Verifiable facts
Statements about events, measurements, dates, records and other claims that can be examined against evidence.
Our Method
Understand the facts, the context, and the story being told.
TruthBubble separates checkable factual claims from framing, inference and tone. Each assessment explains the evidence available and the limits of what it can establish.
Website checker scope
This page describes TruthBubble's intended assessment framework. The currently available website checker only summarizes ratings from related public fact-check reviews; it does not run the separate framing, inference, tone, safety, or confidence assessments described below.
Start with the claim
A post may contain several factual assertions, opinions, jokes and interpretations. We identify material assertions that can be checked, then assess each against relevant evidence. Subjective opinions are not classified as objectively true or false. “Not fact-checkable” is a separate intake category for content with no material verifiable factual assertion; it is not an eighth factual verdict.
Framing and interpretation
How verifiable facts are selected, compared, emphasized, omitted or used to suggest conclusions.
Beliefs and value judgments
Moral, religious, philosophical or subjective positions that evidence may inform but cannot always conclusively resolve.
01 · Factual verdict
Seven official overall factual verdicts
These public labels correspond to the official internal enum codes shown below. They describe the material factual proposition, not a creator’s overall character or reliability. The factual verdict remains separate from framing, inference, tone and safety assessments.
Supported
SUPPORTEDThe central factual claim is supported by strong available evidence.
Mostly Supported
MOSTLY_SUPPORTEDThe main claim holds up, although some details need correction or qualification.
Partly Supported
PARTLY_SUPPORTEDSome important parts are supported, while others are incorrect, unsupported or unresolved.
Misleading
MISLEADINGThe presentation creates a materially inaccurate impression through omissions, implications or selective context.
False
FALSEStrong evidence contradicts the central factual claim.
Unsubstantiated
UNSUBSTANTIATEDThe available investigation does not establish sufficient support for the claim.
Insufficient Evidence
INSUFFICIENT_EVIDENCEThe evidence or research coverage is not sufficient for a dependable conclusion.
Not fact-checkable · intake category
Used when content has no material verifiable factual assertion. It is not an overall verdict.
Confidence · High, Medium or Low
Confidence describes the strength and coverage of the assessment. It is not a percentage chance that a post is true.
02 · Framing and context
True details can still leave the wrong impression
A presentation may omit important context. Selective emphasis by itself does not establish misinformation. The relevant question is whether the presentation causes a substantively inaccurate impression.
ACCURATE_CONTEXT
Accurate context
The presentation gives the context needed to understand the claim.
MINOR_CONTEXT_MISSING
Minor context missing
Some useful detail is absent, but it does not materially change the takeaway.
SELECTIVELY_FRAMED
Selectively framed
The post emphasizes selected information; that alone does not establish a false or misleading claim.
MATERIALLY_MISLEADING
Materially misleading
The framing creates a substantively inaccurate impression.
CANNOT_ASSESS_FRAMING
Cannot assess framing
The available content does not provide enough context to assess its presentation.
Illustrative example
Example: a true number, missing denominator
A post accurately reports that cases doubled, but omits that the count rose from one to two. The number may be supported while the missing baseline changes the impression. Selective emphasis alone is not enough; the review asks whether the omission creates a materially inaccurate takeaway.
Illustrative example
Example: a plausible explanation
A change happens after a new policy. The timing may make a connection plausible, but timing alone does not establish that the policy caused the change. A reasonable hypothesis is not automatically misinformation.
03 · Inference and reasoning
Evidence and inference are different
“Not definitively proven” does not mean “impossible” or “false.” An interpretation can be reasonable without conclusive proof. We consider the creator’s actual level of certainty, available evidence, relevant alternative explanations and whether the conclusion is proportionate to that evidence.
WELL_SUPPORTED
Well supported
The inference is proportionate to the available evidence.
PLAUSIBLE_BUT_UNPROVEN
Plausible but unproven
The interpretation is reasonable, but the evidence does not establish it.
SPECULATIVE
Speculative
The conclusion goes beyond what the available evidence can support.
CONTRADICTED
Contradicted
Relevant evidence weighs against the inference.
NOT_APPLICABLE
Not applicable
The content does not make a material inference to assess.
04 · Tone and communication
Describe the message, not presumed intent
Tone labels describe communication. They are not definitive findings about a person’s intentions. Passionate arguments, satire, legitimate criticism and emotional reporting are not inherently misleading.
- Informative
- Analytical
- Persuasive
- Satirical
- Emotional
- Alarmist
- Outrage-driven
- Derogatory
- Hostile
05 · Safety signals
Separate safety from factual accuracy
Harm and incitement assessments are separate from factual verdicts. A serious assessment requires specific evidence, careful attribution to the speaker and safeguards against confusing quoted violence with endorsed violence.
TruthBubble does not present automated accusations of dangerous intent as established facts.
Evidence & Sources
Show what supports the assessment
Sources examined
Where source information is available, identify the publisher, review title, original link, the claim it examined and relevant limitations. A link makes material inspectable; it is not an endorsement or proof by itself.
Supporting and contradicting evidence
A careful assessment should consider evidence that supports and challenges the specific factual claim. The public website search preserves each published review’s original rating instead of combining ratings into a new TruthBubble finding.
What's still uncertain
Sources can be unavailable, outdated or conflicting, and research coverage can be incomplete. Insufficient evidence means there is not enough evidence or coverage for a dependable conclusion; it does not prove a claim false.
Media coverage has limits
Recent, local, paywalled, unpublished, or poorly indexed material may not appear in a search. A lack of coverage is a limit on what can be found, not proof that a claim is false.
AI-assisted automation and human review
AI-assisted automation contributes materially to research and analysis in the separate Unified V2.1 workflow used by supported Android private-beta flows. Automated work can miss evidence, misread a source or overstate a conclusion. Review the sources and limitations rather than treating an output as an authority.
The public website checker only searches for related published reviews; it does not run Unified V2.1 or assign an independent TruthBubble verdict. This website does not document a per-result human review process for Android beta assessments, so it does not claim every assessment is manually checked or approved.
High, Medium and Low confidence describe the strength and coverage of an independent assessment, not a numeric probability. Stronger, more directly relevant and consistent evidence may support higher confidence; important gaps, disagreement or limited coverage lower confidence. No confidence label is returned by the public website search.
Corrections and new evidence
You can email a result and relevant evidence through our general contact page. There is no dedicated correction or appeal workflow, tracking process, or response-time commitment documented here. Sending an email does not guarantee review or a change.
Verdict example
See how a structured result reads
Demonstration · fictional contentA fictional post about a town park“This clip proves the town park has been closed all month.” Open example
Overall factual verdict
INSUFFICIENT_EVIDENCE · Low confidence
Central takeaway: a short clip alone cannot establish when the park was closed or for how long.
What checks out
In this fictional scenario, the clip appears to show a barrier at a park entrance.
What needs context
The clip has no confirmed location or date.
What does not hold up
The statement that it has been closed “all month” is not established by the clip alone.
Framing assessment
CANNOT_ASSESS_FRAMING · There is not enough context to judge the presentation.
Inference assessment
PLAUSIBLE_BUT_UNPROVEN · A closure is possible, but timing and duration are not established.
Tone
Informative · A description of communication, not a finding about the author’s intent.
Source links
No real source links are included in this fictional demonstration.
Remaining uncertainties
The location, date and relevant closure records are unknown in this example.
Assessment is not a promise of certainty
Evidence may be unavailable, outdated or conflicting. Coverage can be incomplete, and new information may change a result. For high-stakes decisions, review original sources and consult qualified expertise.
Read about limitations →