The word tells you almost nothing now
Search any marketplace for flashcards, planners, or dev templates and you will see two words fighting for your trust: handmade and AI-assisted. Neither is a quality signal anymore. Plenty of handmade things are lazy. Plenty of AI-assisted things are excellent. And a growing pile of both is slop: generated in bulk, never reviewed, priced like someone spent an afternoon on it.
I make AI-assisted products and I will defend that honestly. What I will not defend is the pretence that the label is the point. So here are three questions that cut through it, whoever made the thing and however they made it.
Who made the judgement calls?
Every product is a stack of decisions. What goes on the cover. Which twenty words make the cut and which forty get dropped. Whether the blue is warm or cold. A tool can generate options all day, but a human has to choose, and choosing well is the whole job.
The tell is specificity. A careful maker can tell you why the deck has forty cards and not fifty, why the guide skips a chapter everyone else includes. Slop cannot answer that, because nobody decided. It was set to a number and shipped. If you can find the maker and the answer is a shrug, you have your answer too.
Can you point to the effort?
Effort leaves marks. A colour pass that fixes the off-brand cyan a generator slipped in. A worked example instead of a definition. Consistent typography across forty cards instead of the drift you get when nobody checked. You do not need to know how it was made to see whether someone went back over it.
Preview images and sample pages are where this shows. If a shop lets you see real interior pages, that is confidence. If every listing hides behind one glossy thumbnail and a wall of keywords, that is usually because page two would not survive the look. Ask for the sample. The answer is information either way.
Is there one real person who answers?
The cheapest thing to automate is the apology. When a file is broken or a card has a typo, slop shops route you into a support maze or simply go quiet. A maker who stands behind the work answers in the first person, because the person who chose the subject and pressed ship is the same person reading the inbox.
This is the question that ages best. Tooling changes every year. Whether a human is accountable for the result does not. If you can email and get a straight reply from someone who clearly made the thing, most of the risk is gone, whatever the render pipeline looked like.
Judgement is the part that does not scale
NeuraGrowth products are AI-assisted and we say so on every listing. The rendering and drafting use modern tooling. The subject choice, the culling, the final ship-or-reject call, and the reply when something is wrong are mine, and those are exactly the three things the questions above test for.
So do not buy on the label. Buy on the answers. Run the three questions on us too, and if any of them comes back weak, that is a fair reason to walk. The whole point of running this in the open is that you get to check.
Originally published at neuragrowth.co. I run a one-person digital-products studio and write up what breaks in production.
If you write CLAUDE.md files, I keep a set of working templates here: neuragrowth.co/free/claude-md-templates.
