ChatGPT Plus Payment Declined: Check the BIN Range First, Then Your Account

ChatGPT payment failed, subscription declined, and the page throws “Your card has been declined.” — here is the answer up front, and that answer comes in two layers. Get them backwards and you will burn an entire day:

• Layer 1: whether the issuer policy on this card’s BIN range supports AI-subscription merchants at all. For some BIN ranges the issuer policy explicitly does not support the AI-subscription scenario — take one of those to ChatGPT and the failure is a BIN-range problem, not something you misconfigured. At this layer, no amount of troubleshooting, topping up or IP switching will help; the only fix is moving to a BIN range that supports the AI-subscription scenario.
• Layer 2: the BIN range does support AI scenarios and the card itself tests clean — then the failure sits on the user side and the account side. Limits, card status, balance, an account that is too new, an unstable IP — nearly everything at this layer you can fix yourself, and swapping cards is the least useful thing you can do.

So step one of any diagnosis is always to settle layer 1: does the BIN range of the card in your hand state that it supports the AI-subscription scenario (the BIN-range notes on the card-issuing page spell out which usage scenarios that range supports). Once you have confirmed support, work through this article in order — error text, real meaning, how to fix it — and take layer 2 apart one reason at a time. The error text itself barely tells you which one it is, which is exactly what this article is here to solve.

This article is written for people who do not have an overseas virtual card yet and just failed to subscribe with a Chinese bank card. If this is your first subscription and you want the forward-facing walkthrough instead, read the complete ChatGPT Plus subscription guide.

1. Two layers first: is the BIN range supported, then which bucket

With AI-subscription support on the BIN range confirmed, a declined ChatGPT subscription can only happen in three places. Work out which one you are in first, or the rest of your troubleshooting is just guesswork:

• Bucket 1: the transaction never left the country — the card has no international card network rails, or the issuer keeps this class of cross-border digital-subscription merchant switched off by default. Most Chinese bank cards fail here. Signature: you finish the card number, hit Subscribe, get a near-instant decline, and there is no transaction record and no SMS in your banking app at all.
• Bucket 2: the card side declined it — the request reached the issuer and was stopped by a limit, the card status or the balance. This is what cardholders hit most often and also the easiest to fix yourself. Signature: the failed attempt shows up in the card’s transaction history, with a decline reason attached.
• Bucket 3: merchant-side risk control — nothing is wrong with the card; OpenAI or its payment provider blocked you on account age, IP or frequency. Signature: the card’s transaction history shows nothing at all, while the error talks about too many attempts, or says the payment was not approved.

2. Error-message decoder: what each line actually means

The handful of lines you see when ChatGPT declines your credit card are heavily compressed copy on the OpenAI payment page, and the same sentence can sit on top of completely different causes. The table below maps the six messages you are most likely to have searched for onto their real meaning and the correct next move:

Pay attention to the last two rows: once either of those appears, every attempt you make for the rest of the day has zero diagnostic value — you will keep getting declined, but the cause is no longer the one you are troubleshooting. This is where a lot of people reach the wrong conclusion that “this card does not work”. The retry counts and cooldown durations in the table need one caveat: the gateway’s exact thresholds and cooldown durations are external parameters the operators never publish. These are values our support team has observed over time, not an official figure.

3. Why a Chinese bank card is almost certain to be declined by ChatGPT

Conclusion first: this is not bad luck, it is structural. For a Chinese bank card to land a ChatGPT Plus subscription, it has to clear three gates at once, and most cards stop at the first one:

1. Whether it has international card network rails at all. OpenAI charges over the international acquiring rails of Visa and Mastercard. A CNY-only card has UnionPay rails only, so the request never reaches the issuer side. Typical signature: instant decline, no transaction record in the banking app, no SMS.
2. Whether the issuer allows this merchant category. Dual-currency and multi-currency cards do have international rails, but issuers keep per-merchant-category (MCC) switches and caps on cross-border merchants, and cross-border digital subscriptions, cloud services and AI merchants sit in the off or tightly capped bucket by default at many banks — with a switch that is usually nowhere in the mobile banking UI. A decline at this layer is what produces “Your credit card was declined. Try paying with a debit card instead.” — it is telling you that the card-type-plus-merchant-category combination is not allowed, not that a debit card would actually go through.
3. Whether the 3DS verification code reaches you. Even past the first two gates, the payment provider may still demand a 3-D Secure challenge. A code from a cross-border merchant has to travel over international SMS rails to a mainland number, and delay, loss and interception are all common; when the countdown runs out, the attempt is void, and trying again writes another failure record. This is also where a good share of the “I got locked out for no reason” stories come from.

