How we compare exchanges

Current recommendations: country-recommendations-2026.11-v3

The live matcher ranks researched, exact-product options using dated CoinGecko venue Trust Scores (55%), product evidence (20%), country relevance (20%) and confirmed preference fit (5%). Components are normalized from 0 to 1, then multiplied by their weights. Trust is the external /10 score divided by 10; product evidence is 1 for confirmed service, 0.9 for exact conditional product policy and 0.75 for documented market-listed products with official platform policy. Missing trust is never invented and keeps ordinary venues out of the main ranking. Regional exceptions need direct service evidence and show a parent venue reference only, not a local rating.

Country relevance begins at a neutral 0.5, adds 0.25 for reviewed official service presence, and up to 0.25 for exact-region app download interest (1 − (chart rank − 1)/100, bounded at zero). Missing usage stays neutral, not “unpopular”. Preference fit uses confirmed selected features with priority weights 3/2/1; no preferences is neutral at 0.5. Equal final scores are ordered by name then entity. Global volume, adoption rank, licensing and affiliate payments have no weight.

“Available, subject to verification” requires fresh exact-entity general and product support with no unresolved conditions. “Check at signup” is a researched conditional candidate: fresh official platform policy, independently documented requested product, and no known exclusion. It is not confirmed onboarding; absence from a restricted list never means confirmed permission. Known or stale restrictions and conflicting evidence block candidates. Service restrictions are exchange, entity and product specific, not blanket country or licensing gates. Actual retail/professional limits and local routing still apply; regional offerings never inherit global perpetual access. Unspecified citizenship triggers a conditional check where nationality bans exist, not an assumption that every ban applies. Unknown mandatory features appear only in separate alternatives; explicitly unsupported requirements exclude.

We normally show five options, with show-more up to seven, never padding blocked services or changing the recommended product. If fewer are responsibly researched, separately labelled other-product alternatives may be shown without relaxing must-haves. Beginners can browse leverage research and demo-oriented education; under-18 visitors receive education only. Identity-verification refusal does not produce ready-to-open recommendations.

CoinGecko Trust Score uses spot venue methodology, not futures-product safety or a solvency guarantee. CoinGecko methodology. The cohort is 50 CMC derivatives-ranked venues plus six regional spot exceptions, not the 50 safest venues. Scores and market snapshots expire within 30 days; country/incident policy within seven days, or sooner when sources expire. Monthly manual snapshots are not live measurements.

The current documented Binance recovery caution subtracts 20 points from its internal ranking while fresh. The external score remains 10/10. The November 2 notice reports approximately 95% affected-asset replenishment and phased withdrawal recovery, not full resolution or total current withdrawal failure. Draft historical penalty proposals and draft quality snapshots never enter the public catalog or ranking.

Country interest uses dated app-chart positions: Storelift iOS free Finance snapshots and AppBrain Google Play Finance data for India. These are download-interest proxies, not active users, futures usage or exchange market share. The supplied metric and observation date appear in Details; local apps remain separate. The country explorer retains its stricter compatibility resolver (compat-2026.11-v2); its positive counts are not the conditional recommendations shown by the matcher.

Cryptange AI recomputes these same recommendations server-side and selects at most four vetted claims. It cannot add exchanges, invent fees or scores, change eligibility or reorder recommendations. Raw questions, saved answers and history are not sent to the model. Missing measurements and fees are explained plainly.

Country research: what each signal means

We expand coverage by reviewing original exchange documents, not by treating a search result, a translated website or a local marketing page as permission to open an account. Every published signal identifies its source, the country, the exchange service or entity, the product scope and the date we checked it.

• Account and product access: explicit official support is evaluated separately for general account access and your exact product, within the same entity. Restrictions and final onboarding checks still apply.
• Official service descriptions: a documented country service can be a lead to explore. It does not, on its own, confirm your residence, citizenship, verification requirements or every product.
• Official guides: country-specific instructions describe the documented service only. A guide is not an admission guarantee.
• Local currency and P2P: a named funding or currency market is a funding lead for buying or exploring. It is not proof of resident eligibility, bank support for everyone, active order-book spot access or derivatives access.
• Historical campaigns: ended campaigns show past market presence only. Their rewards are never presented as currently available.
• Measured popularity: only reviewed usage-chart records can support a country popularity answer, with the exact measure, observation date and source. An app-store Finance chart is not an exchange-only ranking or market share. With no reviewed snapshot, we say so.

Country profiles combine published adoption context and editorial guidance; neither becomes exchange-access evidence. In the country explorer, service and funding signals stay separate from strict compatibility. In live recommendations, reviewed official service can improve country relevance, but never override restrictions. A restriction for the same entity still blocks an exploration lead. Local entity evidence is not transferred to a global platform, or vice versa.

New checks receive their actual review date and a bounded expiry; adding coverage never renews older rows automatically. Out-of-date, unconfirmed or contradictory material cannot become confirmed access. Search candidates stay private until the underlying official document has been checked. Country access, local regulatory status and measured popularity remain separate.

Popularity answers are deterministic and labelled Research summary, not AI. Cryptange AI can explain reviewed claims but cannot supply missing country measurements, predict prices or make best, safest or legal assertions.

The legacy scored method below (version 1) needs published editorial assessments and is retained separately; it is not the live v3 recommendation ranking.

This is our proposed version-1 editorial method, not an industry standard, a certification or a prediction of safety or profit.

1. Eligibility checks come first

For your country, we check general account access and the exact product you chose (spot, futures, spot copy or leveraged copy) separately. Each check is supported, restricted, unknown or not applicable, with a source and review date.

• A reviewed restriction always wins over a favourable record.
• No research means unknown — never “supported”.
• Expired or conflicting evidence keeps an exchange out of ranked matches.
• Your “must have” requirements are filters; an unverified feature doesn't pass.

Your answers are your own statements. A match means an exchange appears compatible with your answers; the exchange makes the final decision.

2. Baseline assessment (Q)

Q = 0.45 × safeguards + 0.25 × reputation + 0.20 × experience fit + 0.10 × relevant costs

Each input is a reviewed 0–100 editorial assessment for your product and experience level. If any is missing, no score is calculated.

3. Your priorities (P)

P = Σ(rank weight × priority fit) ÷ Σ(rank weight), with weights 3, 2, 1

Safeguards/reputation fit uses (0.45 × safeguards + 0.25 × reputation) ÷ 0.70. Ease uses experience fit; costs use relevant costs. Payments, assets and tools score the share of your selections with verified support. If a selection can't be verified we show “Fit not calculated” rather than guess.

4. Final score (M)

M = 0.80 × Q + 0.20 × P (or M = Q if you chose no priorities)

We rank by M, then Q, then a stable exchange ID. We show up to three matches and never fill slots for the sake of it. The rounded number is a suitability assessment out of 100 — not a percentage of safety.

5. Offers and commission have zero weight

Welcome offers, their expiry, and any commission we may earn are not inputs to Q, P or M. See the affiliate disclosure.

6. Freshness policy

Proposed operating defaults: campaigns 24 hours, fees 7 days, country/product access 7 days, assessments 30 days. A source can expire sooner. Freshness is checked at the moment you get results. This is not real-time verification.