Why do we believe strangers on the internet?

Before checkout pages, this question belonged to philosophy. Now it belongs to anyone with a search tab open. In BrightLocal’s February 2026 survey, an industry survey with an industry survey’s incentives, 97% of consumers read reviews for local businesses, and 41% said they always read them before choosing. Reading reviews is believing strangers.

Philosophy split early on whether that habit can be rational. David Hume argued in 1748 that a report deserves belief only when experience vouches for the reporter: check first, trust after. Thomas Reid answered from the other side in 1764: we arrive built for this, with a principle of veracity that inclines us to speak truth and a principle of credulity that inclines us to believe. Even about liars he was an optimist: where they lie once, they speak truth a hundred times. The Stanford Encyclopedia still files the whole field under those two names, and its warning about the trusting side has a familiar ring: without some check, a person could be justified in believing the say-so of a random blogger on an arbitrary website.

Hume wrote his rule against miracle reports, and it survives contact with a product page untouched. Here is the passage, from Section X of the Enquiry:

We entertain a suspicion concerning any matter of fact, when the witnesses contradict each other; when they are but few, or of a doubtful character; when they have an interest in what they affirm; when they deliver their testimony with hesitation, or on the contrary, with too violent asseverations.

Read it twice: it is a working checklist. Contradiction. Scarcity. Character. Interest. Manner. Every suspicious thing about a review section in 2026 was already on that list.

Hume’s ledger compresses into four questions you can ask of any stranger’s word, on any platform, in about a minute.

Stake. What does the speaker gain if you believe them? Interest is the first thing Hume priced, and the modern web declares it in markup: an affiliate link is a stake you can view with right-click. A stranger with nothing to gain outweighs a polished page with a commission behind it, and a stranger testifying against their own side outweighs everyone.

Independence. Condorcet proved the arithmetic in 1785: majorities of independent judges, each better than a coin flip, approach certainty as the group grows. The theorem’s fine print is the useful part: it assumes that correctness is independent across individuals. Psychology shows what happens when it fails. In Asch’s conformity experiments, with wrong answers spoken aloud first, 75% of participants bent to the visible majority at least once. Fifty voices are only worth fifty when they never heard each other; a hundred reviews echoing one viral thread add almost nothing to the count.

Specificity. A detail that could be wrong is collateral. Terrible support risks nothing; a day-count, an amount, a quoted line from an agent can each be contradicted, which is exactly why they are worth more. A caution, though: specific lies are cheap to write too, so this test never works alone. It buys extra weight only from voices that already cleared the first two.

Date. Testimony ages, because products keep shipping. The recency guide takes this one up in full; the short version is that an undated report leaves you guessing which version of the product it describes.

Run the tests on three real reports about one claim, pulled from a record of 56:

Stake: the first witness is a customer whose refund is on the line, a stake the format itself discloses, and in the replies under that post an Uber Eats driver adds that the pattern embarrasses him, testimony running against the platform that pays him. Independence: two platforms, three years apart, strangers to each other. Specificity: a driver’s confirmation, a quoted support line, a repeated-requests mechanism, each one falsifiable. Dates: 2021, 2023, 2026, the same mechanism holding across five years. Hume never saw a food app. His ledger clears these witnesses anyway.