Skip to content
AI Tools

ChatGPT Phone Verification, Explained in Full

OpenAI's phone rules for ChatGPT and the API have shifted quietly since 2023, and most guides are years out of date. This is the definitive timeline of who needs a number today, why, and how a virtual number fits in.

SX SMSXR.com Research Team 20 min read
Abstract neural network converging on a single verification checkmark in green on a dark background
On this page

You do not need a phone number to use ChatGPT in 2026 — but you almost certainly do to create your first OpenAI API key. That single distinction explains most of the confusion online, where guides written in 2023 still insist a mobile number is mandatory. OpenAI quietly dropped the universal phone requirement for ChatGPT signups by early 2024, yet kept it exactly where abuse risk is highest. This guide is the full timeline of what changed, who is asked today, and why.

Because OpenAI's policy has moved so many times, most articles you find are simply out of date. Below is a single, sourced reference that tracks the phone requirement from its mandatory beginnings through today's much narrower rules, with an honest look at the regional limits, the VoIP block, the little-known reuse ceiling, and where a separate verification number legitimately helps.

The short version: who is actually asked for a number today

Let us settle the practical question first. In 2026, whether OpenAI asks for your phone number depends almost entirely on what you are signing up for and how the system reads your request.

If you are creating an ordinary ChatGPT account to chat, write, or browse, you generally will not be prompted for a phone number at all. An email address — or a linked Google, Microsoft, or Apple account — is enough. OpenAI's own help center confirms that phone verification is no longer a required step for standard ChatGPT signups. A phone prompt can still surface if OpenAI's anti-abuse systems flag your attempt as risky — an unusual network, a datacenter or flagged IP range, or rapid repeated signups from the same device — but it is now the exception rather than the rule. If a friend tells you they were asked and you were not, that is not inconsistency on OpenAI's part; it is risk scoring doing its job differently for two different signups.

If you are a developer creating your first API key on platform.openai.com, that is a different story. Phone verification is still mandatory there. OpenAI ties that first key to a verified mobile number to keep paid, programmatic access accountable: an API key can drive thousands of automated requests, so the company wants at least one real-world identifier behind it. The good news is that this is a one-time gate. Once your first key is verified, later keys generated on the same account do not ask again, and rotating or revoking keys never re-triggers the phone step.

There is also a third, narrower path. In a small set of countries, OpenAI offers phone-only signup, where the number itself is the account identifier rather than an add-on. We will cover that in its own section, because it carries consequences the other two paths do not — including a restriction on enabling multi-factor authentication.

So the honest one-line answer is: chat freely without a number, but expect the request the moment you cross into the paid API — and choose that first number carefully, for reasons that become clear later in this guide.

Three diverging paths splitting from a single entry point, one leading to an open door, one to a locked gate with a key slot, one to a numbered ring, abstract minimal tech illustration on dark charcoal with green accents

How the phone requirement has changed over time

Nobody documents OpenAI's phone-policy evolution end to end, which is exactly why so much bad advice persists. Here is the sourced sequence of events, from the era when a number was non-negotiable to the layered, situational rules of today.

Date What changed Effect on users
Jan 2023 Phone verification mandatory for every ChatGPT signup; one number allowed up to 2 accounts Everyone needed a working mobile number to start
Jan 16, 2023 WhatsApp added as a code-delivery channel in select countries First alternative to plain SMS for receiving codes
Early 2024 Phone requirement quietly dropped for ChatGPT signups; kept only for the first API key Casual users could sign up with email alone
Apr 1, 2024 ChatGPT made available with no login at all for basic use Trying ChatGPT no longer required any account
Jan 15, 2025 Phone-only signup beta introduced, where the number is the account A new, number-first path in limited regions
2026 (today) Phone-only signup live in 13 countries; one-number reuse limit raised to 3; VoIP block expanded internationally Situational, region-aware, abuse-hardened rules

Two threads run through this table. First, the direction of travel has been toward less friction for ordinary users. In January 2023, at the height of ChatGPT's explosive launch, phone verification was a hard wall: no mobile number, no account, and each number could back at most two accounts. By the spring of 2024 the pendulum had swung so far the other way that you could open ChatGPT and start typing with no login whatsoever.

