ChatGPT Deep Research: How Long Does It Take?

A ChatGPT deep research run takes somewhere between five and thirty minutes for most queries. The range is wide because the process is adaptive: the model decides how many search iterations to run based on how complex your question is and how much the sources it finds agree. Keeping your research brief outside the chat — for example in fenbs — means you can step away while it runs and return to find your research question, the sources you already noted, and the next questions all in one place.

Typical duration ranges

For a focused question with a well-defined scope — "What are the current US SEC rules on disclosing AI use in financial products?" — a run often completes in five to ten minutes. For a broad or contested topic where sources disagree and the model needs to reconcile them — "How do major economies regulate large language models?" — expect fifteen to thirty minutes or more.

Server load also affects timing. During peak hours, the same query may take longer than it would overnight. If you need the report for a meeting, start the run with a comfortable buffer.

What affects the duration

• Query scope: a narrow question with a clear answer finishes faster than a broad one with conflicting sources.
• Number of search iterations: the model decides how many passes to run; you cannot set this directly.
• Source accessibility: pages behind paywalls or with slow load times extend the run.
• Contradictions in sources: when pages disagree, the model runs additional searches to reconcile them.
• Server load: high-traffic periods mean longer waits for the same query.

Signs of progress versus a stalled run

ChatGPT shows a progress indicator while deep research is running. It typically displays the current search step or a status message. Seeing this update every minute or two means the run is still active.

A run that stops showing progress for more than ten minutes without completing is likely stalled. In that case, refresh the page and check whether the conversation still shows the run in progress. If the interface shows no activity, the run may have timed out silently. Restart with a narrower question, or contact OpenAI support with the conversation details if the problem recurs.

Do not close the tab or navigate away mid-run unless you need to. Some runs can survive a page reload, but others do not, and there is no reliable way to tell in advance.

How query scope changes duration

The most reliable way to shorten a deep research run is to narrow the question. Instead of "Tell me about AI in healthcare," try "What FDA guidance exists for AI-assisted diagnostic tools as of the current regulatory year?" The second question has a defined scope, a specific agency, and a time frame. The model can plan its searches more efficiently.

If you have a broad topic, break it into pieces. Run one focused query first, read the report, then run a follow-up on the gaps it identifies. This approach takes longer in total but produces more usable reports than one giant query that the model has to interpret.

What to do while the report runs

Deep research runs in the background while you wait. You do not need to stay on the page, though closing the tab entirely can sometimes interrupt it. Opening a second browser tab or window lets you continue other work while monitoring the progress tab.

Use the wait to prepare for what comes next. Write down the two or three specific claims you most need the report to confirm. Note any sources you already know are relevant so you can check whether the report found them. This preparation makes your review of the finished report faster and more focused.

When to stop a run and try a different scope

If a run exceeds forty minutes without completing, or if the progress stops updating, it is worth stopping and restarting with a narrower question. A run that takes too long is often a sign that the question was too broad or that the sources available online do not resolve the question well enough for the model to synthesize a coherent report.

For issues where deep research is taking consistently longer than expected, see chatgpt deep research taking too long.

How to do this with fenbs

1. Create a free account and open a project for the research topic.
2. Before starting the run, add the research question as a task with a note on the specific claims you need confirmed.
3. Write any sources you already know are relevant to Links and listings.
4. While the run is in progress, use the time to add the gaps you expect the report to address as open questions in the task.
5. When the report finishes, update the task with the key findings and save any unresolved questions as the next step.

This workflow means a long run produces something durable rather than a report you will have to re-read to reconstruct later. The research question stays outside the chat and travels with the project.