Your boss wants “a quick SWOT” for Thursday. Or you are about to raise prices, hire someone, or launch a new service, and you want to think it through properly before you bet money on it. You open a blank 2×2 grid, type “Strengths” in the corner, and stare at it.
Here is how to do a SWOT analysis with AI without getting generic filler: give ChatGPT or Claude one clear decision and your real material (numbers, customer comments, competitor pages), let it interview you before it writes anything, have it sort what you know into the four boxes with the evidence next to each point, then make it argue against its own list. Finish by turning the grid into three actions with a TOWS matrix. A solid first version takes about 45 minutes, and most of that time is you adding facts the AI cannot know.
Below: the five steps, prompts you can copy, a worked example, a version for your own career, and the mistakes that make an AI SWOT useless.
No password needed – your reset opens instantly.
No password. Just your email, and you’re in.
Why most AI SWOT analyses are useless (and how to fix that)
Type “do a SWOT analysis for my bakery” into any chatbot and you get a neat grid in seconds. Strengths: “quality products, loyal customers.” Threats: “competition, economic uncertainty.” It looks finished. It would also fit every bakery on the planet.
That is not the AI being lazy. It simply knows nothing about your bakery, so it fills the boxes with the most likely words. And it says them with confidence. OpenAI’s own help page warns that ChatGPT can sound sure of itself even when it is wrong, and can oversimplify complex issues (OpenAI Help, Does ChatGPT tell the truth?).
MindTools lists the classic SWOT mistakes: lists that are too long, points that are vague, not seeing your own weaknesses, and treating the SWOT as the whole plan (MindTools, SWOT Analysis). A chatbot left on its own makes every one of them.
The fix is a change of roles. You bring the facts. The AI brings the structure, the questions you forgot to ask, and the critic you do not have time to find. That split is what the steps below are built on.
What you need before you start
- A chat assistant. ChatGPT, Claude or Gemini all work. The free plans are enough for a first SWOT. New to them? Our ChatGPT tutorial for beginners and Claude tutorial for beginners get you set up.
- One decision. Not “our business” in general. Something like “should we add a lunch menu” or “should I take on retainer clients”.
- Your raw material. Whatever you have: last year’s sales by month, a handful of customer reviews, survey answers, a competitor’s pricing page, notes from a team meeting. Messy is fine.
- 45 minutes and, if you can, one colleague who will tell you the truth.
How to do a SWOT analysis with AI in 5 steps
Step 1: Name the one decision (5 minutes)
A SWOT about “the company” turns into a list of everything. A SWOT about one decision gives you something to act on. MindTools makes the same point: apply it at the right level, such as a product or product line, rather than the vague whole-company level.
Write one sentence: “We are deciding whether to ___ in the next ___ months.” That sentence goes at the top of every prompt.
Step 2: Let the AI interview you first (10 minutes)
This is the step that kills the generic grid. Instead of asking for the SWOT, ask for questions.
[your one sentence]
About us: [2 to 4 lines: what you sell, to whom, team size, where]
Do not write the SWOT yet. First ask me up to 10 short questions, one group at a time, that you need answered to make each box specific to us. Ask about numbers, customers, competitors and what we are bad at. If I don’t know an answer, note it as a gap instead of guessing.
Answer in plain words. Voice typing helps if you think better out loud. Paste in the material you gathered too: reviews, a survey export, a spreadsheet summary. If you have a long report or PDF, our guide on analyzing long PDF files with Claude shows how to feed it in without losing the details.
Step 3: Build the grid with evidence next to every point (10 minutes)
Now ask for the SWOT, with one rule: every point has to say where it came from.
Rules:
• Strengths and Weaknesses are internal (things we control). Opportunities and Threats are external (market, customers, competitors, rules, technology).
• Max 5 points per box. Pick the ones that matter most for THIS decision.
• Each point must be specific enough to disagree with. No “good quality” or “competition”.
• After each point, in brackets, write the evidence: what I told you or what is in the material. If there is no evidence, write [ASSUMPTION].
• Number them S1, S2, W1, O1, T1 and so on.
• End with a list of the open questions that would change the picture most.
The [ASSUMPTION] tag is the useful part. It shows you, at a glance, which points are facts and which are guesses dressed up as facts. MindTools gives a good test for this: accept precise, checkable statements like a specific cost advantage, not “better value for money”.
A note on Opportunities and Threats: these come from outside, so the AI’s training data is often out of date. If your assistant has web search turned on, ask it to look up current information and show the links, then open the links yourself. OpenAI’s advice is to use search when accuracy matters and to verify by visiting the sources. For a deeper look at a market, follow our guide on how to use AI for market research first and bring the results here.
Step 4: Make the AI argue with itself (10 minutes)
Most teams are too kind to their own strengths and too vague about their weaknesses. A chatbot has no feelings to protect, so give it the job nobody in the meeting wants.
Go through the SWOT point by point:
1. Which “strengths” are just normal in our market (every competitor has them)?
2. Which points are in the wrong box (internal vs external)?
3. Which weaknesses are missing? Ask what a frustrated customer or a new employee would say.
4. Which threats are we underestimating?
Then give me a revised SWOT, still max 5 points per box, with the evidence tags kept.
That first question is straight from MindTools: a feature is only a strength if it gives you a clear advantage. If all your competitors also deliver on time, on-time delivery is the entry ticket, not a strength.
Then do the human part. Show the grid to one colleague, or a customer you trust, and ask: “What’s wrong with this?” The AI can challenge your list. Only people who see your business from the outside can tell you what it actually feels like.
Step 5: Turn the grid into actions with a TOWS matrix (10 minutes)
A SWOT is a snapshot. On its own it does not tell you what to do, which is why so many end up in an appendix. The TOWS matrix, developed by Heinz Weihrich in 1982, fixes that by crossing the boxes (MindTools, The TOWS Matrix):
- SO: use a strength to grab an opportunity.
