Practical guide

AI positioning brief verification checklist

A positioning brief is ready for use only when a reviewer can reconstruct its audience, evidence, alternatives, claim boundaries, contradictions, and approval. Fluent wording is not the acceptance test. This checklist turns the brief into a controlled artifact that can be passed, repaired, narrowed, or held before its messages reach a campaign, product page, sales deck, or launch plan. Use the.

By Two Prune

A positioning brief is ready for use only when a reviewer can reconstruct its audience, evidence, alternatives, claim boundaries, contradictions, and approval. Fluent wording is not the acceptance test. This checklist turns the brief into a controlled artifact that can be passed, repaired, narrowed, or held before its messages reach a campaign, product page, sales deck, or launch plan. Use the CHECK review: Confirm scope and version, Hear the likely interpretation, Examine evidence and calculations, Challenge alternatives and contradictions, and Keep approvals plus residual uncertainty.

Before review: assemble the verification packet

Freeze the brief version and collect the evidence it relies on. The packet should contain the decision the brief supports, product or offer version, intended audience and situation, research table, interview or observation references, alternative analysis, positioning statement, message-proof map, disqualifiers, source dates, rejected variants, and named owners.

Do not review a pasted statement in isolation. A sentence can look plausible while pointing to an expired product configuration or a different audience. Record the reviewer, date, channel, geography, and planned release window. If the packet changes after review, create a new version rather than silently replacing the reviewed evidence.

C — Confirm scope, identity, and completeness

Check that the brief answers one bounded positioning decision. Verify the offer name and version, audience, buying situation, geography, time horizon, real alternatives, intended action, owner, and exclusions. Confirm that every required field is present and that related artifacts use the same definitions.

Mark hold when the brief blends materially different products or audiences, has no observable decision, or names a category that the current offer does not fit. Mark repair when the decision is sound but a field, source, qualifier, or owner is missing.

The UK Government Communication Service's OASIS guide links objectives, audience insight, strategy, implementation, and evaluation. A positioning brief is not a government campaign plan, but that sequence provides a useful completeness test: the proposition should connect to an objective and audience evidence, and the later evaluation should match the intended action. GCS OASIS guide

Sources: [1]

H — Hear what a reasonable reader may understand

Write the explicit and implied meaning of every material statement. Ask representative readers to paraphrase the position without coaching. Record what they believe the product does, for whom, compared with what, under which conditions, and with what expected result. Test titles, diagrams, callouts, footnotes, and omitted information as well as the central sentence.

A qualification fails if readers are unlikely to notice it or if a shorter channel variant removes it. An internal phrase such as “evidence-led” can imply a demonstrated outcome even when the team meant only that source links are available. Review the likely interpretation, not the author's private intention.

The US Federal Trade Commission's advertising-substantiation policy says objective express and implied claims should have a reasonable basis before dissemination and that the amount and type of support should match what the claim communicates. Use that source to review US-facing objective claims; it does not replace jurisdiction-specific advice. FTC Policy Statement Regarding Advertising Substantiation

Sources: [2]

E — Examine evidence, numbers, and transformations

Trace each objective assertion to a dated source. Confirm that the source covers the same product version, audience, comparison, outcome, and conditions. Separate direct observation, customer statement, product documentation, calculation, estimate, forecast, and team hypothesis. A source may be genuine but still fail to support the wording.

Recalculate every number from the underlying record. Check units, denominators, time windows, exclusions, baselines, rounding, and whether a relative change has been presented as an absolute one. For a comparison, verify that alternatives use comparable definitions and dates.

When AI summarized or clustered research, sample the original records. Verify that minority observations and contradictions were not lost, quoted language remains accurate, and sensitive information followed approved handling. Retain the prompt or transformation note only when it is proportionate and useful for reconstruction; the source records remain authoritative.

C — Challenge alternatives, contradictions, and disqualifiers

Confirm that the named alternative is what the audience actually uses, including doing nothing, adding a process control, or using a broader platform. Look for evidence that supports another frame. Record it rather than forcing every observation into the preferred story.

