Standard-form disclosures, the terms and conditions on a signup screen, the conflict-of-interest line on a financial recommendation, the EULA nobody scrolls to the end of, are built on one assumption: that a person who is warned will read the warning, and adjust. Real usage data says almost nobody opens them at all.
The deeper problem sits underneath that one. Even when a disclosure is read, disclosing a conflict of interest can free the person giving advice to lean into their bias further, while the person receiving it still doesn't discount enough to catch the difference. The safeguard can make the thing it was meant to guard against worse, not better.
This report covers both halves: the evidence that standard disclosure goes largely unread, the study showing why disclosure can backfire even when it is read, and the real research on what actually changes borrowing and spending behaviour instead.
The story in four parts
A large-scale study of real browsing logs found the standard-form contract gets clicked open by almost nobody before they buy.
A controlled lab study found that disclosing a real financial conflict of interest didn't rein in biased advice. It made the advice more biased, and listeners still didn't discount enough to make up for it.
A real field experiment on payday loans found one specific way of showing borrowing costs cut future borrowing. Two other formats didn't.
What the evidence points to instead: removing the incentive a disclosure is trying to flag, and showing costs the way that specific field test showed they actually change behaviour.
Every standard-form contract, the terms of service on a new account, the licence agreement before a piece of software installs, works from the same legal fiction: that clicking “I agree” means the document was read and understood. A large empirical study of real online shoppers put a number on how often that fiction holds.
Out of a thousand people who visit a page with a EULA on it, how many actually open it?
Bakos, Marotta-Wurgler, and Trossen tracked real browsing behaviour across 90 online software companies and roughly 48,154 monthly shoppers, using clickstream data to see how many visitors ever opened the End User License Agreement (EULA) linked from the checkout page, not whether they said they would read it, whether they actually clicked.
What it teaches: only about one or two shoppers in every thousand ever opened the licence agreement before buying. The rest were, functionally, agreeing to whatever the document said without ever seeing it. A disclosure nobody opens can't change anyone's decision, however carefully it's worded.
That finding is about attention, not honesty: it says nothing about what happens on the rare occasions a disclosure does get read. That's the sharper, more counterintuitive half of the problem, and it's covered on this site's own Disclosure Backfire principle in full: a real lab study found that when a financial conflict of interest was disclosed, the advice built on that conflict got worse, not better, and the people receiving it still didn't discount enough to catch the difference.
If disclosure alone, read or unread, isn't reliably fixing the decision it's meant to inform, the useful question becomes narrower: is there a specific way of disclosing a cost that does change what people do next? A real field experiment on payday lending tested exactly that, on real borrowers, with real money at stake.
The format that named a real, dollar cost over time was the one that worked. An APR figure and a comparison to credit-card rates, both accurate, both a form of disclosure, moved borrowing behaviour only weakly. Showing the actual dollars a borrower would pay in fees if they kept renewing the loan cut subsequent borrowing by about 11% over the next four months.
The dollar-fees-over-time format worked because it named a concrete, personally relevant number at the exact moment a decision was being made, not because disclosure in general works once you find the right words. Two limits are worth stating plainly.
First, the Bertrand and Morse result was delivered by a person, at a counter, at the moment of the transaction. A page-long disclosure buried in onboarding carries none of that timing or personal delivery, and the Bakos et al. finding above suggests most people never reach it regardless of how it's worded.
Second, reformatting a disclosure does nothing about the underlying incentive a conflict-of-interest disclosure is meant to flag. As the Disclosure Backfire principle on this site lays out, a fee-only advisor can disclose their compensation without triggering a backfire, because there's no percentage-based upside left to lean into. A better-worded disclosure of the same conflict doesn't remove that upside; it just states it more clearly.