Second, OpenAI has simultaneously tightened the places where friction still matters — the API, VoIP acceptance, and per-number reuse — precisely because those are the vectors abusers exploit. The reuse ceiling actually went up (from two verifications per number to three), a small concession to legitimate users, while the VoIP block went from a US-only measure to an international one. The company loosened the front door while reinforcing the locks that guard paid and high-volume access.

One detail worth naming: the exact day the ChatGPT requirement disappeared was never announced. OpenAI's help article on phone verification was simply rewritten, and archived versions of the signup flow place the change somewhere between late 2023 and the first quarter of 2024 — it was quietly dropped by early 2024. That silence is part of why the internet still carries so many "ChatGPT needs your phone number" pages that were accurate once and simply never updated. When you read any guide on this topic, check its date against the table above before trusting it.

ChatGPT account versus OpenAI API: a requirement matrix

The clearest way to reason about phone verification is to stop treating "OpenAI" as one thing. A ChatGPT account and an OpenAI API account are governed by different rules, even when they share the same login. This matrix lays them side by side.

Requirement ChatGPT account OpenAI API (first key)
Phone number to sign up Not required (email or SSO is enough) Required for the first key
Repeat verification for later use Not applicable Additional keys after the first do not re-verify
VoIP / Google Voice accepted No — mobile SMS only No — mobile SMS only
Email verification Yes, standard Yes, standard
Upgrade to paid tier Verified email needed for Plus/Pro/Team Billing is separate from phone verification
Code delivery SMS, or WhatsApp in 12 countries SMS, or WhatsApp in 12 countries
Can the number be changed later No — changing a linked number is unsupported No — the first-key number stays associated
Reuse of one number across accounts Up to 3 verifications total Counts against the same 3-use ceiling

A few rows deserve emphasis. Upgrading to ChatGPT Plus, Pro, or Team does not add a phone requirement — those tiers need a verified email address and a payment method, not a number. So paying OpenAI money is not what triggers verification; specifically minting your first programmatic API key is. This surprises many teams: a whole company can run on ChatGPT Team without anyone entering a phone number, yet the one engineer who sets up API access will be asked for one.

The "repeat verification" row matters for developers who manage several projects. The phone gate applies to the account's first key, not to every key. Once you have cleared it, you can create, rotate, and delete keys freely. What you cannot do is stretch one phone number across an endless series of accounts — the three-verification ceiling applies globally to the number, no matter which product it was used for.

Notice also what the two columns share: the VoIP block, the SMS/WhatsApp-only delivery, and the permanence of the linked number are identical on both sides. OpenAI runs one phone-verification system with two different entry conditions. If you want to see how this first-key moment compares with other platforms that still gate signup behind a code, our broader service overview lists which apps require a number and which do not.

Regional limits and the VoIP block, without the myths

Two hard constraints trip people up more than any other, and both are official policy rather than rumor.

The VoIP block. OpenAI does not accept landlines, VoIP numbers, Google Voice, or premium-rate numbers for verification. Its help center states plainly that it accepts only mobile numbers capable of receiving an SMS. A second official article adds that OpenAI has "always blocked VoIP in the US" and has since expanded that block internationally. This is why a Google Voice number that worked for some other service will fail here, often with a message like "Detected suspicious behavior from phone numbers similar to yours" or a raw error=enforcement_failed in the page URL. The block is deliberate: VoIP lines are cheap to mass-produce, so they are the first thing an abuse-prevention system rejects. There is no appeal path for a rejected number type — the only fix is a genuine mobile number.

Supported and unsupported countries. OpenAI publishes a supported-countries list of roughly 170 territories. It does not serve a defined set of regions, including China, Hong Kong, Macau, Russia, Belarus, Iran, North Korea, Cuba, Syria, and Venezuela. Attempting to sign up from one of these produces an "unsupported country" error. This is a policy-level geo-restriction applied to the account, not a temporary technical fault, and it governs the account itself rather than just the phone step. We are documenting it here for clarity, not advising anyone to work around it — access rules are OpenAI's to set, and you should follow OpenAI's Terms of Use and your local laws.

Code delivery channels. Verification codes arrive by SMS in most places. In twelve countries, OpenAI can also deliver the six-digit code over WhatsApp instead:

WhatsApp-delivery countries
United Arab Emirates Nigeria
Egypt Pakistan
Indonesia Saudi Arabia
Israel Turkey
India Ukraine
Malaysia Vietnam

Everywhere else, SMS is the only channel — there is no email delivery and no voice-call option for phone verification. Knowing your channel matters in practice: if you are verifying with a number from one of those twelve countries and the SMS seems slow, the code may be waiting in WhatsApp. Conversely, if you deliberately want SMS-only delivery, choosing a number outside that list removes the ambiguity.