Run four adversarial tests:

GovS 011 calls for objectives, audience insight, propositions and messages, channels, measurable outcomes, and accountable governance. It supports checking that a proposition and its evidence remain connected across delivery; it does not establish that a particular commercial position is correct. Government Functional Standard GovS 011

  • substitution: would the sentence still appear true if a competitor name replaced the offer?
  • boundary: which audience, workflow, configuration, or jurisdiction makes the claim wrong?
  • contradiction: what credible observation points in the opposite direction?
  • freshness: what changed product fact, market condition, or source expiry would invalidate the brief?

Sources: [3]

K — Keep approvals, dispositions, and uncertainty

Give every checklist row one disposition: pass, repair, narrow, hold, or not applicable with reason. Name the evidence owner, product owner, communication owner, and any qualified reviewer needed for the channel. Record disagreements instead of averaging them into approval.

The final record should show the approved wording, rejected stronger wording, supporting sources, qualifications, unresolved uncertainty, expiry or review trigger, downstream artifacts allowed to reuse it, and the person who can withdraw approval. Approval of one brief version is not approval of future variants.

Copyable checklist

| Test | Evidence to inspect | Pass condition | Failure disposition | | --- | --- | --- | --- | | Decision and version | brief header, product record, release plan | one bounded decision and current offer version | repair or hold | | Audience and situation | research table, sampling note | audience and buying situation are evidenced and specific | narrow or gather evidence | | Real alternatives | interviews, workflow observation, competitor or process records | alternatives reflect current behaviour, including no change | repair | | Explicit and implied meaning | reader paraphrases, channel mockups | reasonable interpretations stay within the evidence | rewrite or hold | | Claim support | source ledger and dates | every objective assertion has suitable prior support | remove, narrow, or obtain evidence | | Calculations | underlying data and formula | units, denominator, window, exclusions, and result reconcile | repair and rerun | | AI transformation | source sample and transformation note | summary preserves material meaning and contradictions | repair or discard summary | | Product truth | current documentation and configuration | described mechanism exists under stated conditions | narrow or hold | | Contradictions | negative cases and dissenting research | contradictory evidence is resolved or visible | qualify, test, or hold | | Disqualifiers | exclusions and channel rules | conditions that make the position wrong are explicit | narrow | | Approval and expiry | owner record and review date | required owners approve the exact version | hold | | Residual uncertainty | open-question log | uncertainty and next evidence gate are visible | publish with limit or hold |

Worked verification: facilities energy analytics

This synthetic brief positions an analytics tool for facilities managers who compare energy use across multi-site estates. The draft statement says: “The only AI platform that cuts building energy costs by 20% without changing your controls.”

CHECK finds six material failures. “Only” relies on an eight-month-old comparison that excluded two current alternatives. The 20% figure comes from a modelled scenario, not observed customer outcomes. “Cuts costs” converts an estimate into an achieved result. “Without changing controls” conflicts with product documentation requiring a new data-approval step. Interview notes include two facilities teams whose main problem is missing meter data rather than analysis. The brief also omits that the current connector supports only one building-management-system version.

The reviewer does not polish the sentence. The disposition is hold. The team narrows the audience to multi-site facilities teams with compatible interval-meter data, replaces the outcome promise with a description of the comparison workflow, adds manual review and connector conditions, and lists spreadsheet analysis plus existing building-management tools as alternatives.

The revised internal position reads: “For multi-site facilities teams with compatible interval-meter data, the tool assembles source-linked comparisons for human review across sites and periods.” Evidence owners verify the product mechanism and connector boundary. Representative readers no longer infer autonomous control or a measured cost reduction. The brief passes for a limited sales-enablement test, with the modelled savings scenario retained only as an explicitly labelled planning example.

The retained artifact includes both versions, all checklist rows, source links, reader interpretations, calculation notes, the contradictory interviews, product approval, channel limitation, and a 60-day review trigger. It does not claim that the position will improve demand or that passing the checklist establishes legal compliance.

Two Prune boundary

Two Prune can assess whether a participant verifies a positioning brief, traces evidence, challenges AI-generated claims, handles contradictions, and communicates a bounded disposition in a realistic simulation. It does not approve marketing claims, determine legal compliance, establish product-market fit, or make employment decisions. Customers remain responsible for market research, product facts, qualified review, release ownership, and real-world commercial choices.

Sources

  1. 1.GCS OASIS guide · gcs.civilservice.gov.uk
  2. 2.FTC Policy Statement Regarding Advertising Substantiation · Federal Trade Commission
  3. 3.Government Functional Standard GovS 011 · UK Government