A cookie banner is the first sentence most visitors read on a site, and it is usually the least examined sentence on the page. It ships with the template. It sounds like legal furniture. Nobody writes it on purpose, so nobody checks whether it is true.
Ours was not true. Until yesterday morning, the banner at the bottom of this shop told every visitor that payment processing on the site runs through Stripe. This shop has no Stripe account connected. It cannot take a card payment at all. The free downloads check out at zero dollars, and that is the whole of the commerce here right now. A banner announcing a payment processor we do not use sat under every page, including the privacy page, calm as a street sign pointing at a road that was never built.
The costly part is how it survived. The site had been through design passes, screenshot reviews, and a full read of every page. The banner was in the screenshots. It was small, it was at the bottom, and it looked like the kind of text that is true by definition, the way a copyright line is true. It took a reviewer reading the footer the way a visitor would, with no knowledge of what the site was supposed to say, to ask why a shop with nothing to sell was naming its payment processor.
The fix was already in the building
The store software knows whether a Stripe key is configured. It has to know; checkout depends on it. The banner text simply never asked. The claim was written once, as a constant, by someone picturing the average store, and the average store in that picture had payments set up.
So the repair was not research and it was not copywriting. The generator now checks the one fact it already holds and prints the Stripe sentence only when a key exists. Without one, the banner says what the site actually does with a visitor's browser, which is save the cart and the preferences to local storage, and the privacy page says the plainer thing too: this shop takes no payments online and collects no card details. The same pass caught the checkout page promising confirmation emails on a shop with no email service connected. Those promises are now gated on the email key the same way. When a key gets set up, the sentence comes back on its own, because it will have become true. That part has already happened once: the email service was connected last night, the checkout promise switched on at the next build, and a real free order produced the confirmation email the page had promised, with the download link inside it.
That is the whole principle, and it generalizes past banners: any sentence a program prints about itself should be one the program can verify at the moment it prints it. Prices, currencies, reply times, stock claims, shipping promises. If the software holds the fact, the sentence has no excuse for guessing.
What a banner owes
Accuracy, first, in the boring sense. Name the storage you actually use. Name a processor only if one is actually processing. A visitor who reads the small print and finds it false has learned something about the shop that no product page can unteach.
Brevity that comes from having little to confess. Our banner got shorter when it got honest, because a site that stores a cart locally does not need three clauses. Long banners are often long because they are describing a site the owner wishes they ran, trackers and processors and partners included.
And an owner. Template text with no owner drifts toward the average, and the average is a fiction that fits no particular shop. Somebody has to read the footer once a season the way a stranger would, including the parts that look too dull to be wrong.
The check takes ten minutes on this site now: read the banner, read the privacy page, then look at the settings and count how many of the named services are actually connected. Yesterday the count and the copy disagreed. Today they match, and the match is enforced by the build rather than by anyone's memory, which is the only version of this fix worth trusting next season.