OddFlag AI · FREE BETA · A SECOND OPINION, NOT A SAFETY GUARANTEE
OddFlagAIBack to checker

How the check works

Bring a suspicious message, link, selected phone number or screenshot. OddFlag shows a short “Show me the angle” view: what is being asked, quoted pressure or promises, and one practical next step. Missing information remains unknown, and the full review is available beneath the summary. Optional context - who contacted you, what they asked and whether you already acted - helps tailor that guidance.

Three different kinds of evidence

Quoted excerpts come from the text you supplied. Source facts come from the named external service or dated dataset. AI interpretation is separately labelled and may be wrong. One does not become proof simply because another agrees.

Message patterns and AI

Pattern review looks for combinations of code or secret requests, payment pressure, account actions, remote access, secrecy and URL structure. It can also highlight explicit affection, obligation or blame tied to a request for money, secrets or access, and wording that discourages independent checking. Such wording can occur in legitimate situations. A word or emotion alone does not prove fraud or establish anyone’s intent, and the checker does not diagnose relationships. Native pattern checks run on device; web rules run on our service. The rules focus on English. Missing context, unfamiliar languages and code words can be missed. Optional OpenAI review considers chosen context and available evidence, with a separate sharing choice. Web screenshot interpretation uses OpenAI; native apps read selected screenshot text on device and let you review it before any optional text upload.

Link source checks

With your separate online-evidence choice, eligible links are checked against DNS and authoritative domain-registration records. Google Web Risk threat-list results are included only when that service is enabled and available. Each result shows which checks completed, their source and time, and which were unavailable, blocked or not requested. We do not visit the suspicious destination, inspect its page, follow redirects or run its scripts. DNS, registration age and no threat-list match never establish safety.

Selected phone numbers

Number-plan metadata helps interpret the number you deliberately selected. A separately chosen online review can compare it with a bounded, dated public FTC complaint snapshot and a small sourced official-contact directory. The checked number is not sent to the FTC. Coverage is limited, freshness checks can fail and snapshots need operator updates. Reports are unverified; a reported caller ID may have been spoofed. Number format does not prove assignment or reachability, and even a published official-number match does not authenticate an incoming caller. We do not identify private subscribers, dial numbers or block calls.

Three qualified outcomes

High caution means strong warning signs warrant pausing. Caution means details warrant independent checking. Limited evidence means the service cannot decide; it never means safe. Missing or unavailable checks remain visible. Scams can be missed and legitimate messages can be flagged. We claim no measured detection accuracy.

A useful next move

Next steps reflect what was requested and what you already did. They can include pausing a payment, securing an affected account, using a trusted app or number and keeping records. Follow-up answers prepare a new review; they do not submit automatically. We do not contact organizations, file reports or recover money for you. Use independently trusted channels for consequential decisions. Read terms and privacy.