What you will achieve
You will test whether a practical voice brief produces recognisable, useful writing across two different tasks. You will finish with reviewed passages, specific editorial feedback and a revised brief that another writer can use. Product facts remain fixed while you assess expression.
This is an editorial calibration exercise. Use the brand-guidance and writing controls currently available in your AdCyte workspace. If approved-example or managed-brand controls are unavailable, keep the examples and review record in the accompanying brief and apply them through the supported writing or manual editing workflow.
Before you begin
- Prepare reviewed video evidence and current approved facts for the selected product or campaign.
- Choose a short factual explanation and a more interpretive campaign passage as the two tasks.
- Gather a small set of approved writing examples and a few rejected passages with reasons.
- Identify the editor who can make the final voice decision.
- Check the current generation and rewrite permissions and charges before using paid actions.
For the initial setup, use Set up brand guidelines and a brief. This guide tests that brief rather than repeating the setup process.
1. Choose two tasks with different editorial demands
The first task should test clarity and restraint. For example, explain how the storage box in a reviewed demonstration folds flat. The second should test perspective and personality, such as a short account of the campaign's idea using supplied campaign context.
Specify the reader, destination and purpose for each. Keep the scope modest enough that an editor can compare the work carefully. You are assessing whether the voice travels across tasks, not asking a single sample to represent every future format.
Record the facts that must remain unchanged. If the campaign passage needs interpretation, identify it as such and keep it separate from product claims.
Check: the tasks differ in expression while sharing a clear, controlled factual basis where appropriate.
2. Turn the voice into observable rules
Write a few rules an editor can inspect. For example: lead with the useful answer; explain necessary technical terms; use connected paragraphs; allow occasional dry wit; avoid unsupported superlatives and stock marketing phrases.
Separate mandatory requirements from preferences. A required product name or forbidden claim is mandatory. A preference for a quieter closing sentence is an editorial choice that can be discussed. This distinction prevents every disagreement becoming an argument about whether the brand has been violated.
Include limits. If humour is allowed, say where it should not distract from an instruction or qualification. If brevity matters, specify that necessary conditions must remain. A useful voice brief gives the writer room to think without leaving the important boundaries vague.
Illustrative editorial brief
Two tasks, one factual boundary

Style references guide expression. They do not supply new product facts.
3. Annotate the examples
For each approved example, identify the feature to reuse. “The opening answers the question directly” is more useful than “we like this”. For each rejected example, explain the problem and the desired alternative.
Label examples as style references. Check them for old offers, outdated product names or claims that should not enter the new draft. Keep the current approved facts in a separate section of the brief.
An example can demonstrate rhythm without becoming a sentence template. Ask for the underlying editorial choice rather than a near-copy with the product name changed. The goal is a recognisable voice, not a family of paragraphs wearing different badges.
4. Prepare the first pair of drafts
Use the reviewed evidence and applicable brand guidance in the supported AdCyte writing workflow. Supply the task-specific purpose and destination where the workflow allows. If a control does not exist, retain that requirement in the editorial handoff rather than inventing a step.
Save or record the resulting drafts before making substantial changes. Keep enough context to identify which brief and source material informed them. Do not publish these test outputs merely because they have been generated.
If a task lacks sufficient evidence, narrow it or obtain the missing approved information. A voice test should not proceed by allowing the writer to invent product detail so the paragraph has something to say.
Check: both drafts can be traced to the intended task and source context.
Constructed before-and-after example
Make the revision explainable

These passages were written for this illustration, not generated by a tested product workflow.
5. Review voice and facts in separate passes
First check factual accuracy: product identity, demonstrated action, current approved statements and qualifications. Resolve those issues before praising the writing for its personality. An entertaining error remains an error with better timing.
Then review the voice against the agreed rules. Does the opening do its job? Is the vocabulary appropriate? Do paragraphs develop a thought? Does the humour help? Does the factual answer remain clear while the campaign passage allows more interpretation?
Use simple judgements such as meets, partly meets or needs revision, with a reason. A numerical score can be useful internally, but it should not replace the explanation or imply a scientifically validated measure of brand character.
6. Make targeted changes and record why
Edit the passages that need attention using the controls available in the workspace. If targeted AI editing is unavailable, make the change manually through the supported workflow. Preserve the source facts and keep a copy or revision reference for the before-and-after comparison.
Write one short reason for each substantive change. “Removed a generic opening so the answer leads” can inform future briefs. “Corrected the product name against the current record” belongs in the factual review. Keep those categories separate.
Refine the voice rule only when the evidence supports a general change. One awkward sentence may need a local edit rather than a new prohibition applied to every future article. Overcorrection can leave a brand guide sounding like a list of incidents nobody is allowed to discuss.
7. Check the revised pair and update the brief
Review both tasks again after the edits. Confirm that the new rule helps across the pair and has not damaged clarity or introduced another problem. If you regenerate, record the revised brief and inspect the new result rather than assuming the instruction worked.
Retain the strongest short passages as reviewed examples if the responsible editor agrees. Note why they were accepted and the formats they represent. Keep obsolete examples out of the active brief without erasing useful historical context from your records.
Do not assume these edits automatically train the product or change other brands' writing. Use only the feedback mechanisms available and approved in your workspace. The explicit brief remains the practical handover.
Worked review record
These entries illustrate editorial decisions rather than product-generated results.
| Review item | Factual explanation | Campaign passage | Brief adjustment |
|---|---|---|---|
| Opening | Answers the folding question directly | Starts with generic promotion | Lead with the specific idea or reader benefit supported by the brief |
| Detail | Keeps the correct product identity | Adds an unsupported claim about durability | Product claims must come from the approved fact record |
| Rhythm | Clear connected sentences | Several fragments interrupt the explanation | Use short sentences selectively within developed paragraphs |
| Humour | Unnecessary beside the instruction | One restrained observation fits | Keep humour occasional and away from material qualifications |
| Final decision | Accept after factual check | Revise opening and unsupported claim | Retain the accepted passage as a scoped example |
Troubleshooting
| Problem | Action |
|---|---|
| Both drafts sound generic | Replace broad adjectives with examples and concrete editorial reasons |
| The voice becomes exaggerated | Narrow the instruction and remove repetitive flourishes |
| An example imports an old claim | Separate style references from current factual authority |
| Reviewers disagree | Identify the rule or reader need behind the disagreement and name the final editor |
| A new feature is unavailable | Use supported writing controls and an explicit manual review record |
Completion checklist
- Two distinct tasks have clear readers and purposes.
- Approved facts and style examples are separated.
- Voice rules describe observable writing choices.
- Factual accuracy and expression were reviewed separately.
- Substantive edits have reasons and identifiable versions.
- The final brief and examples have an accountable editorial owner.
Next step
Use the revised brief in AdCyte's content workflow for the next suitable output. Read How do you stop AI-written content sounding the same? for the wider editorial approach.
FAQs
Do we need the newest managed-brand features for this exercise?
No. The review method can use current writing controls and an external editorial brief. Verify which features are available in your workspace and avoid relying on unreleased example, calibration or learning controls.
Should we score every draft numerically?
Only if the score helps your team discuss specific criteria. A short judgement with a reason is often more useful. Do not treat an internal voice score as an objective measure of brand quality or factual accuracy.
Can the accepted examples become rules for every channel?
Use them within their stated scope. A campaign passage may not suit a product instruction or social opening. Keep the core identity consistent while explaining how the expression should adapt to the reader and destination.
<!-- Public copy ends. -->


