Words are the part AI touches most
Of everything a brand does, language is where AI shows up first and most often - every tagline, post, email, and product line. That is exactly why the verbal identity has to be nailed down: a name and a voice guide are the rails that keep all that AI-generated language sounding like one brand instead of the category average. Tonight you build both for Cadence, our sample AI note-taker.
Verbal identity: name, tagline, voice, never-say 6 min live
Verbal identity is not just a logo with words. It is four connected parts: the name you go by, the tagline that carries the promise, the voice that governs how you always speak, and the never-say list that draws the line. Together they are one system - and the most important thing to know about it is that examples teach a model far better than adjectives ever will.
LiveThe four parts, and how they fit together3 min▶
Each part has a distinct job, and each one is a place AI helps you explore:
- Name - the word people say and search. It carries the most weight and is the hardest to change later.
- Tagline - one line that carries the positioning as a feeling, not a feature list.
- Voice - the consistent way the brand speaks across every channel: a set of principles, each shown with a do and a don't.
- Never-say list - the banned words, claims, and tones. Small list, huge leverage - it removes most off-brand output in one line.
Self-studyWhy examples beat adjectives, every time2 min read▶
Adjectives are the sameness trap in disguise. Ten brands all ask AI to be "bold, innovative, and authentic" and get ten versions of the same beige paragraph, because those words describe the average. A model has no shared definition of "warm" - but it can copy a sentence you show it. So a voice guide built on sample lines and a never-say list steers hard off the average; one built on adjectives just rejoins it. This is the whole reason the verbal identity is the highest-value brand doc in the AI age.
Naming with AI - and the trademark rail 5 min live
AI is a naming machine: it will hand you a hundred candidates in a minute across every style. That speed is the gift and the trap. You still have to screen for meaning and memorability - and you absolutely must check trademark and domain before you commit, because an AI-generated name may already belong to someone else. AI names, AI does not clear them.
LiveThree naming styles, then the check that is not optional3 min▶
Most names fall into one of a few styles. AI can generate all of them on demand - your job is to pick the style that fits the strategy, then screen the output:
- Descriptive - says what it does ("MeetingNotes AI"). Instantly clear, but weak differentiation and hard to protect.
- Evocative - a real word borrowed for its feeling ("Cadence" - the rhythm of a good conversation). Memorable and ownable.
- Invented - a coined word ("Notion", "Zapier"). Maximum trademark room, but you have to spend marketing to give it meaning.
Self-studyWhich naming style fits when2 min read▶
| Naming type | When it fits |
|---|---|
| Descriptive ("MeetingNotes AI") | You need instant clarity and search traffic; you accept weak differentiation and a hard trademark path. |
| Evocative ("Cadence") | You want memorable and ownable, and you have a story to explain the link. Easier to protect. |
| Invented ("Notion", "Zapier") | You want maximum trademark and domain room, and you can invest to give the word meaning. |
| Compound / blend ("Otterly", "Fireflies") | You want a bit of both - a hint of meaning plus room to own it. Watch that it still says easily. |
Cadence is evocative: it borrows the rhythm of a well-run conversation, which is exactly what a calm note-taker protects. That is why it survived the shortlist - the meaning matched the strategy, not just the sound.
Voice rules + the never-say list 4 min live
Voice is where the AI age rewards you the most. A voice guide is a handful of principles, each shown with a do and a don't, plus a never-say list of banned words, claims, and tones. That never-say list is the highest-leverage line in the entire brand system: it removes most off-brand output before you write a single positive rule.
LiveFour Cadence voice principles, shown not told2 min▶
Cadence is calm, capable, and quietly on your side. Here is that turned into rules a model can follow - each with a do and a don't:
- Calm, never hyped. Do: "Your notes are ready." Don't: "Revolutionize your meetings!"
- Plain over clever. Do: "We wrote down what was decided." Don't: "Synergize your action items."
- On your side, not in your way. Do: "You talk. We will capture it." Don't: "Leverage our AI engine to maximize output."
