Answers
Does generation stop when you close the tab?
Closing a tab is a browser event, not necessarily the end of the work you started. Here is a practical way to check what happened to a long-running generation, using only what you can observe in your own account.
If you are asking whether generation stops when you close the tab, the honest answer is that the verified facts available do not cover this behaviour. We cannot say the work continues, and we cannot say it stops. Picture the situation: you start a long generation, you need to switch tabs or devices, and you want to know whether the result will be there when you come back. The question is about the status of a persistent job, not about the quality of the output. Since the available facts do not settle it, the practical route is to look for observable signals before and after you close the tab, without assuming the product does something it has not been confirmed to do.
Why the question stays open
Closing a tab is a browser event. It is not necessarily an event in the job you started. The question is not covered by the verified facts available, so we avoid both optimism and doom. What you can do is run an operational check. Treat it as a method proposed here, not as a universal standard.
A practical check for the status of a persistent job
The steps below are a suggested procedure, written to reduce uncertainty using things you can actually observe. They do not describe how the product works internally.
- Note the start time and the job identifier. What counts as a check: you have a time reference and something recognisable (a title, a prompt, a visible identifier) that lets you find the same job after closing the tab.
- Before closing, record the visible status. What counts as a check: a screenshot or a note showing the status on screen (running, completed, waiting). If no status is visible, write that down as no signal. Do not read it as zero or as an error.
- Reopen the same page or the same account and compare. What counts as a check: does the job appear under the same identifier? Has the status changed since your note? If it does not appear, do not jump to the conclusion that it is lost. First confirm you are in the same account and the same section.
- If the job cannot be recovered, decide whether to recreate it. What counts as a check: you have ruled out account and section problems. Only then is recreating it a practical option, not something you infer from how the product behaves.
RankAI's contribution: tell a missing signal apart from a negative result. Without documentation on tab behaviour, the risk is mixing up three different things: the job finished, the job is still running but not visible, the job was never persistent. The method above is meant to separate them with observations, not guesses. One useful criterion is repeatability: if you run the same sequence (start, note the status, close, reopen) and get the same result, you have a stable indication for your case, not a general rule. A second criterion is account consistency: many false alarms come from reopening a different session. Finally, do not treat a missing indicator as proof that nothing was saved. In the product facts, when a visibility figure is unavailable, it is shown as N/A, never as zero. That does not answer the tab question, but it stops you from reading an empty space as a confirmation.
A limit you can verify
The limit is simple: the facts available do not settle what happens to a generation when the tab closes. Any firm conclusion either way would be an unsupported inference. You can verify only what you observe in your own account, using the criteria above.
When a measurement is not available, remember that RankAI shows it as N/A, never as zero. For more on frequently asked questions and product limits, see the FAQ page.