You asked for output that is “professional but friendly, authoritative but humble, technical but accessible”. Each pair sounds reasonable. The model produced a five-paragraph essay that swings between sentences — one sentence sounds like a McKinsey memo, the next like a friendly Slack message, the third like a textbook footnote. Re-prompting “blend them more smoothly” yields a smoother but equally inconsistent voice. The problem is not the blending. The problem is that you asked for two voices and the model averaged them, and averages of opposing voices read as schizophrenic.
This page walks through why tone adjectives stacked with “but” or “and” fail, and how to pick one primary tone with explicit nuances anchored by examples.
Common causes
1. Two tones joined by “but” or “and”
“Formal but friendly” reads to you as a delicate blend; reads to the model as two incompatible style targets to satisfy at once. It picks one per sentence, randomly.
How to spot it: your tone instruction has the word “but” or “and” between two adjectives.
2. No example of the desired blend
You described the blend in prose. The model has no anchor for what your “professional but friendly” actually sounds like. Different training-distribution averages of each word collide.
How to spot it: tone described in 1-2 sentences with no sample paragraph.
3. Tone wishes hide stakeholder politics
Sometimes “professional and accessible and innovative and on-brand” exists because you have 4 stakeholders. Output suffers because the model is asked to satisfy a committee.
How to spot it: 4+ adjectives in the tone slot.
4. Conflicting tones are inherent to the domain
Some tasks genuinely sit at a tension: a sales email needs warmth and authority. The model has to choose per sentence and you cannot fix this with adjectives alone.
How to spot it: tasks like sales, fundraising, conflict resolution — domain inherently mixes registers.
5. Tone instruction is a single phrase, not structured guidance
“Make it warm and authoritative” leaves the model with no recipe. Specific rules (“Use ‘we’ in section 1, switch to ‘I’ in section 3”) would land far better.
How to spot it: one short tone phrase, no structural mapping.
Before you change anything
- List every tone adjective in your prompt.
- For each pair, ask “do these conflict?” — if yes, you cannot satisfy both equally.
- Find one example of writing that hits the blend you want.
- Decide one primary tone before re-prompting.
- Map secondary tones to specific structural positions if relevant.
Information to collect
- Full tone instruction as written.
- Output that mixed tones unevenly.
- One sample of writing that hits the blend you actually want.
- Model, temperature, system prompt.
- Whether the inconsistency is sentence-by-sentence or paragraph-by-paragraph.
Shortest path to fix
Step 1: Pick one primary tone, demote others
Bad: "Be professional but friendly."
Good: "Primary tone: professional.
Friendliness shows up only through: contractions ('we're' not 'we are'),
'you' instead of 'one', no exclamation marks.
Otherwise default to professional register."One winner, others reduced to mechanical rules.
Step 2: Anchor with a one-paragraph example
Paste a 2-3 sentence sample of the exact blend you want:
Voice example (write in exactly this voice):
"Quick note on the rollout: we're holding launch until Friday. The
auth flow is failing on Safari; we want it solid before customers see it.
Tracking the fix in INC-4123."Examples > adjectives, always.
Step 3: Map nuances to structural positions
If you genuinely want two registers, give each one a place:
Section 1 (opening): warm, "we", contractions.
Section 2 (technical detail): formal, third person, no contractions.
Section 3 (call to action): warm, direct, second person.The model is much better at “use X in section 2” than at “blend X and Y throughout”.
Step 4: Convert tones to measurable rules
Tone rules:
- Contractions: allowed
- Exclamation marks: forbidden
- Second person: required
- Industry jargon: max 1 per paragraph, defined on first use
- Average sentence length: 12-18 wordsMeasurable rules eliminate the per-sentence drift.
Step 5: Pin tone in a system prompt or project instruction
In-message tone instructions drift across turns. Move tone to project / system prompt where it persists.
Step 6: For inherently mixed-register tasks, write a recipe
For a sales email:
Recipe:
- Open: warm, name the recipient, reference a recent shared context.
- Middle: shift to authoritative — concrete numbers, specific named cases.
- Close: warm again — invite next step, no hard-sell.A recipe is a structural anchor that adjectives cannot provide.
How to confirm the fix
- Output sentence-to-sentence reads coherent, not stitched.
- A blind reader cannot identify which “tone” — it just sounds like a voice.
- Running the same prompt 3 times produces 3 outputs in a consistent voice.
- The measurable rules pass on every sentence.
If it still fails
- The two tones may be genuinely incompatible — cut one.
- Reduce sample size: have the model write 2 sentences first, then expand.
- Provide 2-3 examples instead of 1; tone gets pinned more reliably with more anchors.
- For brand work, treat tone as a config file: examples + rules + position mapping, all locked.
Prevention
- Default: name one primary tone. Everything else is a mechanical rule or a positional nuance.
- Save tone anchors (one short paragraph per tone) for reuse.
- Never use “but” or “and” between adjectives in tone instructions.
- Audit production prompts quarterly for accumulated tone wishes.
- For team workflows, agree on one voice and one anchor — not “the brand vibe”.
- Test tone with a blind read: hand the output to someone who has not seen the prompt; if they describe a coherent voice, it landed.