Skip to main content
Brand voice is what makes generated content sound like you. In Notra it lives in Brand Identity, and every generation path, the Create Content wizard, schedules, event triggers, Iris and GEO Write, picks one identity and writes with it.

​
What is a Brand Identity?

A brand identity is a named profile. A workspace starts with one called Default, created from the website you entered during onboarding, and you can add more, for example Marketing, Technical or Internal. Each identity stores:
  • Company name: your product or company name
  • Website: the site Notra analyzes for the profile, sitemap and guidelines
  • Description: what you build and why it matters
  • Tone profile: Conversational, Professional, Casual or Formal
  • Custom tone notes: free-text guidance layered on top of the profile
  • Language: the language content is written in
  • Custom instructions: rules the writer must follow
  • Who are you writing for?: your target audience
The Brand Identity page has four tabs: Company Info (the fields above), References, Sitemap and Brand Guidelines.

​
How Notra Learns Your Brand

​
Automated analysis

When you add a website, Notra runs a three-step analysis and fills in Company Info for you:
1

Scraping

Notra fetches your homepage and key pages and extracts the main content.
2

Extracting

An AI model reads the content and identifies your company name, description, tone and style patterns, and the audience your messaging speaks to.
3

Saving

The results are written into the identity. Review and adjust any field afterwards.
During onboarding this runs in the background right after you create the workspace. On the Brand Identity page you can re-run it whenever your site changes, and adding a new identity with a website runs it for that identity.

​
Manual configuration

Every field is editable. Changes auto-save a moment after you stop typing, and a toast confirms the update.

​
References

The References tab holds real posts and writing samples, for example tweets, that help the AI match your style beyond what a tone profile can express. Add a few pieces you are proud of; plans differ in how many references are included.

​
Sitemap and Brand Guidelines

The Sitemap tab tracks your indexed pages and monitors site health for AI discovery. GEO Write uses it to pick internal links for new articles. The Brand Guidelines tab pulls logos, colors, typography and landing page screenshots from your website. Generate them once, refresh when your design changes, and edit individual entries.

​
Tone Profiles Explained

Each tone profile selects a different prompt template for every format:

​
Conversational

Best for: developer tools, startups, open source projects
  • Warm and authentic
  • Direct address (“we” and “you”)
  • Clear and specific without being dry
  • Founder-to-community feel
Example:
We’ve shipped some meaningful improvements this week. The new cache component support catches unsupported auth calls early and provides clear migration guidance. Email verification flows now support secure link completion with proper error handling for expired links.

​
Professional

Best for: enterprise products, B2B SaaS, corporate communications
  • Clear and confident
  • Polished without being stiff
  • Focus on business value
Example:
This release delivers significant enhancements to the platform. New cache component validation ensures proper auth implementation with actionable error guidance. Email verification workflows now include comprehensive link-based completion flows.

​
Casual

Best for: consumer apps, community-driven products, creative tools
  • Friendly and approachable
  • Relaxed language
  • Enthusiasm visible
Example:
Hey folks! We’ve been busy this week. The cache system is now way smarter about catching auth issues before they become problems. Plus, email verification links actually work the way you’d expect, nice and reliable.

​
Formal

Best for: finance, healthcare, compliance-heavy industries, academic tools
  • Precise and structured
  • Traditional language
  • Complete sentences
  • Conservative terminology
Example:
The engineering team has completed several deliverables for this release cycle. Runtime validation capabilities have been enhanced to detect unsupported authentication patterns in cached contexts. Email-based verification procedures have been implemented with comprehensive error state handling.
Custom tone notes refine the profile without replacing it. “Technical but enthusiastic about developer experience” on top of Conversational works better than trying to describe the whole voice from scratch.

​
Language

The Language field tells the writer which language to produce. The rule is strict: the final text is written primarily in that language, and English is only used for terms that the language commonly borrows, such as technical terms and product names. This applies even when your custom instructions and references are in English.

​
Target Audience Impact

Who are you writing for? changes both filtering and formatting.

