The Prompt Stack.
A method for getting answers out of AI you can act on — instead of the confident, fluent, sometimes-wrong ones it hands you by default. One quick check on any answer, then four questions you run in order. It works on a holiday deal or a doctor's letter; I push it hardest where a wrong answer costs real money.
- ROLE
A job title buys nothing. Set the stance you want it judged by — sceptical, not a cheerleader.
- FILTER
Split what it knows from what it's filling in.
- RISK
Name the one thing that would prove the answer wrong — something you can check.
- VERDICT
One answer, with a confidence level. No fence-sitting.
// On this page
How do you tell if an AI answer is true before you act on it?
AI sounds exactly as sure when it's wrong as when it's right. A made-up figure arrives in the same calm, helpful voice as a correct one, so tone tells you nothing — and tone is what most of us are quietly judging on. So before any of the four stages, I run one quick check on the answer in front of me. It takes under a minute, and it's aimed at one specific trap: an answer that's confident, fluent, and quietly about something other than your problem.
- 1. Name the exact thing. Make it tell you precisely what it's looking at — the specific page, document or item — before it summarises anything. "Which one is this, exactly?"
- 2. Hand me one fact I can check in under a minute. One date, one number, one name I can hold against the real thing myself.
Check One, Bin the Lot. Fail either question and the whole answer goes — not the wrong line, all of it. The good-looking paragraphs aren't a consolation prize. They're the part that nearly fooled you. Passing both questions doesn't make the answer right — it means it's about the right thing, and worth running the four stages on.
Say you've pasted in your phone contract and asked for a plain summary. Before you read a word of it, make the AI tell you what it's looking at: which provider, which plan, the monthly price printed at the top. If it says £29 when the page in front of you says £39, that's someone else's contract being described in a very reassuring voice. Then make it hand you one fact you can check fast. A made-up number gives itself away the second you look. A right answer about the wrong thing never does: every fact correct, neatly laid out, and about a contract you've never seen in your life.
That clears the first trap. The four stages below are for the answer once you trust it's about the right thing — they stop an answer that's technically correct but missing what matters from talking you into a decision.
What are the four stages of the Prompt Stack?
One question gets you one cheerful answer. Four questions, asked in order, each clear the ground for the next. Here's each one with an everyday example — the exact prompts I run on real money are in the proof layer below.
01 — ROLE Set the stance, not a costume.
Left to itself, the AI sells you the sunshine. So I give it the opposite job first: set the stance, not a costume — sceptical, looking for reasons the view is wrong, and licensed to decline rather than guess. A grand job title doesn't make it more accurate — the instruction does. "Act as a top analyst" buys you no extra accuracy; "assume this is a bad idea until the facts say otherwise" changes the whole answer.
Weighing up a holiday deal, I tell it to be the tight-fisted friend who assumes every deal hides a catch and finds that catch before saying a single nice thing. It stops describing the infinity pool and starts hunting the resort fee and the "airport" that's a two-hour coach ride away.
The exact prompt I run on a real position is in the proof layer below ↓
02 — FILTER Split what it knows from what it's filling in.
Most AI answers mix what's on the page with what the model invented on top — same tone, same confidence. The filter step makes it sort the answer into two piles: what the thing in front of you actually says, and what it's guessing. A sharper version is fixed-versus-moving — settled facts that don't change in one pile, anything that moves, or that someone has put a spin on, in the other. Whether a claim belongs in the first pile or the second is the one call the model shouldn't make on your behalf.
On that holiday deal, "breakfast included" goes in the first pile only if the deal page says so. If it's a confident guess, it goes in the second, where you can see it for what it is.
The filter run on a real company, with the invented lines struck out, is in the proof layer below ↓
03 — RISK Name the one thing that would prove it wrong.
"What are the risks?" gets you a checklist, and a checklist isn't the point. The question that matters is the one almost nobody asks: what exact thing would prove this wrong — a tripwire, not a vague worry. A risk you can't see in advance is background noise. A risk with a specific signal is something you can act on.
On the holiday deal: "if the total at checkout comes out over £900, the headline price was a lie." Now you're not weighing a foggy feeling — you've got one number to watch, and the deal either trips it or it doesn't.
The exact prompt I run on a real position is in the proof layer below ↓
04 — VERDICT One answer, with a confidence level.
Nobody standing at the checkout with their card out has ever been helped by a balanced six-paragraph essay on both sides. The last step stops the hedging: one answer, and how sure it is — low, medium or high. A confidence label you can hold it to later is worth more than a summary it can hide behind. The honest upgrade, when it matters, is to ask the same thing again a different way and see if the answer holds — a verdict that survives a re-ask is one you can lean on.
