OpenAI is quietly testing a feature that lets ChatGPT users pay to undo their own usage limits. Hit the weekly cap on a $20 Plus plan and, for some users, a prompt now appears offering to restore the allowance to full for roughly $8. On the $200 Pro plan, the equivalent reset is reported to run up to about $80. The company never announced it; it was discovered by a subscriber who ran into it at the point of being locked out, and an OpenAI spokesperson later confirmed the company is exploring ways for capped users to buy more usage.
The answer-first version: your flat monthly subscription now has a pay-to-continue button, and it shows up at the worst possible moment. The reset restores your usage to 100% and pushes the next weekly renewal about seven days out. It is cheaper than upgrading, which is the point — but it is also a new charge that did not exist a month ago, applied to a limit most users cannot see coming, offered at the instant they are least able to say no.
The mechanics, as reported, are straightforward. When a ChatGPT Plus subscriber exhausts their weekly message allowance, instead of only being told to wait, some accounts now see an option to pay to reset. Redeeming it restores usage to full and resets the weekly clock. The price sits at around $8 for Plus; on Pro it scales up to roughly $80, still framed as a stopgap against a full plan change. The feature was first surfaced by a Reddit user on the $20 plan who described a black prompt appearing at login once their allowance ran dry — not a setting they went looking for, but one that found them. That detail — a user, not a press release, breaking the news of a paid feature — is itself worth noting: the first public account of how OpenAI plans to charge for extra usage came from someone who had already been charged the inconvenience of being locked out.
OpenAI has not disputed the reports. A spokesperson described the effort as exploring ways for people who exhaust their caps to buy additional usage, part of a broader push to make its products more functional, and called it an early experiment with no plans for a broad rollout. That last clause matters: this is a test, it may never ship widely, and the prices may move. But it is live enough that paying customers are hitting it in the wild.
There is a real case for this, and it is worth stating plainly before taking it apart. Inference costs OpenAI money on every request, and the heaviest users cost far more than a flat $20 covers. Faced with that, a company has three honest options: raise everyone’s price, cut everyone’s allowance, or let the heavy users pay for what they actually consume. A cheap, optional top-up is arguably the most consumer-friendly of the three — the light user is not subsidising the power user, and the person facing a deadline can pay $8 to finish rather than being told to come back next week or find $180 to upgrade to Pro.
Seen that way, the reset is a kindness: a low-cost escape hatch from a hard wall. Plenty of metered services work exactly like this, and nobody calls a data top-up a scandal. If OpenAI shipped this as a clearly documented, opt-in feature with the weekly cap shown in plain numbers, it would be an unremarkable and defensible piece of product design.
It did not ship that way. Three things turn a defensible top-up into something that reads as a pressure tactic, and none of them is the price.
First, it was undisclosed. Users are discovering it by hitting it, not by reading a changelog. A charge you learn about only when it is offered to you is, by design, a charge you cannot plan around. Second, the underlying limit is illegible. As we have written before, the flat AI subscription is quietly becoming a meter, and this is the meter made explicit — except the dial has no numbers. ChatGPT’s weekly allowance is not shown to Plus users as an absolute quantity, so you cannot see how close you are to the wall until you are through it. Third, and most pointedly, the offer arrives at the moment of maximum friction. You are mid-task, the work is unfinished, and the button that makes the problem disappear costs $8. That is not a neutral menu choice; it is a decision extracted under pressure.
A top-up you chose in advance is a feature. A top-up that appears the instant you are locked out, priced against work you have already started, is a toll. The number is the same; the position you are in when you pay it is not.
This does not arrive out of nowhere. OpenAI has already built the plumbing for paying past a limit elsewhere in its product line. In Codex, its coding agent, hitting a usage cap on Plus or Pro prompts you to buy credits to keep going, and eligible users can switch on auto top-up so their balance refills automatically when it runs low. The ChatGPT reset is the same instinct pointed at the consumer chat product: when you run out, the answer is increasingly a purchase rather than a wait.
The direction of travel is consistent, and it is the thing worth watching. Each individual step is small and defensible — a credit here, a reset there, all optional, all cheap. The cumulative effect is that the meaning of a subscription changes underneath the people paying for it. You bought a flat plan; you are being eased, one prompt at a time, toward pay-as-you-go, without the plan’s name ever changing. We made a version of this argument about the hidden cost of AI tokens; the reset button is that hidden cost surfacing as a literal button.
Strip away the reset button and the deeper grievance is the limit it attaches to. A Plus subscriber is not shown their weekly allowance as a number they can watch tick down; they are shown it, in effect, only once — at zero, when the work stops. Every other metered service you use tells you where you stand. Your phone plan shows the gigabytes remaining. Your electricity meter has a dial. ChatGPT’s weekly cap, for consumer tiers, is a wall you discover by walking into it, and the pay-to-reset prompt is the first time many users learn a hard weekly cap was even there.
That is what turns an ordinary top-up into something sharper. If the allowance were visible, you could ration it, plan around it, or decide in a calm moment whether the reset was worth $8. Because it is invisible, the first decision point is also the worst one: mid-task, blocked, with the fix one click away. The charge is small; the absence of information around it is not. A meter you can read lets you make choices. A meter you cannot read makes them for you, and then offers to sell you out of the corner it put you in.
None of this is a reason to panic, and $8 to finish a deadline may genuinely be worth it to you. But it is worth being deliberate rather than reflexive about it:
OpenAI is not doing anything a metered business is not entitled to do, and the price is not the scandal. The objection is narrower and more durable than outrage: a subscription is a promise about predictability, and a pay-to-continue button that appears unannounced, against an invisible limit, at the moment you are most stuck, quietly breaks that promise while keeping the subscription’s name. You can still choose to pay. It is worth noticing that the choice was arranged so that you would.
Originally published at theaidownside.com — evidence-first reporting on the costs and trade-offs behind AI products.