How we score

A Grill Score is an editorial assessment of one speaker in one exchange, scored against the criteria on this page. It is our judgment, supported by the record we show you. It is not a measurement, and the decimal in an average is arithmetic, not precision.

We score statements and responses in this exchange. We do not rate anyone's character.

Three categories, scored separately

Each category is a whole number from 0 to 10. The three are shown side by side and matter more than the average: a direct answer does not cancel out an inaccurate one.

Answered (for the person responding)

Did the response address the question that was asked?

• 10Answers the question asked, directly and completely.
• 8Answers it directly, with minor hedging or added context.
• 5Answers part of it, or answers a different but related question.
• 3Mostly pivots; the question is acknowledged but not answered.
• 0Refuses, or ignores the question entirely.

Clear question (for the person asking)

Used instead of Answered when we score the questioner. Was the question clear and answerable?

• 10One clear question, answerable from what the speaker knows.
• 8Clear, with a minor ambiguity.
• 5Compound or partly unclear; a fair answer would have to choose which part to answer.
• 3Built so that any short answer misleads.
• 0Not a question in practice; a statement or attack phrased as one.

Accurate

Do the factual claims match the available evidence? Scored only on claims we checked against a source we show you.

• 10Every checked claim holds up.
• 8Holds up, with minor errors that do not change the point.
• 5Mixed: some checked claims hold up, others do not.
• 3Mostly not supported by the record.
• 0The central claim is contradicted by the record.

Fair

Did the exchange use accurate premises and represent the other side faithfully? Judged on what is observable: false premises, distorted or truncated quotations, misrepresented positions, or a fair restatement. We do not guess at hidden motives.

• 10Accurate premises; the other side's position stated fairly.
• 8Mostly fair, one minor slant.
• 5A premise or framing that the record does not support, alongside fair points.
• 3Repeated false premises or a clearly misrepresented position.
• 0Built on a distorted quotation or a fabricated premise.

In-between scores

A score between two anchors (for example 6 or 7) means the response sits between those descriptions. We write one line of reasoning for every score, so you can see why it landed where it did.

A worked example, scored with rubric v1

One real episode, as published. Rubric v1 used one shared ladder for all three categories; the v2 anchors above apply from the first episode published after Oct. 17, 2026.

When a category is not scored

• Not applicable (N/A): the category does not fit, for example Answered for someone who was never asked a question.
• Insufficient evidence: we could not check the claims against a source we can show you. Missing evidence is never scored as 0.

Unscored categories are shown as N/A or "Not enough evidence", never as a number.

The average

• The average is shown only when at least two categories are scored. With fewer, we show the component score alone and no average.
• It is the plain average of the scored categories, rounded to one decimal place (5.25 becomes 5.3).
• Every score names the speaker and the exchange it refers to.

Findings on a receipt

Each receipt shows the claim, what the record says, the source, and our finding, in the words used in the episode. Findings fall into these groups:

• Supported: the record backs the claim (for example TRUE, HOLDS UP, MOSTLY TRUE).
• Mixed: part holds up and part does not, or it is accurate but leaves out something that changes the picture (PARTLY TRUE, MIXED, OVERSTATED, STRETCHED, LEFT OUT, MISLEADING).
• Not supported: the record contradicts it (FALSE).
• Not checked / not enough evidence: we could not find a source we can show you, so we do not rule on it.

An unanswered question is not the same as a false claim: "Not answered" is about the response, a finding is about a factual claim. Where a published finding does not fit one of these groups on its own, the receipt says "See the record" and the record rows carry the result.

PAID and DUE

On screen, a receipt is stamped PAID when the claim checks out against the record and DUE when it does not or when the question is still unanswered. The written finding on the receipt is the precise result; the stamp is shorthand.

The Dodge Clock

The clock shows how long a yes-or-no question goes without a yes or a no.

• Starts when the question is finished.
• Stops at the first clear yes or no, or a clear "I don't know" or "I won't answer".
• Pauses while the questioner interrupts, adds to the question, or someone asks for clarification.
• A compound question gets one clock per part.

Elapsed time alone does not prove evasion; it is shown next to the tape so you can judge.

Rubric history

• v1 (published episodes Oct. 13 to Oct. 17, 2026): three questions, Answered (or Clear question), Accurate, Fair, 0 to 10, with one shared ladder for all three. The Bannon episode (Oct. 17) showed counts (questions answered, claims checked, claims that held up) instead of 0 to 10. Those episodes keep their published scores.
• v2 (this page, adopted Oct. 17, 2026): separate anchors per category, N/A and insufficient-evidence rules, average only with two or more scored categories. Applies to episodes published after Oct. 17, 2026.

Every published episode is labelled with the rubric it was scored under. Browse the episodes or read the Corrections.