Book it or don't, in one line, with a confidence level. One verdict you can decide on, not a fog you have to referee.
The exact prompt I run on a real position is in the proof layer below ↓
Same four stages, run on real positions — where a confident wrong answer costs money. Below are the exact prompts, a real worked example, and the posts where each one met a live decision.
How do you run all four stages in one prompt?
Copy, paste your position at the bottom, run. Or use the per-stage prompts further down for tighter control between steps.
Running each stage separately lets you review and correct the output between steps. Worth doing when real money is involved. Per-stage prompts below ↓
What does each stage look like as a real prompt?
01 — ROLE Set the stance before the question.
AI defaults to agreeable. Ask "what do you think of this stock?" and you get a balanced-sounding summary that reads, on closer inspection, as mildly positive about almost anything. The first stage hands the AI a stance before it sees your question — so it argues from a position, not toward one.
Sharpen the stance for specific jobs — a "value-oriented short-seller", a "compliance officer reading this filing for the first time", a "risk committee member who has to defend approving this trade." It's the adversarial stance each one carries that does the work, not the title. Most of what the rest of the prompt is for gets done right here.
02 — FILTER Separate facts from filler.
Most AI output on a company looks like analysis and is actually narrative — verifiable facts stitched together with plausible-sounding extrapolations, in the same tone, with the same confidence. The filter step asks for two lists: what's in the source material, and what the model invented on top of it.
After the output: scan List A and move anything that's actually an inference. Models smuggle assumptions into List A constantly — phrases like "the company is well-positioned to…" are not facts. Once the lists are clean, everything that follows is built on something auditable.
03 — RISK Make the downside explicit.
"What are the risks?" gives you a checklist, and a checklist isn't what this stage is for. The third question — what you'd actually see if it was going wrong — is the one that matters. A risk you can't observe in advance is background worry. A risk with a specific signal is something you can act on.
If the AI can't give you a specific observable signal for (c), it doesn't really have a view — it's pattern-matching off the consensus. That's a useful signal in itself.
04 — VERDICT One action. With a confidence level.
By this point there's something worth reading. The verdict step is short on purpose: one practical action, a stated confidence level, and a sentence on why. An AI forced to commit to "low confidence" is more useful than one allowed to hide behind a balanced summary.
The confidence label builds up into a track record over time. Was it right when it sounded sure? What happened when it wasn't? That calibration is most of the value.
What does the FILTER stage actually remove from an AI analysis?
A typical AI response on a company mixes verifiable facts and invented narrative in the same breath — same tone, same confidence. Stage 02 pulls them apart. Here's what it cut from a META Q4 2024 analysis.
- Revenue $48.4bn — up 21% year-on-year
- Operating income $23.4bn — margin 48%
- Family of Apps MAU: 3.35bn
- Reality Labs operating loss: $5.0bn
- Q1 2025 guidance: $39.5–41.5bn
- "continued dominance in digital advertising"
- "exemplary management execution"
- "unassailable position in social media"
- "Reality Labs showing encouraging progress"
- "signals strong management confidence"
Which posts apply the Prompt Stack to a real decision?
Each of these posts is the Prompt Stack applied to a real decision — with the actual model output, the finding, and what I did with it.
- The single prompt change that made AI analysis worth using
Stage 2 — FILTER. One instruction, a real worked example, and the stale figure it caught before a trade was placed.
- 5 questions to ask AI before buying any stock
Stage 4 — VERDICT. Five questions for the hour before you place the order, when the research is done and the decision is about to be made.
- 7 AI prompts for covered calls
Full stack applied to options income. Seven prompts built around real chain data — position fit, IV regime, strike selection, roll decisions.
- 5 AI prompts for earnings call analysis
The stack applied to earnings calls — reading management language, surfacing what wasn't said, and stress-testing the guidance before you act on it.
- Best AI tools for earnings analysis
Which tool does which job on a live results release. The stack is only as useful as the tool running it — this is where each one breaks down.
- AI limitations in options trading
Where the stack doesn't save you. What happened when AI fabricated options chain data — and what it consistently missed on volatility regime.
For the full per-tool failure catalogue — every AI fabrication caught across these tests, with the prompt, the output, and the screenshot — see The Lessons.
The Field Guide expands on this.
A four-stage method, five copy-paste prompts, and a one-page reference card. Tested on a real portfolio. Not a demo.
Get the Field Guide →Free. No email required.
I started testing AI to find out whether its answers could be trusted — on my own investing first, where being wrong costs money. The Prompt Stack is the method that survived. More about me →
// Open methodology · The Prompt Stack is also published as an open repository at github.com/CtrlCursor/prompt-stack under CC-BY 4.0. Share, adapt, credit.