4. Once the BIN range is right: real decline reasons by frequency

This section is about layer 2: the BIN-range notes state support for the AI-subscription scenario, the card itself tests clean, and you still get declined. The answer is that the reasons are an entirely different set — and the overwhelming majority are ones you can fix yourself.

The distribution below is not an estimate. It is our own ledger’s attempt-by-attempt attribution of failed authorizations across the OPENAI subscription merchant family (three billing lines: web subscription, Credit, Google Play), as of 2026-07-17; the full version is on the ChatGPT platform payment page. Hold on to two premises while reading it: ① it counts declines on BIN ranges that support the AI-subscription scenario, after the user already has a virtual card — failures on Chinese bank cards never reach this layer, and cases where the BIN range itself does not support AI subscriptions are not in here either; ② the distribution differs by merchant family, so do not transplant this set onto another vendor — for Claude or Cursor, read the attribution on their own platform pages.

Group those four (same basis: attempt-by-attempt attribution of failed authorizations in the OPENAI subscription merchant family, on BIN ranges that support the AI-subscription scenario, as of 2026-07-17) and the picture is blunt:

• Limits and caps together: 53% (per-transaction 36% + daily/weekly/cumulative 17%) — more than half of all declines are a limit that was not set high enough.
• Card status: 28% — not activated or frozen; the card is still in the list and the balance is still there, but every authorization gets declined.
• Insufficient balance: 19% — usually a shortfall of a few dollars.

Why limits run out so easily

Because most people read the limit as “available per month”. In reality the limit on the card detail page has four dimensions: per-transaction, daily, monthly and cumulative. Of those, the cumulative limit is lifetime and does not reset monthly, and subscription renewals keep drawing from the same pool: the first month clears $20, the second month draws from the same pool, and the month the pool runs dry is the “cumulative limit exhausted” decline.

One more product fact that has to be stated plainly: on the user-facing card detail page these four are read-only, there is no raise-limit button. Per-transaction, daily and monthly are set by the BIN-range configuration and cannot be self-adjusted; the only one you can raise yourself is the available cumulative limit, and the way to do it is to top up the card. So when an article or a support agent says “raise your limit”, the concrete action is always top up the card, not hunting for a control that does not exist.

While we are here, one naming collision worth clearing up: what this site calls self-service limit increase refers to the cap on how many cards you may open, which is not the same thing as the spending limits above. The rule on card count is: at most 5 active cards at a time by default, and 5 cumulatively (closed cards included). Once you hit the cap you can request an increase on the card-issuing page; approval is decided automatically from your account usage record, and the card-issuing page is authoritative on how many cards you have left.

A charge higher than the list price is normal

ChatGPT Plus is listed at $20.00 per month, but the charge shapes that actually appear in our ledger come in three forms (our own ledger, OPENAI subscription merchant family, as of 2026-07-17): the standard $20.00, $21.28 with state sales tax, and $24-25 when an account has an unpaid invoice that gets collected together with the next charge. That last arrears-catch-up shape is the single most common reason behind “I funded it for $20 and still got declined”.

State sales tax also explains a second thing people notice: rates differ by state, so two people on the same plan can be charged a dollar or two apart. Seeing someone charged $20.00 while you are charged $21.28 does not mean anything went wrong. The tax amount is decided by where the billing address is, so it is not a variable you get to operate on — just budget for it.

So do not budget off the list price: for the $20 plan, fund the card to ≥$26 as a rule — that number exists to cover the $24-25 arrears-catch-up shape specific to ChatGPT, and the tax-inclusive tier ($21.28) is covered along the way. The $200 tier (Pro) works the same way: budget ≥$210. The platform minimum issuance limit is 30 USD (the card-issuing page is authoritative), so opening a card for the $20 plan at the minimum covers all of these shapes. For full pricing across plans and where the charge differs from the sticker, see what a ChatGPT subscription actually costs.

5. Troubleshoot in frequency order (start here if you already have a card)

Follow the frequency order from the last section — check the highest-share causes first, do not start guessing at the exotic ones. But before any of that comes one step you cannot skip: rule out layer 1 first.

1. Step 1: confirm this card’s BIN range supports the AI-subscription scenario

   Go to the card-issuing page and read the usage-scenario notes for the BIN range this card belongs to; check that AI subscriptions are listed. If support is not stated, stop troubleshooting here — on a BIN range whose issuer policy explicitly does not support AI subscriptions, a ChatGPT attempt is a BIN-range problem, all five remaining steps will not change the outcome, and the fix is a BIN range that supports the AI-subscription scenario. Only once support is confirmed does each step below mean anything.

2. Step 2: open the card’s transaction history and find that failed attempt

   This is the watershed of the whole diagnosis. You can find it → the problem is on the card side, continue to step 3; there is no record at all → the decline happened at merchant-side risk control, the card is innocent, skip straight to section 6 and look at the account and the IP. Whether you do this step decides whether you fix one concrete problem or spend a whole day guessing.

