central claim the meme makes or implies — then scores it across the dimensions below. The goal is informed awareness, not a verdict from an oracle.
- true — The claim is accurate and fairly framed.
- mostly true — The core claim is accurate, but the framing omits context.
- misleading — Real elements are used to distort meaning — correlation dressed as causation, cherry-picked data, stripped context.
- false — The claim is factually wrong or fabricated.
- unverifiable — Cannot be assessed from the image alone and no sourcing is provided.
- humor — Pure comedy — no serious factual claim is being made.
- satire — Exaggeration or irony used to comment on a subject; not meant to be taken literally.
A meme’s accuracy and its purpose are different questions. Comedy and satire are legitimate formats — being funny does not make a meme dangerous.
- humor — Intended purely to be funny — no narrative, grievance, or factual claim being pushed.
- satire — Commentary through exaggeration or irony, typically political or cultural.
- amplification — Promotes a narrative beyond comedy or satire — a shareable claim meant to be taken seriously.
Three vectors from the framework are checked on every meme:
- Statistical illiteracy — Correlation dressed as causation, decontextualized numbers, misleading charts.
- Emotional identity triggers — Stacks patriotism, sympathy, fear, or outrage with an antagonist — often with a share or purchase prompt.
- Domain complexity — An alarming claim about a technical subject the viewer cannot independently evaluate.
- Dual-process theory — The meme forces a fast emotional reaction before analysis can begin.
- Truthiness effect — An image or chart makes the claim feel true, independent of the facts.
- Confirmation bias — The meme flatters a belief the viewer already holds.
- Statistical illiteracy — The viewer cannot evaluate the numbers being presented.
The framework’s pre-share checklist, applied specifically to the meme being analyzed:
- Does this make me feel strongly before I’ve evaluated it?
- If it shows a correlation, does it show a mechanism?
- Who benefits from me sharing this?
- Where did this originate?
- Does this confirm what I already believe?
Is there a gap between what the meme urges and what its likely origin benefits from — for example, an actor opposing something abroad while subsidizing it at home? The analysis reports that gap when one is visible.
Traceable — a real source or author is cited · Partial — a source is named but cannot be verified · Untraceable — the origin is concealed.
Each analysis reports verdict confidence (how sure the assessment is given what is visible) and an emotional manipulation score (how much the meme relies on emotional override of analysis), both on a 0–100 scale.