An abstract world map rendered as a network of glowing nodes, some nodes bright and connected, a few dimmed and disconnected, representing supported and unsupported regions, minimal tech illustration in green on dark charcoal

The rules that surprise people: reuse limits and the number you can't change

Beyond the signup question sit two policies that quietly cause the most regret, especially for developers, and neither is widely known because both live deep in OpenAI's help center rather than in the signup flow itself.

One number, three verifications. A single phone number can be used to verify OpenAI accounts up to three times. That is an increase from the limit of two enforced back in 2023, but it is still a ceiling, and it counts verifications across ChatGPT and the API platform combined. Push past it and you meet "Too many signups with this number." A number can become reusable again — but only after the account it was tied to is deleted and OpenAI completes its 30-day data-purge window, and only if the number was never banned. In practice, a personal mobile number burned across a few test accounts can lock you out of creating another for a month or more, with no way to accelerate the purge.

You cannot change the number on an existing account. This is the single most consequential and least-known rule, documented in OpenAI's own help article on the subject. OpenAI does not support changing the phone number attached to an account — not through settings, and not through support tickets. Whatever number you verify with is associated with that account indefinitely. For most people that is fine. But it turns the first choice into a permanent one, which is why "just use my personal number" can backfire if you later change carriers and lose the number, want that number kept private, or verified in a hurry on a number you no longer control.

Put those two rules together and a clear principle emerges: choose your verification number deliberately, because OpenAI treats it as a near-permanent identifier with a hard reuse budget. For a developer who only needs to clear the one-time first-API-key gate, using a separate number for that single verification is a reasonable way to keep a personal line out of a corporate AI account — as long as you understand the number stays linked and cannot later be swapped out.

Your SIM, free number sites, or a paid virtual number

Once you know a number is required — for a first API key, or because a risk check demanded one — you have three realistic ways to receive the code. They are not equal, and the differences map directly onto OpenAI's rules described above.

Your own SIM Free public-number sites Paid virtual number (SMSXR)
Works with OpenAI's VoIP block Yes — real mobile line Often no — many are VoIP-based and rejected Yes — real mobile numbers
Code privacy Private to you Public — every visitor sees your code Private to your session
Number freshness Fresh, unless you used it before Heavily recycled, frequently at the 3-use ceiling already Not resold for the same app
Ties your identity to the account Yes, permanently (no later change) No No
Cost Free, but spends your number's 3-use budget Free From ~$0.40 per code, refunded if no code

Your own SIM is the simplest option and perfectly sensible for a personal account you intend to keep forever. Its drawbacks are the two permanence rules: the number can never be detached, and each verification spends one of its three lifetime uses.

Free public-number sites fail against OpenAI more often than against almost any other service, for structural reasons. Their numbers are frequently VoIP-sourced, which OpenAI's filter rejects outright. The numbers that do work are shared with the entire internet, so they tend to arrive already at the three-verification ceiling — and even when a code comes through, it is displayed publicly for anyone to read, which is a poor foundation for an account you cannot later re-secure with a different number.

A paid virtual number addresses each failure mode: mobile-grade lines that pass the VoIP filter, codes visible only to you, and numbers that are not recycled for the same service. The trade-off is a small per-code cost. For a one-time API-key verification, that cost buys a clean number with its reuse budget intact.

Using a separate number for verification: the privacy angle

There is a legitimate reason people look for an alternative to handing OpenAI their personal mobile number, and it has nothing to do with evading rules. It is about not permanently binding a private line to an AI company's account database — a database you have just learned you cannot detach the number from later. Your phone number is one of the most durable identifiers you own; it outlives email addresses and usernames, and it links accounts across services. Keeping it out of systems that do not strictly need it is ordinary data hygiene.

