Which tool to choose
For this specific task — reasoning about a choice, weighing arguments, returning an ordered table — you don't need a subscription. All three of the big assistants do it for free and well. The choice depends on what you already have on hand.
A mental decision, no documents (changing jobs, cities, a purchase): use Claude. The free version runs on a capable model and gives access to Projects and Artifacts. The Artifact returns the table to you in a separate panel that you can edit and copy out cleanly.
You have a document to analyze (a contract, two quotes, a specification): use Gemini. The large free context handles long files; for documents that exceed the free plan's window you need a subscription, but for a single contract you won't get there.
You also want to look for confirmation online (reviews, market data, prices): use ChatGPT. Free accounts can search the web for up-to-date information, analyze data, and upload files.
A warning about the free limits: they're tight. Free accounts have a message cap on the best model within a window of a few hours; once you exceed the cap, the conversation switches to a lighter model until the reset. For a single decision they're enough. If the conversation drags on, save the table by copying it out before you bump into the limit.
How to do it
The path doesn't change whether you're in a browser or an app: open the assistant, paste the prompt, answer the questions it asks you.
Define the decision in one sharp sentence. Not "the job," but "accept the Milan offer at 38k or stay in the current position at 32k." The more precise the choice, the less the AI invents context.
Write your real priorities, in order. Salary? Proximity to home? Growth? Without this, the AI uses average textbook priorities that aren't yours.
Paste this prompt. The operating syntax:
I have to make this decision: [write the choice here in one sentence, with the options].
My priorities, in order of importance:
1. ...
2. ...
3. ...
Non-negotiable constraints: [budget, timing, family, health, etc.]
Before answering, ask me at most three questions to understand my
real situation. After my answers, build a table with:
- a PROS column and a CONS column for each option
- a weight from 1 to 5 for each item, based on my priorities
- a final row with the total score for each option
At the end, indicate which option turns out most aligned with my priorities
and clearly state what would make it change. Use only the information
I've given you; if you're missing a piece of data, ask me instead of inventing it.
Answer the questions it asks you. It's the step that separates a generic table from one tailored to you.
Check: the table is useful if every item traces back to a priority or a constraint of yours. If you see arguments that have nothing to do with your life ("great for networking" when you don't care at all), write:
Remove the items that don't fall within my priorities and reweight the table.
- Stress the result. Add:
Now play devil's advocate against the winning option: what's
the scenario in which I regret having chosen it?
This is where the con you hadn't considered often comes out.
A concrete example
Marta, a freelance graphic designer, has to decide whether to take a spot in a coworking space at 180 euros a month or keep working from home. Stated priorities: 1) concentration, 2) low costs, 3) contact with other professionals.
The AI asks her three questions: how many hours does she work from home without interruptions? Does she already have clients from word of mouth? How much do 180 euros weigh on her monthly revenue? Marta answers: barely two hours because of the kids, yes, word of mouth brings her work, 180 euros are 9% of her revenue.
The table that comes out: for "home" the heaviest pro is zero cost (weight 4), the heaviest con is zero concentration with the kids (weight 5). For "coworking" the heaviest pro is guaranteed concentration (weight 5) plus new contacts (weight 4), the con is the cost (weight 4). Total: coworking 13, home 7. In the role of devil's advocate the AI adds the real risk: if clients decline, 180 fixed euros become a burden, so it suggests a pay-per-day formula instead of the full subscription. Marta chooses pay-as-you-go coworking: three days a week, cost halved, concentration saved. Decision made in twenty minutes, with the doubt about costs resolved instead of ignored.
When it does NOT work (and how to fix it)
If the AI invents numbers or data you didn't give it
If average salaries, prices, or statistics you didn't provide pop up in the table, they're made-up estimates, not facts. Rewrite:
Use only the numbers I gave you. For the data you're missing, write
'to be verified' instead of estimating them.
For real public data (average rental price in an area, the cost of a service) use ChatGPT with web search on, so it cites a source instead of guessing.
If it agrees with you on everything
Assistants tend to confirm the position that shows through in your question. If you wrote the prompt making it clear what you'd want, the table will go along with it. Fix: ask for two separate tables, one defending option A and one defending option B, then compare them. The difference between the two shows you where you're cheating with yourself.
If the decision is too big for a single table
For choices that intertwine money, family, and health together, a single table flattens everything. Break it up: ask for one table for each dimension ("only economic aspects," then "only personal aspects"), and at the end:
Merge the previous tables, giving each dimension a weight, and
recalculate the total score.
You avoid letting the economic factor, easier to quantify, crush the others.
If you bump into the message limit halfway through the reasoning
It happens with the free version during a long conversation. On ChatGPT, once you reach the limit you're switched to a less powerful model. So you don't lose the thread, copy the table into a text file as soon as it's ready. Then you can pick it up after the reset, or paste it into another assistant and continue there without starting over from scratch.
A tip from someone who really uses it
Never ask the AI "what should I do." Ask it to build the table and weight it, then you decide by looking at the score. The real value isn't the final number: it's that the 1-to-5 weight forces you to declare what really matters. Often, while you assign the weights (or while you argue with the ones the AI put in), you realize you've already decided in your gut. The table serves to bring that decision to the surface, not to replace it.
Frequently asked questions
Which AI is best suited for important decisions like changing jobs?
For pure reasoning, free Claude works fine: the model handles complex argumentation and the Artifact gives you the table in an editable panel. If you want to add real data alongside (average salaries, the cost of living in a city), open ChatGPT in parallel with web search. There's no "right" AI: there's the combination of reasoning and fact-checking.
Do I have to pay for a subscription to make a good pros-and-cons list?
No. This task requires neither top models nor premium features. The free limits — a handful of messages every few hours — cover a single decision. A subscription makes sense only if you make complex decisions every day or work on very long documents that exceed the free context.
Can I trust the weights the AI assigns?
The weights are a starting point to challenge, not a verdict. The AI sets them based on the priority order you gave it, but it doesn't know how much an emotional factor really weighs for you. Treat them as a draft: change by hand any number that feels wrong. Often it's precisely in correcting a weight that you understand what you want.
Isn't it better to decide with your gut instead of relying on a machine?
Here the AI doesn't decide for you, and using it doesn't switch off your instinct: it puts it to the test. The table forces you to write down in black and white the cons that the gut tends to hide when a choice attracts you. If after seeing the score your gut screams the opposite, that's valuable information: it means a priority you didn't declare counts more than you believed. The decision stays yours, but made with all the cards on the table.