Source Policy and Evidence Verification
A source must support the specific claim beside it. A contract response can show a market, but not a resident’s eligibility; a screenshot can show a setting, but not a completed withdrawal.
When official pages appear to disagree
First compare product, account, region and effective date. A difference may be legitimate rather than an error. Paradex, for example, distinguishes Retail and Pro orders by how they are submitted. An API result and a retail fee statement cannot be compared as if they always described the same order. We keep the qualifying rule with the number instead of selecting whichever fee makes the platform look cheaper.
If two current documents still disagree about the same product, we show the disagreement and narrow the conclusion. A newer announcement controls only if it applies to that field and is effective. We do not average conflicting limits, assume the largest number wins or silently replace a missing value with a platform-wide maximum.
A reviewed document, with its actual boundary
This classification is based on how the order is submitted, not on a permanent user profile.
- Content reviewed
- Supports
- Order classification depends on submission method. Standard API and interactive retail flows must be distinguished.
- Does not prove
- A fee for every order, legal professional-client status or a trader’s geographic eligibility.
Dates belong to the observation
A publisher’s update date, our content-review date and an automated request date answer different questions. A fresh HTTP response does not refresh an old fee. Revising an article does not retest every linked endpoint. Published observations retain their collection dates. Tools use documented schedules or visitor-entered assumptions rather than fetching market quotes. Immutable research editions keep their original records, with a separate corrected edition when necessary.
Open the evidence beside the number to see its scope. A page-level “reviewed” label means the article was materially reviewed; it should not be read as a same-day verification of every rule. When a value is essential to a calculation but lacks usable evidence, the tool must identify the gap instead of treating it as zero.
Evidence-state glossary · six distinct records
- Link checked (retrieved)A request or recorded review succeeded. Relevant content and support for the cited field require a separate content check; HTTP 200 alone is not proof. Show the observation date and its scope. A legacy link check does not become a content review; dynamic values need their own timestamps.
- Automated check limited (restricted)Authentication, geography or an anti-bot control prevented a fresh public retrieval. A preserved capture may provide historical context, but the field cannot be called freshly retrieved.
- Awaiting verification (transient)A timeout, temporary server error or first failed check interrupted retrieval without proving that the source disappeared. Retain the record for recheck; do not promote it to unavailable until a repeated or hard failure is documented.
- Link unavailable (unavailable)The expected source failed, disappeared or no longer contained the cited material. The dependent field is lowered in confidence or withheld until a replacement source is reviewed.
- Superseded (superseded)A newer official record replaced the source or rule. The earlier record stays in history when it explains a material change, but it does not control current output.
- Archived (archived)A dated historical copy is retained for change tracking or reproducibility. Archived material is never presented as a current live or retrieved value.
Some older records preserve only an HTTP observation. They cannot retroactively demonstrate a content review. We keep that limitation visible instead of inventing a successful retrieval or a review date.
Example: an access-limited check is not reviewed content
The stored country-source audit recorded HTTP 403 for Kraken MTF — Interpreting the logs on . Its content-review field is empty.
This is a historical automated check, not a new result from today. It proves the request was refused in that run. It does not prove the current contents, that the document is gone, or that a country’s residents cannot trade. We do not turn the failed request into a product conclusion.
Captures must demonstrate what their captions claim
A useful screenshot shows the relevant field, its product context and enough of the interface to interpret it. A cropped fee without its account tier can mislead; a homepage cannot prove a listed contract or a successful withdrawal. We do not manufacture exchange screenshots or imply that viewing a public interface involved a funded account. An illustration is explanatory artwork, not operational evidence.
A retained capture keeps its capture date even when nearby prose changes. Cropping should improve legibility without hiding a condition. If the necessary context cannot be shown clearly on a phone, a readable table or short explanation with the original source is preferable to a tiny, decorative screenshot.
Secondary sources and material corrections
Independent reporting can establish an attributed report, identify an incident or point us toward a primary document. It does not become the current exchange rule merely because several sites repeat it. When the controlling source is missing, the conclusion stays limited. Reports of a restriction or service incident need the applicable operator notice before they change a claim of confirmed access.
When a source moves, the replacement must support the same field; a working homepage is not a repaired evidence link. Material corrections identify the affected claim and consequence in the research changelog or dossier history. Contact the editorial desk through the corrections process with the page, disputed field and controlling source. Credentials and private account documents are not needed.
Sources and research
The Paradex example was checked against its public Retail and Pro Orders documentation on 9 October 2026. The restricted example preserves the separately dated country-source audit, including its empty content-review field. These records intentionally demonstrate different levels of evidence.
Product eligibility is developed in the country methodology; execution and score interpretation belong to the research methodology. Neither example involved account approval, a funded trade or withdrawal testing.