A dedicated verification number solves that cleanly. SMSXR provides real mobile numbers across 183 countries that receive OpenAI's SMS codes, so you can verify without exposing your own number. The service is built around the mechanics of exactly this kind of one-time code:

  • Pay per code, not per rental. You are charged for the code you receive, not for renting a line by the month. As of July 2026, OpenAI-capable numbers start from around $0.40 for the UK and the Netherlands, with pricing spanning roughly $0.30 to higher tiers by country.
  • Automatic 10-minute refund. If no code lands within about ten minutes, the charge is refunded automatically — useful given OpenAI's strict acceptance rules, where a first attempt can occasionally be filtered.
  • Numbers are not resold for the same app. A number used for an OpenAI code is not handed to another customer for the same service, which matters enormously here because of the three-verification ceiling: a number that has already been used elsewhere for OpenAI may arrive with part of its budget spent.

For price context, here are the cheapest OpenAI-capable countries on SMSXR as of July 2026:

Country Price per code
Philippines $0.30
Argentina $0.40
United Kingdom $0.40
Netherlands $0.40
Brazil $0.50

Core Western markets are similarly modest — the US at $1.50, Germany at $1.10, Spain at $1.10, France at $1.70. If you want live availability and the full country list for this service, the OpenAI verification page shows current numbers and rates.

Two honest caveats. First, because OpenAI blocks VoIP, only genuine mobile numbers work — one more reason free VoIP-based services fail here. Second, remember the "can't change it later" rule: if you verify your first API key with a virtual number, that is a fine one-time use, but the association is permanent, so treat it as a considered choice rather than a throwaway.

A single mobile SIM chip at the center of a protective hexagonal shield, with an incoming message pulse arriving from outside, abstract minimal tech illustration in green on dark charcoal

Multi-factor authentication and phone-only accounts

Phone verification at signup is separate from the ongoing security choices on your account, and the two are worth untangling, because conflating them leads people to think they need SMS on their account forever.

For multi-factor authentication (MFA), OpenAI supports a broad set of options: authenticator apps, push approval to a trusted device, SMS or WhatsApp codes, passkeys and hardware security keys such as YubiKey, and a downloadable recovery key. You are not locked into SMS. If privacy is your concern, an authenticator app or a passkey lets you secure the account without leaning on a phone number at all — a meaningfully stronger and more private setup than SMS-based two-factor, since it is immune to SIM-swap attacks and does not depend on you keeping any particular number.

Phone-only accounts are the exception to almost everything above. In the countries where OpenAI offers number-first signup, the phone number is the login identity: you sign in with the number, and password resets arrive by SMS because there is no email on file initially. Those accounts carry a specific documented limitation — they cannot enable MFA until an email address is linked to them. This is the one configuration where your number does triple duty as identity, recovery channel, and (eventually) second factor, which is the strongest argument in this whole guide for thinking carefully about which number you attach.

The phone-only program currently spans thirteen countries — the United States, the Netherlands, France, Spain, the United Kingdom, Poland, Germany, Japan, India, Pakistan, Algeria, Uzbekistan, and Ukraine — and OpenAI has stated that no other regions are planned. So for most of the world, the email-first path remains the default, and phone-only signup is not something you will encounter. If you are in one of the thirteen and see a number-first flow, know that linking an email afterward is what unlocks the full security menu.

Troubleshooting: an error decoder for failed verifications

When a code fails, the message usually tells you exactly which rule you hit. OpenAI's errors are terse but consistent, so here is how to decode the frequent ones and what actually fixes each.

  • "Detected suspicious behavior from phone numbers similar to yours." Almost always a VoIP or flagged-number rejection — the wording is OpenAI's official phrasing for its number-reputation filter. Retrying the same number, waiting, or switching browsers will not help. The fix is a different, genuine mobile number; VoIP and Google Voice cannot clear this check.
  • error=enforcement_failed. The same family of abuse-prevention block, surfaced as a URL parameter rather than a friendly message. It usually accompanies the suspicious-behavior text. Treat it identically: the number type or reputation is the problem, not a transient glitch on your side.
  • "Too many signups with this number." You have crossed the three-verification ceiling on that number. No retry interval fixes this. Use a different mobile number, or — if you control the old accounts — delete one and wait out the 30-day purge window for the number to become eligible again.
  • "Unsupported country." The account region is on OpenAI's unsupported list (China, Hong Kong, Macau, Russia, Belarus, Iran, North Korea, Cuba, Syria, Venezuela, among others). This is a policy restriction on the account, not a phone problem, and no choice of number changes it.
  • No code arriving at all. First confirm whether your number's country is one of the twelve WhatsApp-delivery regions and check that app too — codes there can arrive over WhatsApp instead of SMS. Otherwise, remember there is no email or voice-call fallback: a delivery failure means requesting a fresh code or switching to a different number. Carrier-side filtering of short-code SMS is rare but real; a number from a different country often resolves it immediately.

