Three ways to describe a supplier; only a dated observation is checkable

ON THIS PAGE

1. Key findings
2. Three forms, one checkable
3. The line drawn in one worked example
4. Why implication gets the same treatment as a false statement
5. A parallel rule for a different kind of claim
6. Why an entry ages rather than settles
7. The test a reader can run on any directory
8. Limitations and open questions
9. Sources

A research directory and a rumor site can describe the same vendor in the same week and read completely differently, and the difference usually comes down to one sentence shape rather than to how much either site actually knows. What was observed, on what date, and nothing claimed past that, is the shape that holds up. Nearly everything else does not.

Three forms, one checkable

This publication's own review rubric, the document a fresh adversarial reviewer reads before checking any draft, names the distinction directly. A sentence about a named supplier can take one of three forms: a dated observation, an inference drawn from that observation, and a claim about intent. Read directly from the source on 2026-09-19, the rubric states that only a verifiable statement of fact, with its support in the piece or in this publication's own ledger, is acceptable outright. An inference or an opinion is acceptable only when a reader cannot mistake it for a statement of fact. A claim about intent is treated as its own separate and much harder category, because intent is a fact about someone's state of mind, and a reader checking a dated observation against a source can verify it. A reader checking a claim about intent has nothing to check it against.

The line drawn in one worked example

The rubric does not leave this distinction abstract. Its own worked example reads: "No certificate could be found on the vendor's site on the date fetched" is a fact. "The vendor hides its certificates" is a claim about intent, and intent is almost never checkable. The two sentences can describe the identical underlying observation, the same site, the same date, the same absence of a document. One states what was actually looked for and not found. The other adds a motive nobody checking a website can establish, and the rubric's own text is explicit that absence of evidence is not evidence of wrongdoing: a missing certificate is a fact, and a reason for its absence is not something a fetched page can supply.

Why implication gets the same treatment as a false statement

The rubric extends past sentences that are literally false. It states that a sentence which states nothing false but leaves a reader with a false impression of a named business is treated the same as one that is false outright, and that a true fact placed so that it implies an untrue one fails the same way. Two true, separately checkable facts read together can add up to a claim neither one makes on its own, and the rubric's standard is the impression a reasonable reader takes away, not whether each individual clause survives a literal fact-check. A directory that wanted to imply wrongdoing without ever writing a false sentence could do so simply by choosing which true facts to place next to each other. The rule against implication is what closes that route.

A parallel rule for a different kind of claim

FTC's own advertising guidance, read 2026-09-19, sets a related standard for a different situation, a company making a claim about its own product rather than a publication describing someone else's: "Before a company runs an ad, it has to have a 'reasonable basis' for the claims," and a reasonable basis "means objective evidence that supports the claim." The same guidance states that statements from satisfied customers usually are not sufficient to support a claim that requires objective evaluation. The parallel is not exact, since one rule governs an advertiser's claims about itself and the other governs a publication's claims about someone else, but both rest on the same underlying idea: a claim earns the right to be stated as fact by carrying evidence a reader could actually check, and a claim that cannot be checked stays labeled as what it is, an inference, an impression, or an unverified claim, rather than being written as settled fact.

Why an entry ages rather than settles

A dated observation is honest about its own limits in a way an undated claim is not. "No certificate could be found on the vendor's site on the date fetched" says nothing about the vendor's site the following week, and it is not meant to. What stays unknown about any dated observation is exactly what happened after it was made, which is why an entry in a record built this way ages rather than settles: it does not become wrong when the underlying fact changes, because it never claimed to describe anything past the date it names. A claim about intent has no comparable expiration, because it was never tied to a checkable fact in the first place, and that is a second, quieter reason the distinction matters: one kind of sentence can be superseded by a later date, and the other kind never actually was pinned to one.

The test a reader can run on any directory

The test this rubric applies to a draft before publication is one any reader can run on any site making claims about named businesses. Does the sentence carry a date. Does it state what was actually observed rather than why it happened. And when two true facts sit next to each other, do they add up to a claim neither one states on its own. A site that fails this test consistently is writing impressions and calling them facts. A site that passes it consistently is doing the slower, less satisfying work of describing only what it actually checked.

Limitations and open questions

This piece describes this publication's own review rubric as read directly from its source on 2026-09-19; the rubric itself can be revised after that date, and a reader checking this description later should verify it against the current document rather than this piece. This piece takes no position on how consistently this publication's own published articles satisfy the standard described here, which is a separate question this piece does not audit.

A dated observation says what was found and when. An accusation says why, and why is usually the one thing nobody checking a website can actually know.

Sources

This publication's own review rubric, REVIEW-PROTOCOL.md, read directly on 2026-09-19, for the three-form distinction and the worked example quoted above. FTC, advertising guidance materials published at ftc.gov, read 2026-09-19, for the reasonable-basis standard quoted above.