- ST: use a strength to blunt a threat.
- WO: fix a weakness so you can take an opportunity.
- WT: defensive moves that shrink a weakness and avoid a threat. MindTools notes these protect you from loss but will not create success on their own.
For each of SO, ST, WO and WT, suggest 2 concrete actions. Each action must name the labels it is built from (for example S1 + O2). No action without labels.
Then pick the 3 actions with the best effort-to-impact ratio for a small team, and for each one give: the first step we can take this week, what it costs roughly in time, and the signal that tells us within 30 days whether it is working.
The labels keep the AI honest: an action that cannot point at a box came from nowhere. You end with three actions, not sixteen bullet points. If one of them is a bigger project, our guide on using AI to create a project plan takes it from there.
A worked example: a two-person design studio
This example is made up to show the process. The numbers are illustrative, not data.
Decision: “Should we offer monthly retainer packages instead of only one-off projects, starting in January?”
In Step 2, the AI asked eight questions. The ones that changed the picture: “How many past clients came back for a second project?” (answer: about half), “What do clients complain about?” (slow replies when both designers are busy), and “Do competitors near you offer retainers?” (they did not know, which became an open question).
After Step 4, the grid looked like this:
- S1 About half of past clients returned for a second project (studio’s own records).
- S2 Clients praise the brand workshops in reviews (12 Google reviews mention them).
- W1 Reply times slip to several days in busy weeks (client complaints).
- W2 No written process, so work depends on the two founders [critic added this].
- O1 Several returning clients asked for “someone on call” for small updates (emails).
- T1 Clients can now make simple social graphics themselves with AI tools [ASSUMPTION, check with clients].
The critic moved “creative, high-quality design” out of Strengths, because every studio says that. It also flagged that T1 rested on a guess, so the studio asked three clients. Two said yes, they now do small graphics themselves.
The three TOWS actions they kept:
- SO (S1 + O1): offer a small retainer to the returning clients who asked for it, before building a public package.
- WO (W1 + O1): a retainer includes a reply-time promise, so they first write a simple support process with AI (our guide on writing SOPs with AI covers this).
- ST (S2 + T1): lead the retainer with strategy and workshops, the part clients cannot do with a template tool.
Total time at the keyboard: under an hour. The most valuable minute was the critic pointing out the missing weakness.
Personal SWOT with AI: for your career or a job change
The same method works for yourself. MindTools describes a personal SWOT for career decisions: what you do better than most, what skills you lack, what is changing in your industry, and what puts your role at risk (MindTools, Personal SWOT Analysis).
Your evidence here is your own history: performance reviews, feedback, projects you shipped. Use the same prompts, swap “us” for “me”, and change the decision line to something like “Should I apply for the team lead role?” One extra instruction helps a lot:
The result feeds straight into a self-evaluation or your interview preparation.
SWOT your competitor (carefully)
MindTools also suggests running a SWOT on your competitors. AI makes that faster: paste in their pricing page, their public reviews and their job ads, and ask for their SWOT with evidence tags. The US Small Business Administration lists what a competitive analysis should look at, including competitors’ strengths and weaknesses, your window of opportunity, and barriers to entering the market (SBA, Market research and competitive analysis).
Two cautions. Only use public information. And treat everything about a competitor’s inside (their costs, their team, their plans) as [ASSUMPTION] unless you have a real source for it.
Mistakes that make an AI SWOT worthless
- Asking for the SWOT in your first prompt. You get the generic version. Interview first.
- Accepting points without evidence. If a box is full of [ASSUMPTION] tags, you have a to-do list for research, not an analysis.
- Letting the boxes grow. Five points each, maximum. A longer list just hides the points that matter.
- Mixing internal and external. “A new competitor opened nearby” is a threat, not a weakness. The critic step catches most of these.
- Stopping at the grid. No TOWS step, no actions. The grid alone changes nothing on Monday morning.
- Pasting in confidential numbers without checking your company’s AI policy. If you are not sure, round or anonymize them. The analysis works just as well with “revenue down about 10%” as with the exact figure.
FAQ
Can ChatGPT do a SWOT analysis?
Yes, and quickly. But on its own it can only produce a generic one, because it does not know your business. Give it your facts and one decision, ask it to interview you first, and it becomes genuinely useful. Check anything it claims about the outside world, because it can be wrong while sounding certain.
Is there a free AI SWOT analysis generator?
The free plans of ChatGPT, Claude and Gemini work as a SWOT generator with the prompts above. Dedicated generator websites mostly do the same thing with a form in front. The quality depends on the facts you put in, not on the tool.
What is the best prompt for a SWOT analysis?
The one that forbids guessing. The key lines are: one decision at the top, “ask me questions first”, max 5 points per box, evidence after each point, and [ASSUMPTION] when there is none. The Step 3 prompt above has all of them.
How do I present the SWOT?
One slide with the grid, one slide with the three TOWS actions. Ask the AI to cut each point to one line for the slide. Our guide on making a presentation with AI covers the rest.
How often should I redo it?
When the decision changes or the market does. Because you saved the prompts and the evidence, updating it later takes minutes, not a new workshop.
If your SWOT leaves you with two or three options and no clear winner, score them side by side with a decision matrix built with AI.
Your next step
Pick one decision you are sitting on right now. Write the one sentence, paste the Step 2 prompt, and answer the questions. In 45 minutes you will have a SWOT that is actually about you, and three things to do next week.
Want more walk-throughs like this, where AI does the busywork and you stay in charge? Launchpad is free.
No password needed – your reset opens instantly.
No password. Just your email, and you’re in.
Or start on our home page and see what else you can hand off to AI this week.