An alert is a claim
A trading signal proposes an action under certain conditions. To evaluate it, you need the instrument, direction, timestamp, entry rule, validity window, exit or expiry, and risk assumptions. “Buy now” without these details is not a reproducible instruction.
Latency matters. An alert recorded at one price may reach you after that price is no longer available. A short-expiry trade can be especially sensitive to a few seconds. Record receipt time as well as the provider's timestamp and never assume the advertised entry was executable for you.
Request the whole record
Look for all issued signals, including losses, cancellations and periods of silence. Ask whether results use live executions, hypothetical quotes or a demo account. Check whether a losing entry can be edited or removed after publication.
A reported win rate needs a sample size, a period, actual payouts and costs. Even 18 wins out of 25 observations does not establish a stable 72% future probability. The sample may be selected, the market may change, and related signals may not be independent.
Know how the provider is paid
A provider may earn from subscriptions, referrals, deposits or trading activity. These incentives can conflict with your interest in trading less or not at all. A disclosure does not eliminate the conflict, but hiding it makes evaluation harder.
The label AI describes a method or product claim, not an audit. Ask what data are used, when they become available and how errors are handled. A confident explanation can be generated after a prediction; it is not evidence that the prediction had an edge.
Test without making it a race
Define your evaluation method before collecting results. Log every eligible signal, the quote you could actually observe, the offered payout, the decision rule and the settlement. Include missed signals according to a prewritten rule instead of selectively excluding inconvenient ones.
Use a demo or paper record to inspect clarity and execution assumptions. Do not equate a positive trial with proven live profitability. If the signal cannot be independently checked, “insufficient evidence” is a valid conclusion.
Try this on paper
Take one example signal without placing a trade. Make a checklist for asset, timestamp, receipt time, entry, expiry, payout, full loss, quote source and provider incentives. Mark anything missing. Write which claims could be verified and which would remain marketing.