Which tool to choose
For the SWOT itself any generalist assistant will do: the value is in the prompt, not in the tool. A useful distinction:
- For reasoning about the idea (internal strengths and weaknesses): Claude and ChatGPT are solid and can play devil's advocate if you ask them.
- For the external opportunities and threats that depend on the real market: use an AI with web search (the online versions of ChatGPT or Gemini, or Perplexity, which cites its sources, or Grok for what's moving right now). That way the threats aren't invented but anchored to verifiable facts.
How to do it
The path is identical on computer and on phone: you write the idea, ask for the table, then put the items to the test.
- Summarize the idea in 3-4 lines. What you offer, to whom, how you make money, what sets you apart. A SWOT on a vague description produces a vague table.
- Ask for the table with the remedy built in. The working syntax:
Act as a skeptical strategy consultant. I'm doing a SWOT analysis of this idea: I sell online cooking courses to people who live abroad and miss Italian cooking; I make money through monthly subscriptions; I stand out because I teach little-known regional recipes. Return a table with four columns — Strengths, Weaknesses, Opportunities, Threats — with 3-4 concrete items specific to my idea, not generic. For each Weakness and each Threat add a "what I would do" line with a practical action. Don't sugarcoat: if there's an obvious weak point, put it first.
- Challenge the SWOT. After the table, ask: "which is the threat I'm most underestimating, and why?" It's the question that brings out the risk no one wants to look at.
- Verify the numbers. Every market figure the AI cites must be checked against a real source before you use it to decide.
A concrete example
Davide wants to open a scooter rental service in a seaside town. He writes the idea to ChatGPT with the prompt above. The table shows him strengths he already sensed (little local competition, summer tourism) but also two weaknesses he'd suppressed: the seasonality that leaves six dead months and the maintenance cost of vandalized vehicles.
Under each weakness the AI proposes a remedy: deals with hotels for income even in low season, and a security deposit via app. Davide then asks for the most underestimated threat: the AI points to a possible change in the municipal regulation on scooters, which he hadn't considered. Davide phones the town council before investing. The SWOT didn't decide for him, it showed him where to look.
When it does NOT work (and how to fix it)
If the SWOT is obvious and generic
If you read items like "opportunity: growth of the digital market," the AI has worked on nothing. Put specific details back into the prompt — your city, your price, your typical customer — and ask for items "that hold only for my idea and not for any company."
If AI invents market data
A "plausible" growth percentage or sector size can be entirely made up. Treat every number as a suspect: ask the AI for the source, and if it doesn't provide one or it doesn't hold up, discard the figure. For external opportunities and threats use an AI that searches the web and cites.
If AI agrees with you on everything
The models tend to please whoever's writing: it's the flaw known as sycophancy. If the table is all rosy, force it: "play devil's advocate, write the SWOT as a competitor of mine would see it, one who wants to prove I'll fail." The hostile version brings out what the polite one hides.
A tip from someone who really uses it
Do the SWOT twice with two different AIs, on the same description. Where they agree, you've found the solid points; where they diverge, you've found the gray areas to think about more. Comparing two answers costs two minutes and protects you from taking the first table you read as gospel.
Frequently asked questions
What do I have to give the AI for a good SWOT?
The essentials in a few lines: what you offer, to whom, how you make money and who the competitors are. The more concrete the picture, the more the four boxes fill with real things instead of clichés.
Is an AI's SWOT worth as much as a consultant's?
It's an excellent starting point, not a substitute. A consultant knows your sector and puts the responsibility of the advice on the line; AI lines up quick hypotheses. Use it to arrive prepared at the consultant's, not to avoid them when the stakes are high.
Can I trust the SWOT to decide whether to invest?
No, and here's the misunderstanding to clear up: the SWOT is a map of the risks, not a green light. It shows where to look, but the decision to invest weighs money, time and consequences that only you can assess. AI lists the threats; you're the one who chooses which ones you're willing to take.