​
Developer audience

When the audience is developers, engineers or technical teams:
  • PR links included: every referenced change links to its pull request when one is available
  • Technical terminology: proper terms without over-explaining
  • Implementation detail: APIs, components and architectural changes are fair game
  • Author attribution: contributors are credited with their GitHub handle
Example:
Cache component support with actionable error guidance #423 - Runtime guardrails now catch unsupported auth calls in cached contexts and provide clear migration guidance with the correct usage pattern. (Author: @alex)

​
Non-developer audience

When the audience is customers, end users or stakeholders:
  • No PR numbers or links anywhere in the output
  • Plain language that explains what changed in terms of benefits
  • Outcome focus: what users can now do
  • Author attribution still included
Example:
Improved error messages: When you use authentication features incorrectly, you’ll now see helpful guidance on how to fix it. (Author: @alex)
The audience also decides what gets filtered out. A change that only matters to engineers is dropped entirely for a customer audience rather than being translated into vague copy.

​
Custom Instructions

Custom instructions add rules beyond the tone profile:
Be specific. “Use simple language” is vague. “Explain technical concepts using analogies from everyday life” gives the writer a rule it can follow.

​
How Brand Voice is Applied

When generating content, the identity is formatted into structured tags and placed in the prompt:
Two rules travel with these tags:
  1. <company> is the publishing identity. Sources are evidence only, so a repository, workspace or team with a different name never causes the writer to hesitate or attribute the work elsewhere.
  2. If <language> is present, the final text is written in that language.
The tone-specific system prompt for the chosen format wraps all of this, and any skills you enabled on the Skills page, such as the human-writer skill, run on the result.

​
Multiple Identities

Use separate identities when one voice does not fit every channel:
  • Marketing for blog posts and LinkedIn, Conversational, customer audience
  • Technical for changelogs, Professional, developer audience
  • Internal for team updates, Casual
The Create Content wizard, schedules, event triggers and GEO Write each let you pick the identity for that run. One identity is marked as the default and is used when nothing else is selected.

​
Maintaining Consistency

Same voice everywhere: a changelog and the tweet announcing it come from the same identity, and scheduled, event-triggered and manual runs all use the same rules. Consistent structure: changelogs always follow Summary, Highlights, More Updates, with categories in a fixed priority order. Quality controls:
  • Summaries target 120 to 180 words
  • No made-up PR numbers or commit details
  • Runs with nothing audience-relevant are skipped instead of padded

​
Examples Across Tones

Same feature, four tones: Feature: New GitHub webhook verification for enhanced security
We’ve added webhook signature verification to make sure incoming GitHub events are actually from GitHub. If a webhook fails verification, we’ll reject it and log the attempt. This prevents potential security issues from forged webhook payloads.
GitHub webhook signature verification has been implemented to ensure all incoming events are cryptographically authenticated. Invalid signatures are rejected with appropriate logging, preventing unauthorized webhook delivery attempts.
Your webhooks just got a lot more secure! We’re now checking signatures on every GitHub event to make sure it’s legit. Sketchy requests get blocked at the door.
Cryptographic signature validation has been implemented for all GitHub webhook payloads. The system verifies request authenticity prior to processing and maintains audit logs of verification failures in accordance with security best practices.

​
Updating Your Brand Voice

Update an identity whenever you rebrand, your audience shifts, or you want to try a different tone. Future runs use the new settings; existing drafts keep the voice they were generated with.
After changing an identity, generate one post from the Create Content wizard to check the result before your schedules run with it.

​
Best Practices

Tone: match your existing marketing and documentation. Fintech tends toward Formal, developer tools toward Conversational. Change it rarely. Custom instructions: start small and add rules as you notice yourself making the same manual edit twice. Keep them about style, not content. Audience: be specific. “Senior engineering leaders at enterprise companies” beats “technical people”, and it decides whether PR links appear. References: add real posts. Three good samples do more for style than a paragraph of description.

​
Troubleshooting Voice Issues

Too technical or not technical enough: adjust the audience description first, then add a custom instruction about detail level. Wrong tone: check the profile, then make sure custom tone notes and custom instructions do not contradict it. Wrong language: set the Language field. Instructions written in English do not switch the output language. Inconsistent style across posts: make sure every schedule and trigger points at the same identity, and that the identity has all fields filled in.
Last modified on September 2, 2026