4 Tests That Your Work Should Pass
How to check your reasoning before someone else does
Most of us are reasonably good at spotting problems in other people’s work. A missing source, an argument that leans too hard on a single example, a conclusion that was clearly chosen before the analysis began. We notice these things quite readily when we’re reading someone else’s output.
But we are much worse at applying the same scrutiny to our own work.
This has always mattered, but it matters more now that AI can produce polished, confident-sounding material in seconds. The hard part today is knowing whether the thinking behind any text you read is any good. That requires a set of standards.
So, what follows is a set of four tests. They work for pretty much anything you can read, from articles, to consulting deliverables, to training materials. The tests come from a framework developed in the 1940s for a completely different context, and I’ll come back to that later.
For now, here are the tests, followed by a custom diagnostic prompt you can use with any generative AI tool.
1. Can others see how you got there?
This is the most basic test, and the one that gets skipped most often. If someone reads your work, can they trace how you arrived at your conclusions? Can they see what evidence you drew on, what assumptions you made, what you chose to leave out?
A lot of professional work relies on implicit authority. The reader is expected to trust the conclusion because of who produced it, what firm they work for or how confident the language sounds. That is fragile, however, and it doesn’t survive contact with a sceptical client or anyone who asks “how do you know that?”
To answer this question satisfactorily, you need to make your sources visible and share your method, even briefly. You also have to be honest about the limits of your evidence, including where the data was thin, where you made a judgement call and where the sample was small or unrepresentative. None of this undermines your credibility - if anything, it strengthens it, because it shows you’ve actually thought about the foundations of your own argument rather than glossing over them.
The AI angle here is important. When you use AI to help produce something, the chain of reasoning can become genuinely hard to trace. The tool generates plausible-sounding claims, and unless you check each one, you may find yourself presenting material whose origins or logic you can’t account for. If a client asks “where did this figure come from?” and your honest answer is “I’m not sure, the AI produced it,” you have a problem.
I have spoken about this in some detail in a recent live with Exploring ChatGPT - here’s a link if you’d like to go deeper into this discussion.
2. Would you accept this from someone you disagree with?
We are not good at separating an argument from the person making it. When a claim comes from someone senior, well-credentialed or institutionally backed, we tend to engage with it more generously. When the same claim comes from someone junior or unfamiliar, we might scrutinise it harder or dismiss it altogether. This works in the other direction too: if someone we find generally unreliable makes a sound point, we’re likely to underweight it.
The test here is straightforward: take your argument and swap the speaker. If the person you find least credible in your field presented the same reasoning you are using in your own writing, with the same evidence, to reach the same conclusion, would you find it rigorous? Would you accept the evidence as sufficient? Would you consider the method sound?
If you wouldn’t, that tells you something about how much work the evidence is actually doing.
This plays out in professional settings all the time. A consultant might scrutinise a competitor’s methodology line by line, then accept something equally flawed from a partner at a more prestigious firm. A policy team might wave through a weak evidence base if the research comes from an institution they already trust. A senior leader’s strategic recommendation might get less pushback than an identical suggestion from someone two levels down.
The standard should be the same regardless of source. That sounds obvious when stated plainly, but in practice it requires active effort. It means being willing to take a nobody’s good argument seriously, and equally willing to challenge a respected authority’s weak one. Most professionals would say they already do this, but fewer actually do.
The Art of Asking Questions is a reader-supported publication. To support my work, please consider becoming a paid subscriber.