- Confident, not boastful. Do: "Every decision, captured." Don't: "The world's most advanced note-taker."
Self-studyThe never-say list does the heavy lifting2 min read▶
A never-say list is short and blunt: the words, claims, and tones the brand refuses. For Cadence that is the hype vocabulary ("revolutionary", "seamless", "game-changer", "unlock", "cutting-edge", "synergy"), the exclamation-heavy tone, and any claim it cannot back up. One line like this removes more off-brand output than pages of positive guidance, because a model finds it far easier to obey a boundary than to interpret an ideal. Ship the never-say list first; refine the positive principles after.
Generate + screen a Cadence name and tagline shortlist ★ 13 min · everyone builds
Run the naming process end to end: explore at volume with AI, screen for meaning and memorability, then apply the human clearance step that no model can do for you. You walk out with a shortlist you actually trust.
Open any chat AI. Feed it Cadence's positioning and personality from b2-b3 so the names are on-strategy, not random.
Explore: ask for 20 name candidates across all three styles - descriptive, evocative, invented - plus a one-line reason each fits a calm, present note-taker.
Screen: cut to a shortlist of five on meaning and memorability. Say each out loud - if it is hard to say or spell, it is out.
Clear (human, required): for each finalist run a trademark search in your market, a domain and handle check, and a cross-language sense check. Drop anything taken or risky - this is the rail.
Then ask for 5 taglines for your chosen name that carry the promise as a feeling. Screen the same way. This is how Cadence and "Be present. We will take the notes." survived.
Build Cadence's voice guide ★ 8 min · build your own
Now the artifact that pays off for the rest of the course: a compact voice guide - four principles with do/don't examples and a never-say list. This is the exact document that will power the Cadence brand GPT and the consistency simulator in b6, so build it to be used by a model, not admired in a PDF.
Give the AI Cadence's personality words and ask it to draft four voice principles, each with a short DO line and a DON'T line - real sentences, not adjectives.
Ask it to propose a never-say list: banned words, banned claims, and banned tones. Push it to include the hype vocabulary the whole category overuses.
Edit hard. Rewrite every example in Cadence's actual voice - this is rail one, your judgement doing the work the model cannot.
Keep it to one page. Save it as "Cadence voice guide". You will paste this straight into the b6 brand GPT and watch the consistency meter climb because of it.
The six-word list that outperformed a style guide. A team had a beautiful 40-page brand book and their AI content still drifted into hype. They added one never-say line to their prompts - six banned words and "no exclamation-mark selling" - and the drift stopped that week. The boundary a model could obey did more than all the prose it could only skim. That is why your voice guide, not the model, is the thing that keeps Cadence sounding like Cadence.
Try it yourself - this week ◐ 30-45 min total
- Finish Cadence's name shortlist if you did not live - and actually run one finalist through a trademark and domain check so the clearance step becomes a habit.
- Complete the Cadence voice guide: four principles with do/don't examples and a never-say list. This is the b6 input, so make it real.
- Run the voice-guide prompt on your own brand (or one you admire). Note the never-say words you have been shipping without meaning to.
- Say your top three names out loud to someone who has not seen them. If they mishear or misspell, it fails the memorability test.
- Optional: draft one tagline for your real brand that only it could say - then check it against your never-say list.
Official sources covered
Taught from public brand-AI curricula and brand-voice guidance. Certificates and paid tooling stay official. This session covers:
Three questions before you go 🎯 ◐ 90 seconds
1 · What does a verbal identity include?
Verbal identity is the whole way a brand speaks: name, tagline, voice principles, and the never-say list that draws the line. Visuals are a separate session (b5).
2 · Why must an AI-generated name still get a trademark check?
AI generates names; it does not clear them. Clearance - trademark, domain, cross-language - is a human rail, because shipping an infringing name is a real legal exposure.
3 · Why do examples and a never-say list beat adjectives for AI voice?
"Be bold and authentic" describes the average, so every brand gets the same beige output. A do/don't example and a never-say line are concrete - the model can actually follow them.