3. Step 3: check all four limit dimensions, not just the balance

   Open the limits block on the card detail page and compare all four — per-transaction / daily / monthly / cumulative — against this charge (for the $20 tier, compare against ≥$26, not $20), then check the available balance on the card. Any one of them below the charge produces a decline, and the error text is identical in every case. Remember the rule from the last section: only the available cumulative limit can be raised by topping up the card; turn on auto top-up while you are there so next month’s renewal does not repeat this.

4. Step 4: confirm the card status is Active

   On the same basis as the last section, not-activated and frozen cards together account for nearly three in ten. Frozen is the easiest one to miss, because the card still looks fine in the list and the balance is still there, while every authorization gets declined. Find out why it froze first (a risk-control trigger, or unpaid fees), fix that cause, then unfreeze — otherwise it will just freeze again.

5. Step 5: copy the card details, never retype them

   Use the copy buttons on the card detail page for the card number, expiry and CVC, and copy the billing address exactly as the card detail page shows it. Retyping produces shifted CVC digits and month-year swaps on the expiry date, and those still show up in real failure records — errors of that kind do not even clear the $0 card-attach verification. If you cannot even attach the card, this is almost always why. The expiry format is MM/YY.

6. Step 6: confirm you are on the regular monthly flow, not a trial

   This is the hidden reason behind a lot of repeated failures: submissions through the trial entry are more likely to be stopped at merchant-side pre-authorization risk control — the signature is that not even the $0 verification appears in the card’s transaction history — the request most likely never reached the issuer, and what got screened is the account or network. The same card in the same environment often goes through on the regular subscription page. If you keep getting declined on the trial page, switch to the regular subscription page instead of burning retries there.

6. The four prerequisites for adding a card: account / IP / billing address / card

If your failure is the kind with no record in the transaction history, work through the four items below one at a time — nearly all of them land in here. These four hold simultaneously or not at all: if any one of them is wrong, the other three being perfect changes nothing. Note that the fourth item, the card, contains the one that is most often skipped: whether the BIN-range notes state support for the AI-subscription scenario.

Account: why brand-new accounts fail at card-add so often

From the cases our support team has handled over time, accounts under a week old with no normal usage record during that week fail at card-add noticeably more; that goes double for accounts registered entirely inside a VPN environment and sent straight to the payment page. Even when such an attempt slips through, later identity-verification prompts are more likely. The correct order is: register, use it normally for a few days (ask questions, upload files, open a few conversations), and subscribe a week later. For the full set of triggers that put an account on risk control’s radar, see the real reasons Claude / ChatGPT accounts get banned.

IP: the question is not datacenter vs residential, it is dedicated vs shared

The common claim online is that you must use a residential IP and a datacenter IP is always declined. Based on the cases we actually handle, that claim is not precise enough: the gap between datacenter and residential IPs is smaller than people assume, and what really decides the outcome is whether the exit is yours alone. Shared proxy-service nodes fail so often because that exit has already been used for large volumes of the same kind of activity and is on blocklists — which has nothing to do with datacenter versus residential. Given a dedicated exit, a residential IP is still better than a datacenter one, but that is a marginal improvement on top.

One more discipline that gets ignored: once the IP has changed, wait 24 hours before adding a card. Switching from a local node to a US node minutes before payment is a textbook high-risk behavioral signal — cross-country and cross-region jumps especially.

Billing address: it is a set of details to copy, not a riddle to solve

Nearly every guide tells you the billing address must match the card’s registered address exactly, and so a lot of people get stuck here because they cannot find any such address. Let us settle it: the billing address shown on the card detail page is issued to you precisely so you can copy it into the merchant checkout page. You do not need to find it elsewhere and you do not need to invent one — open the card detail page and copy the street, city, state and ZIP exactly as written.

Only one thing really matters: the billing address must be in the same country as the IP you pay from, ideally the same state. A US BIN range goes with a US IP and a US address; a US card behind a Hong Kong node is one of the most common mismatches there is. When that condition does not hold, no amount of careful address copying helps.

A last reminder about ordering: after a decline, walk the six steps in section 5 first (BIN-range scenario → transaction history → limits → card status → card details → subscription entry), and leave the billing address for last. It is the one item here that gets messier the more you edit it, and touching it first usually just makes it impossible to tell which change did the work.

7. Three failures in a row will lock you out — the trap newcomers hit most

This one does so much damage because it runs exactly counter to instinct. The first reaction to a decline is “the network glitched, click again”, the second failure becomes “let me try another card”, the third “let me try another browser” — and three attempts in, the lockout is live. From then on, every correction you make (topping up to raise the limit, switching IP, editing the address) produces the same failure, so you end up rejecting one correct fix after another for the wrong reason. That is the real source of “the more I fix it the worse it gets”.

