The only marketing that survives AI slop is the claim someone can go check

AI can fake anything that sounds right. It cannot fake a number you can go check yourself. That gap is the whole thing.

The cost of producing text that sounds right has gone to zero, so sounding right stopped being a signal. What cheap generation still cannot fake is specificity that costs something if it is false: the real number, the launch that did not convert, the figure you would rather round up. Someone making it up does not volunteer those, because they were not produced by doing the thing, they were produced by predicting what doing the thing sounds like.

I build a product where this is the core discipline. When it answers a question about your wallet or a contract, the answer is grounded in live data it can show you, not in confident recall. And when it cannot read something, I would rather it say "I cannot read that right now" than hand back a smooth, empty reassurance. That choice costs something. A vaguer product would be easier to ship and harder to disprove. That is exactly why the honest one is worth more.

The same logic runs through how I talk about it. When I make a claim about what the product does, I want it checkable against the actual code, not just persuasive. It is slower and it is harder and it loses to slop on volume every time. But in a world of infinite cheap plausible text, the one scarce thing is the claim someone can verify.

So the question I keep asking, for both the product and the pitch: can someone go check this themselves? If the answer is no, it is probably slop, even when I wrote it.

For founders here: what is your version of the checkable claim, the thing in your product or your marketing that someone could call you out on if it were false? That is the thing I would lead with.

I wrote the longer version of this here if it is useful: https://dev.to/txdesk/ai-can-fake-anything-that-sounds-right-it-cannot-fake-a-number-you-can-go-check-yourself-52h5