TRUE
Available evidence supports the material claim.
METHODOLOGY AND CORRECTIONS
How VOM answers—and where its limits begin.
VOM combines general AI conversation with current-information research and claim verification. It does not search the web for every message, and no result should be treated as a final authority.
REVIEWED:
Each request follows a route
VOM automatically classifies what you ask. That enables the appropriate response flow, but automated routing can make mistakes.
- 01
Normal conversation
For explanations, writing, translation, planning, or brainstorming, VOM behaves like a conversational AI. This route does not use live web search and can be incorrect or out of date.
- 02
Current question
When an answer depends on recent or changing facts, VOM attempts web research. It may show sources, but it does not assign a verdict or confidence score.
- 03
Verifiable claim
VOM gathers web research, then evaluates the claim in a separate step. The result may include a verdict, explanation, confidence score, and available links.
- 04
Protected request
Private-life allegations and requests for sensitive personal data do not receive a verdict. VOM responds with privacy limits instead of building a personal dossier.
What the verdicts mean
A verdict evaluates the material claim against available evidence. It does not decide whether the speaker intended to deceive.
FALSE
Available evidence contradicts the material claim.
UNPROVEN
Evidence is insufficient or conflicting, or the claim cannot support a responsible conclusion as written.
Sources support inspection, not guarantees
Source links come from grounding information returned during web research. A link identifies material considered by the system; it is not an endorsement of the source or proof that every sentence is supported by it.
VOM does not guarantee a fixed number of sources, a citation for every sentence, independent corroboration, or a clickable link in every result. Coverage can be incomplete, and sources can disagree or change.
A non-time-sensitive claim may reuse a closely matching earlier verification. The automatic bypass currently recognizes selected time markers in English, Spanish, Portuguese, and French; other languages may still reuse an earlier match.
Confidence is not certainty
The confidence score is a model-generated assessment of how strongly the research supports the displayed verdict. It is not a statistical probability, an independently calibrated accuracy rate, or a guarantee.
A score of 90 does not mean there is a measured 90% chance that a claim is true. Read the explanation and inspect the sources; scores should not be compared directly across subjects or languages.
Private-person safeguards
VOM does not assign truth labels to claims about a private person's relationships, health, sensitive traits, whereabouts, or other private life. It also does not provide home addresses, personal contact details, live tracking, or identity resolution from a phone number, email, or photograph.
Public professional information may be researched when it is already publicly supported. Ambiguous names should be presented as possible matches, not confirmed identities.
A protected request is processed to generate a response, but its readable question and answer are replaced by placeholders in saved history. One-way hashes and limited metadata may still be retained for integrity and aggregate measurement.
Eight languages, with visible limits
VOM supports Spanish, Brazilian Portuguese, English, French, German, Arabic, Chinese, and Japanese. The latest message normally determines the response language; an explicit request can override it.
Names, URLs, and source titles may remain in their original language. Automated detection and rewriting can fail, and VOM has not published independently validated accuracy benchmarks for each language.
How to report a suspected error
Answers are generated automatically and are not reviewed by a human before they appear. VOM does not yet offer an in-product appeal, edit historical generated results in place, or maintain a public correction log.
To report a suspected error, send the exact claim, approximate date and time, a screenshot of the result, what you believe is wrong, and links to relevant evidence. A report does not guarantee a particular outcome or response time.
Important limitations
VOM can misunderstand a request, select weak or incomplete sources, miss relevant evidence, misread dates, translate imperfectly, or generate an incorrect explanation even when web research is used.
Grounded research reduces some errors; it does not eliminate them. For medical, legal, financial, safety-critical, or other consequential decisions, consult qualified professionals and authoritative primary sources.
Consistency, not absolute neutrality
VOM is designed to apply consistent, evidence-focused rules and distinguish documented facts from contested characterization. Models, instructions, and the available source landscape can still contain bias.
Public frameworks informing this transparency
These resources help structure this disclosure. VOM does not claim certification, audit, or affiliation with these organizations.
Inspect the evidence. Keep your judgment.
Use VOM to talk, research, and verify, but treat every result as a starting point you can examine.
Try VOM