← all free guides

free field guide workflows

Run Several ChatGPT Agents at Once

Split research into separate jobs that run at the same time in one ChatGPT chat, then merge the results and flag where they disagree.

Run Several ChatGPT Agents at Once

In one ChatGPT chat, several agents can work on different parts of the same job at the same time. This guide shows you how to set up the jobs, then how to merge what comes back so the disagreements jump out.

What this helps you do

Some research has pieces that don't depend on each other. Comparing three tools, pulling their pricing, and turning your needs into a checklist can all happen at once. Doing them one after another is slow, and reading three separate answers makes it easy to miss where they contradict each other.

With GPT-6 Pro in the chat tab, ChatGPT can hand those pieces to multiple agents that run in parallel. You watch their progress in the conversation and open each finished task to see what that agent did. Then you ask ChatGPT for one merged answer with every conflict flagged. That flag is where you make the call.

Before you start

  • A ChatGPT Pro plan. Parallel agents in GPT-6 Pro are rolling out to eligible Pro users. On a managed work account, availability depends on your workspace settings and the rollout.
  • GPT-6 Pro selected. In the model menu, choose GPT-6, then the Pro option. OpenAI explains the options in GPT-6 and other models in ChatGPT.
  • A decision with separate parts. Write down the options you're choosing between and what matters most to you, like budget, team size, or must-have features.

Terms to know

  • Agent: an AI worker that takes a task, uses tools like web search, and reports back.
  • Parallel: at the same time, instead of one after another.
  • GPT-6 Pro: the Pro option for GPT-6 in ChatGPT's model menu. It's the version that can run multiple agents in the chat tab.
  • Independent jobs: tasks that don't need each other's results. Parallel agents work best on these.
  • Merge: combining several answers into one, with duplicates removed and conflicts called out.

Step-by-step

  1. Open a new chat and pick GPT-6 Pro from the model menu.
  2. Paste the first prompt from the Copy this section. Fill in your options and requirements. Describe each job separately so ChatGPT can split them cleanly.
  3. Watch the progress. You'll see the agents' work appear in the conversation as they go. You don't need to do anything while they run.
  4. Open each completed task. Check what each agent did and which sources it used. If one agent pulled pricing from a blog instead of the official site, note it.
  5. Paste the merge prompt. This asks ChatGPT to put everything together and list every place the findings or sources disagree.
  6. Resolve the flags yourself. For each disagreement, open the official source and decide. This is the step that keeps you from acting on a wrong number.

Copy this

Use this to start the parallel research. Replace each bracketed part.

Copy this

I'm choosing between [OPTION 1], [OPTION 2], and [OPTION 3] for [WHAT YOU NEED IT FOR].

These three jobs are independent. Work on them in parallel:

Job 1: Compare the three options on features, ease of setup, and who each one is best for.
Job 2: Pull current pricing, plan limits, and free-tier details from each official website only. Link the exact page for every number.
Job 3: Turn my requirements into a yes/no checklist I can score each option against. My requirements: [YOUR REQUIREMENTS].

Report each job separately when it's done.

When all three jobs are finished, paste this in the same chat:

Copy this

Merge the findings from all three jobs into one answer.

1. Score each option against the Job 3 checklist in a table.
2. Make a separate list called "Disagreements" with every place where the jobs, or their sources, don't match. For each one, show both versions, both sources, and which source is the official one.
3. List anything you couldn't confirm from an official source.
4. Don't pick a winner yet. Tell me the two or three questions I need to answer to decide.

Worked example: planning a trip

Trip planning has pieces that don't depend on each other, so it splits cleanly. Paste this first, with your own details.

Copy this

I'm planning a trip to [DESTINATION] from [START DATE] to [END DATE], leaving from [YOUR CITY]. My budget is [BUDGET].

These three jobs are independent. Work on them in parallel:

Job 1: Find flight options for those dates. List airline, times, stops, and price, and link the exact page for each.
Job 2: Find hotels that fit my budget. List the nightly price, the location, and the cancellation terms, and link the official booking page.
Job 3: Check what's open while I'm there, such as attractions, restaurants, and any seasonal closures or holidays. Use official sites and link each one.

Report each job separately when it's done.

When all three are finished, paste the merge prompt in the same chat:

Copy this

Merge the flights, hotels, and what's open into one trip plan.

1. Put it in a day-by-day outline.
2. Make a separate list called "Disagreements" with every place where the jobs, or their sources, don't match, such as a hotel check-in that lands after your flight, or an attraction listed as open on a day its official page says it's closed. Show both versions and both sources.
3. List anything you couldn't confirm from an official source.
4. Don't book anything. Tell me the two or three questions I need to answer to decide.

Open the official page for each disagreement before you book anything.

Things to know

  • Only split work that's truly separate. If job two needs job one's answer, they can't run at the same time. Ask for those in order instead.
  • ChatGPT decides how to divide the work. Describing each job clearly is the best way to get the split you want.
  • The merge is the payoff. Three good answers that contradict each other without saying so are worse than one answer with the conflict marked. Always ask for the disagreements list.
  • Official sources win. For prices and limits, trust the company's own pricing or help page over any summary, and open it yourself before you pay.
  • Save it for real decisions. For a quick question, a normal chat is faster. Parallel agents pay off when a choice has several separate parts to research.

If something goes wrong

  • You only see one answer, not separate agents: check that GPT-6 Pro is selected, and that your jobs are written as separate, independent tasks. The feature is still rolling out, so it may not be on your account yet.
  • The jobs blur together: start a new chat and number each job, as in the prompt above.
  • Prices look wrong or old: ask ChatGPT to recheck that job using only the official pricing page, and link it.
  • No disagreements were listed: ask, "Check again. Where do any two sources give different numbers or claims?"

Final check

You pass if you have a scored comparison table, a "Disagreements" list with sources for both sides, and you've opened the official page for every disagreement before deciding.

paid brand collaborations

teamwork makes the dream work.

I love working with brands that make AI easier to understand, operate ethically, and share what they know generously. If that sounds like you, drop your info below.

is this a paid opportunity?