ChatGPT Suddenly Asking for Phone Verification? We Checked 24 Sources — the Truth Isn't What You Think
In April-May 2026 V2EX blew up with "Codex is asking me to add a phone number" reports, and the story spread fast. We spent a night checking 24 independent sources and 67 original claims with 3-vote adversarial verification (18 confirmed / 7 killed) to set the record straight: this is not a blanket SMS rollout — it mainly affects the OpenAI Codex pipeline. We debunk the 7 most common rumors, cover the current real status of mainland China / Hong Kong / VoIP numbers, and explain why this and your virtual credit card are two completely independent tracks.
Around the May Day holiday, V2EX suddenly exploded with threads. “Codex is asking me to add a phone number”, “is my account about to get banned”, “is ChatGPT about to say goodbye to Chinese users for good”.
We got curious too — did OpenAI really make a move? So we spent a night going through 24 independent sources and 67 raw claims from around the web, with 3-vote adversarial cross-verification. In the end, 18 claims confirmed, 7 killed.
The conclusion may not match what you have heard.
1. The Truth in One Sentence
This is not SMS being force-rolled onto everyone. The one being hit is mainly Codex, OpenAI’s developer tool (CLI / Desktop / Cloud). People using the regular ChatGPT web or App are basically fine.
To this day (2026-06-05) OpenAI has published not a single official announcement. All the evidence was pieced together from three sides: the GitHub repository (openai/codex#20320 officially tagged auth) + support replies on the official developer forum + field-test threads on V2EX.
OpenAI Support’s exact words on the official forum: “Codex cloud interacts directly with your codebase, so it needs stronger security”. In plain language — Codex can read your code directly and run commands, so when things go wrong the cost is an order of magnitude higher than ChatGPT, which is why OpenAI stacked an SMS layer on it. The main ChatGPT web currently triggers it only on risk-control hits — it is not a universal rollout.
2. Who Actually Runs into It?
Based on the evidence so far, 4 kinds of scenarios:
- Using Codex CLI and one day suddenly getting
token_revokedor 401, then being asked to add a phone on re-login - Using Codex Desktop and logging in again after switching computers / browsers
- Having MFA (two-factor) enabled, then getting an SMS step-up stacked on top when logging in to Codex
- Risk control flags an anomaly — unusual IP location, new device, abnormal behavior
If you just browse ChatGPT on the web / App, most people feel nothing at all.
A V2EX user’s field report in t/1209905 is quite representative: “it worked fine in the afternoon, then in the evening everything suddenly got logged out; some accounts need it and some do not, and logging in again demands addphone”. This shows OpenAI is triggering selectively, not sweeping everyone.
3. 7 Rumors Knocked Down by Adversarial Verification
This section saves you time. The claims spreading hardest in the community but with insufficient evidence, knocked down one by one after 3-vote adversarial voting:
| Circulating claim | Adversarial vote | Reality |
|---|---|---|
| 70% of free-tier accounts got force-pushed SMS | 0/3 | A blogger made it up; no field-test sample |
| The main ChatGPT web demands SMS from everyone | 0/3 | Codex only; the main site triggers sporadically, not universally |
| OpenAI issued a new policy force-pushing existing users | 0/3 | No official announcement has been published to date |
| The phone-only signup beta = force-push on existing users | 3/3 no | That beta is a signup simplification for new users in the US / India |
| Telegram gray-market phone-verification services are already mature | 0/3 | Just one or two comments touting for business; not a market |
| The same number must repeat SMS verification every time | 0/3 | Passing once basically persists |
| The trigger condition is multi-user shared networks + bulk signups | 0/3 | A blogger’s speculation; no evidence |
Adversarial method: each claim gets 3 independent agents attempting to refute it; 2/3 in favor keeps it, 2/3 against kills it. All sources are publicly checkable.
4. What Chinese Users Care About Most: How to Get Past the Number Gate?
Honestly, this gate is almost fully closed for Chinese users. Here are the confirmed facts laid out:
| Number type | Status | Evidence |
|---|---|---|
| +86 mainland China | 100% rejected | The mainland was never on OpenAI’s supported-country list |
| +852 Hong Kong | Blocked in Codex | openai/codex#22954 issue in the official repository |
| Google Voice / TextNow / Twilio | Explicitly rejected by the system | Prompts “this appears to be a virtual number, please provide a non-virtual phone number” |
| 5SIM / TextVerified | Generally fails | Multiple V2EX users field-tested and could not receive the SMS |
| hero-sms | Fails + account bans | One user reported “wasted a whole afternoon and the account got frozen” |
| DogeSMS | Partly successful | $0.26 - $0.80 per attempt; some users field-tested successfully |
| DuoPlus cloud numbers | Some people use it | Weak evidence; a single-thread recommendation |
Reason: SMS-relay channels change too fast — what works today may die tomorrow, and OpenAI’s VoIP detection keeps running. If you need the latest field tests, go to V2EX t/1209905 and read real user feedback from the last 1-2 weeks — not blogs / WeChat advertorials.
5. A Very Common Misconception: Phone Verification vs Virtual Cards Are Two Separate Tracks
While we are here, let us untangle something many people lump together.
The full ChatGPT Plus user journey actually has 3 independent gates:
| Gate | Owned by | Affected by this SMS wave? |
|---|---|---|
| Login | OpenAI risk control (email / phone) | Affected (this is the one in question) |
| Payment | Stripe / card networks / issuing banks | Completely unaffected |
| Usage | Web / App / Codex | Web and App unaffected |
These 3 gates are independent of each other. If your Plus is paid with a virtual card (such as an RDVCC virtual credit card), this SMS wave has nothing to do with the card:
- The card keeps charging as usual
- Plus service continues as usual
- Unless you switch IP / device and get flagged at login — but that is a login-side matter; the card side is fine
So what exactly is a virtual credit card?