A practical rule when nothing seems to work: the issue is far more often the number (VoIP-sourced, over-used, or from an awkward route) than your account, your device, or the code itself. Start by ruling out the number, and you will resolve the majority of failures in one step.

What to do next

If you are a casual ChatGPT user, the takeaway is reassuring: sign up with an email or a linked Google, Microsoft, or Apple account and you will most likely never see a phone prompt. The warnings from the launch era simply do not apply to you anymore, and even trying ChatGPT without any account at all has been possible since April 2024.

If you are a developer, plan for the one moment that still matters — the first API key — and decide in advance which number you want permanently tied to that account, since OpenAI will not let you change it and each number carries a three-use lifetime budget. A dedicated mobile verification number keeps your personal line out of the picture for that one-time step, provided you use a real mobile number rather than blocked VoIP.

If you are in an unsupported region, the honest answer is that OpenAI's geo-policy applies at the account level; this guide documents the rule rather than advising a way around it. And if you hit an error mid-verification, the decoder above will almost always name the exact rule involved. Whatever your situation, following OpenAI's Terms of Use and your local laws keeps the account you build on solid ground. With the timeline, the requirement matrix, and the error decoder in one place, you now hold a reference that reflects how OpenAI's phone rules actually work today — not how they worked three years ago.

Frequently asked questions

Do I need a phone number to sign up for ChatGPT?

No. OpenAI quietly dropped the mandatory phone-verification step for ordinary ChatGPT signups by early 2024, so you can typically create an account with just an email address or a Google, Microsoft, or Apple login. A phone prompt may still appear if OpenAI's systems flag your signup as high-risk, but it is no longer a universal requirement.

Why does OpenAI ask for a phone number for the API but not for ChatGPT?

Creating your very first API key on platform.openai.com still requires a verified phone number, because that key unlocks paid, programmatic access that OpenAI wants tied to a real mobile line to limit abuse. Once your first key is verified, additional keys on the same account do not trigger the phone step again. This is the single most common reason people still hit verification in 2026.

Does OpenAI accept VoIP or Google Voice numbers?

No. OpenAI officially blocks landlines, VoIP, Google Voice, and premium-rate numbers, stating it accepts only mobile numbers that can receive an SMS. The block started in the US and has since been expanded internationally, which is why VoIP numbers now trigger errors like 'Detected suspicious behavior from phone numbers similar to yours.'

How many times can one phone number be used to verify OpenAI accounts?

A single number can be used for phone verification up to three times, an increase from the limit of two that applied back in 2023. If you exceed that, you may see a 'Too many signups with this number' error. A number can become reusable again after the associated account is deleted and OpenAI completes its 30-day data purge, provided the number was not banned.

Can I change the phone number on my OpenAI account later?

No. OpenAI does not support changing the phone number attached to an existing account at all. This makes your first choice permanent, so if you verify with a temporary or virtual number for something like a first API key, be aware it stays associated with that account indefinitely.

Which countries can receive OpenAI codes over WhatsApp instead of SMS?

OpenAI can deliver verification codes via WhatsApp in twelve countries: the United Arab Emirates, Egypt, Indonesia, Israel, India, Malaysia, Nigeria, Pakistan, Saudi Arabia, Turkey, Ukraine, and Vietnam. Everywhere else, codes arrive by SMS only, and there is no email or voice-call delivery option for phone verification.

What does the 'unsupported country' error mean?

OpenAI operates in roughly 170 supported countries and does not serve regions including China, Hong Kong, Macau, Russia, Belarus, Iran, North Korea, Cuba, Syria, and Venezuela. If you try to sign up from one of these, you will see an 'unsupported country' error. This is a policy geo-restriction, not a technical glitch, and it applies at the account level.

How much does a virtual number for OpenAI verification cost on SMSXR?

On SMSXR, numbers for OpenAI verification start from around $0.40 for the UK and Netherlands as of July 2026, with a broad range depending on country. You pay per received code rather than renting a line, and if no code arrives within about 10 minutes the charge is refunded automatically.

SX
SMSXR.com Research Team

We test verification flows and price data first-hand, then write it up plainly.

Need a temporary number right now?

Codes in seconds, automatic refund if none arrives, prices from $0.10. No personal number required.

Get started free