What triff's markers mean

/ Verified

The primary source confirms the claim.

triff located a primary source — an official government database, regulatory filing, or authoritative record — and the figure or fact in the draft matches what that source records.

This is not a declaration that the claim is “correct.” It means the claim is consistent with what a named primary source says as of the time triff ran.

\ Delta

The claim and the primary source record different figures.

triff found a primary source for this claim, and the figure it records differs from what the draft states.

This is not a correction, and it doesn't mean the draft is wrong. Journalists often have access to more current data, unpublished figures, or sources that postdate the last official release. The marker flags the difference — what the primary source records vs. what the draft says — and leaves the judgment to the author.

Example: A draft states unemployment at 4.1%. BLS records 4.2%. The author may have a pre-release figure, a different methodology, or a more current source. triff notes the difference; the author decides what to do with it.

~ Conflicting

Primary sources disagree with each other on the same fact.

triff found more than one authoritative primary source for this claim, and they record different figures. This is not a reflection on the draft — it's a signal that the underlying data landscape is genuinely contested or ambiguous.

Example: Two official government surveys track the same labor market indicator and report different monthly changes. A draft that cites one is not wrong — but a reader relying on the other would reach a different conclusion. triff surfaces the tension.

? Under-specified

Primary sources are sufficient and clear, but the claim is too loose to render a single comparison.

triff retrieved evidence and can read the primary sources, but the draft claim does not pin down enough detail — metric, period, entity, or comparison baseline — for one fair comparison. Different reasonable readings of the same sentence would yield different markers.

This is a claim-quality signal, not a verification failure. The responsibility sits with the author: tighten the claim so a reader (and triff) know exactly what is being asserted.

Example: “Wages have been falling recently” without a defined window or wage measure. One-month and twelve-month official series can point in opposite directions — triff marks the claim “?” until the draft specifies what “recently” and which series mean.

- Unverified

The evidence isn't conclusive.

There are many reasons this can happen:

• The relevant primary source exists but could not be retrieved (access restrictions, rendering issues, or availability at the time of the run)
• No established primary source is known for this type of claim
• The claim is forward-looking, analytical, or otherwise not directly comparable to recorded data
• The data series exists but does not cover the relevant time period, geography, or sector
• Other reasons specific to the claim

When possible, triff notes what a primary source comparison would require and points toward where that source might be found — so the author can follow up directly if they choose.

For routine or supporting claims, an unverified marker may simply mean the data isn't easily accessible. For important or central claims in a draft, it's worth treating as a flag — if a key fact can't be compared to a primary source, that's worth knowing before publication.

A note on judgment

triff does not declare claims right or wrong. It surfaces what primary sources record and where they diverge from the draft or from each other. The author always has context triff doesn't — a source interview, a pre-release figure, domain expertise, or a deliberate editorial choice. Every marker is an invitation to compare, not a verdict.

The goal is to give authors and editors the primary source picture as clearly as possible, so they can make informed decisions about their own work.

Our work is shown, not assumed

For every marker, triff provides the primary source it consulted, the specific data it retrieved, and how it compared that data to the claim in your draft. You don't have to take our word for it — the references are there so you can check them directly.

That said, triff is built on AI, and AI makes mistakes. A source may have been misread, a figure misattributed, or a comparison drawn from the wrong data series. We encourage you to review the provided references before acting on any finding. If something looks off, it may be.

We're in beta

triff runs on a scalable cloud platform, but we're early. You may encounter slower runs, occasional errors, or findings that don't look right. We're actively improving the system and take every piece of feedback seriously.

If you see something unexpected — a finding that seems wrong, a source that doesn't match, or a run that didn't complete — please let us know. That feedback is genuinely important to us and directly shapes how the product develops.

Send feedback →

Reading a triff string

When triff summarizes a run, it displays a compact string of markers in a fixed order: verified, then delta, then conflicting, then under-specified, then unverified.

/ \ ~ ? -

Each character represents one claim. The string always reads left to right in that sequence, regardless of where the claims appear in the document.

Example: ////\\~~??- means the run found 4 verified claims, 2 deltas, 2 conflicting, 2 under-specified, and 1 unverified.

The shape of the string tells you something at a glance: a run full of / characters looks different from one weighted toward \ or -. That at-a-glance picture is intentional.