Plainly put, it is a bank card with no physical plastic — just a card number + CVV + expiry date. Common uses:
- Subscribing to overseas services ( ChatGPT Plus / Netflix / Spotify / Midjourney)
- Facebook / Google / TikTok ad campaigns
- Cross-border shopping (Amazon / Walmart / eBay)
- Cross-border SaaS subscriptions
Why are more and more people using virtual cards? 3 practical reasons:
- Physical cards struggle cross-border — domestic Chinese credit cards very likely get declined at an MCC like OpenAI’s
- One card per use is safer — close the card once a subscription is done, leaving no trace
- USDT top-up flows better — no forex needed; funds are credited on-chain within minutes
To fully understand the concept, read this: What Is a Virtual Credit Card — Definition / Types / Uses / Settlement Explained in One Read.
The virtual-card track was left completely untouched by OpenAI this time. Along the Stripe / Checkout path there is no sign of any VCC BIN range being blocked. If all you want is stable Plus renewals and trouble-free charging:
- · Genuine US BIN (99%+ Stripe approval rate)
- · 3DS enabled by default
- · USDT top-up, credited in 2 minutes
- · No monthly fee, one card per service
6. The Things We Are Not Sure About (Being Honest with You)
We will not draw conclusions for you. A few questions where the evidence is currently insufficient:
- Will OpenAI extend the Codex policy to everyone on the main ChatGPT site? Nothing so far in the short term, but no guarantees — worth rechecking in 1-2 weeks
- How long can SMS-relay channels like DogeSMS hold out? Once VoIP detection tightens, hard to say
- How long will Hong Kong +852 keep working on the main ChatGPT site? Codex already blocks it; the main site has not been exhaustively verified
- How much is the actual decline in Plus renewals among mainland China users? So far no large-scale trend of “abandoning OpenAI for domestic models”
Research window: 2026-04-30 to 2026-06-05, about 5 weeks. If OpenAI extends enforcement across the whole product line in the next 1-2 months, the conclusions here will need re-evaluation.
7. Final Words
No fearmongering.
No shouting “OpenAI is tightening again, stockpile accounts now”.
No drawing conclusions for you.
If you are an ordinary ChatGPT Plus user, on the web or App, paying with a stable virtual card — odds are you will not feel this SMS wave at all.
If you do run into it, knowing which path works and which path is a waste of time is all you need.
If OpenAI makes an official move, we will update this article.
FAQ
Q: I only use ChatGPT on the web — will I be asked for phone verification?
Most likely not. This is a step newly added to the Codex (developer tool) path; ordinary ChatGPT web login is basically fine. It only triggers sporadically if OpenAI risk control deems your login suspicious (unusual IP location, new device, abnormal behavior, etc.).
Q: I currently pay the Plus monthly fee with a virtual card — will I get stuck?
Phone verification is the login stage; the virtual card is the payment stage — these two are completely independent. In this SMS wavethere is no evidence whatsoever that OpenAI simultaneously tightened VCC BIN ranges. Your card keeps charging as usual, and Plus service continues as usual.
Q: Does RDVCC offer a phone-verification bypass service?
No. We only do virtual credit cards + cross-border payments and stay off the SMS-relay line — (1) that is login-side business, not payment-side; (2) SMS-relay channels change too fast to promise stability; (3) it risks violating the OpenAI ToS. If you need a number, please look for the latest field-tested approaches in user communities yourself.
Q: Will OpenAI soon extend forced SMS to everyone on the main ChatGPT site?
Insufficient evidence so far. The Codex team’s support explanation is “Codex can read code directly and needs stronger security”, a rationale that does not hold for the main ChatGPT site. But OpenAI has issued no official announcement to date, and enforcement could in theory shift at any time. We keep monitoring the official X account, help.openai.com, and the OpenAI Status Page, and will update this article on any official move.
Q: I am a mainland China user — if I hit phone verification, is there a reliable approach?
Honestly: there is no 100% reliable approach. +86 mainland numbers are not on OpenAI’s country list, Google Voice and other VoIP numbers are rejected outright by the system, mainstream SMS-relay platforms (5SIM / TextVerified) fail in field tests, and hero-sms even gets accounts frozen. The few that pass in field tests (such as DogeSMS) cost $0.26-0.80, and success is not 100% either. Suggested strategy: (1) keep all related accounts logging in from the same IP, same device, in a stable environmentto minimize the odds of triggering risk control; (2) if you really hit it, check V2EX for field tests from within the past week — not WeChat advertorials.
Related Reading
- ▸ ChatGPT Plus Upgrade Declined? 60% Are Card BIN Issues — The Complete Troubleshooting Guide
- ▸ The Complete ChatGPT Plus Subscription Tutorial — US Setup / Card BIN Choice / Cancellation
- ▸ What Is a Virtual Credit Card — Definition / Types / Uses in One Read
- ▸ The 10 Most Common Reasons Virtual Cards Get Declined — The Complete Self-Rescue Guide
- ▸ RDVCC Visa Virtual Card Product Details
Fact-check window for this article: 2026-04-30 to 2026-06-05. Method: 24 independent sources, 67 raw claims, 3-vote adversarial cross-verification, 18 claims confirmed / 7 killed. Primary sources: the GitHub openai/codex repository (auth-tagged issues), the official OpenAI developer forum, and field-test threads in the V2EX Chinese user community. If you find the situation has changed since reading this, feel free to contact us with an update.
Pain point solved? Try RDVCC Virtual Credit Card
1 USDT issuance · USDT deposits · Works with 100+ international platforms