A cost note while we are here: the decline fee depends on the BIN range — some ranges charge nothing for declines, but others charge $0.60 per failed attempt (itemized on the card-issuing page; whether it applies and how much is shown on the card-issuing page and in the BIN-range notes). On the ranges that charge it, ten blind attempts cost you more than a ruined environment — there is a visible string of fees too.

The correct cooldown discipline

1. First failure: stop, use the test in section 1 to work out which layer you are in, fix the problem, then wait at least 1 hour before the second attempt. Retrying the instant you fix something can still hit the rate limit.
2. Second failure: never make a third consecutive attempt. Cool down fully until the next day, and do not even open the payment page in between. The cooldown exists to let the counter reset, and any attempt during it restarts the clock.
3. What to do during the cooldown: first confirm the BIN range supports the AI-subscription scenario (if it does not, stop waiting and change range); top the card up to ≥$26 ($200 tier: ≥$210), turn on auto top-up, pin the IP down and stop switching, re-enter the billing address exactly as the card detail page shows it, and confirm the account is at least a week old. When the cooldown ends, line every condition up at once before you submit.

A signal you can check yourself: if your error has turned into a rate-limit style message, or a decline that blames the number of attempts, the lockout is live and there is no point trying again that day.

8. The standard sequence for getting through in one go

String every condition above into the correct time order and you get the seven steps below. The order itself matters — account warm-up and IP stability have to start days ahead; they are not things you can patch in at payment time.

1. Step 1 (one week ahead): warm the account up

   Register an OpenAI account with United States as the country. Then use it normally for a week: ask questions, open conversations, upload files. Do not sprint from registration straight to the payment page.

2. Step 2 (2 days ahead): pin the IP down

   Pick a dedicated US exit and do not switch nodes for the next 24-48 hours, and do not bounce between a direct local connection and the proxy. Check yourself on ipinfo.io every few hours and confirm country and city stay unchanged.

3. Step 3: open a card on the right BIN range, keep only this period’s usage on it, turn on auto top-up

   Read the usage-scenario notes on the card-issuing page first and pick a US BIN range that states support for the AI-subscription scenario — get this step wrong and everything after it is wasted. Issuance starts at $1 per card and varies by BIN range; the shared-limit card we recommend for AI subscriptions is $2, shown directly on the card-issuing page. The minimum issuance limit is 30 USD (the card-issuing page is authoritative), and funding the card to ≥$26 covers both the tax-inclusive and the arrears-catch-up shapes. Do not load six months of money onto the card to save yourself the trouble — put this period’s usage on the card, leave the rest in your account balance, then turn on auto top-up from the card detail page: when a charge is due or fails for insufficient balance, the system tops the card up from your account balance automatically.

4. Step 4: copy the billing address from the card detail page

   Copy the US address shown on the card detail page into the checkout page exactly; the address country must match the country of the IP you pay from, and the same state is better. Filling details: street only on line 1, line 2 empty, city / state / ZIP consistent with each other. Whether you may swap in your own set varies by BIN range, and the card detail page and its notes are authoritative; when declined, try the address that shipped with the card first, do not change it as the opening move, and do not copy a universal address circulating online.

5. Step 5: use the regular monthly entry, not the trial

   In ChatGPT, click your avatar → Upgrade plan → pick Plus ($20 per month) as a regular subscription. Submissions through the trial entry are more likely to be stopped at merchant-side pre-authorization risk control (signature: not even the $0 verification appears in the card’s transaction history, so the request most likely never reached the issuer), so do not burn retries there.

6. Step 6: copy and paste every card field

   Use the copy buttons for the card number, expiry (MM/YY) and CVC; do not retype them. The name must match the cardholder shown on the card detail page. Review the whole page once before you submit.

7. Step 7: if it fails, give yourself exactly two chances

   First failure → check the transaction history to locate the cause → fix it → wait at least 1 hour before retrying. Second failure → cool down until the next day and do not click even once in between. Never make a third consecutive attempt. (Again: thresholds and cooldown durations are external parameters the operators never publish; these are values our support team has observed over time, not an official figure.)

9. FAQ

10. Related reading

• ▸ The complete ChatGPT Plus subscription guide — the forward-facing companion to this article, from registration through upgrade
• ▸ What a ChatGPT subscription actually costs — where the gap between list price and real charge comes from, and how to pick a plan
• ▸ Three real AI subscription payment failures — attempt-level attribution for card-attach, charge and renewal failures
• ▸ The top 10 reasons a virtual card gets declined — the general-purpose checklist, not limited to ChatGPT
• ▸ The real reasons Claude / ChatGPT accounts get banned — the full set of account-layer risk-control triggers