Guide

Concepts and glossary

Every term BrandKind puts on screen, defined once. The rest of this guide links back here rather than repeating itself, so this is the page worth reading properly.

How your account is arranged

Workspace: The container everything else lives in. Your content, voices, team members, destinations and reports all belong to exactly one workspace, and nothing crosses between them. A marketing agency typically runs one workspace per client; a council might run one for the whole organisation. You can belong to several, and you switch between them explicitly — see Workspaces and branding.
Piece: One item of content: a blog post, an email, a document, a social post or a press release. A piece is created by a generation run and then lives in your library where you can edit, share, export or publish it. It keeps its full history — the voice it was written in, the scores it earned, the images attached to it.
Format: What kind of thing a piece is. There are six, fixed: blog (long-form article), email (newsletter-style email), doc (formal document), social (short social post) and press_release (press office announcement) and case_study (evidence-led customer story). Format is the first question the generate screen asks, and it shapes both the writing and the preview. A social run may also carry a channel, which narrows it to one platform; the other five formats have no platform variants. One run produces one format; to get the same story in another format, see repurpose below.
Channel: The specific platform a piece is written for. A channel always belongs to a format and never replaces one: LinkedIn, X, Facebook and Instagram are all social, and newsletter is email. Choosing a channel changes the writing, not the layout — the length, the opening line, and the hashtag and link conventions all differ from the first sentence. Channels appear in two places: as the second question on the generate screen, asked only when you are writing a social post, and as Channel versions in the editor, which re-cuts a piece you have already written and approved.
Visibility: Whether a piece is private — only its owner can open it — or shared, meaning everyone in the workspace can. Visibility is a property of the piece and is separate from a public share link, which is a different mechanism entirely.

Words about brand voice

Voice is the only mechanism BrandKind gives you for controlling who the writing is aimed at. There is no separate audience setting, sector switch or tone slider — if you want different writing, you change the voice.

Brand voice: A saved description of how you write, reusable across every piece. It is the single most important thing to get right, and the thing most worth spending time on. Full treatment in Brand voices.
Captured brand voice: A voice preset built by BrandKind reading real examples of your existing writing and working out the pattern, rather than you describing it. Captured voices are generally much better than written ones, because most people are poor at describing their own style.
Custom instructions: A one-off alternative to a saved voice, written directly into the generate screen for a single run. Nothing is saved and nothing is reused. Useful for a one-time experiment; a poor substitute for a voice preset if you write regularly. A run uses either a voice preset or custom instructions — never both.
Brand voice glossary: Words and phrases attached to a voice that must be used, spelled, or avoided in a particular way — product names, house style, banned jargon. The writer honours them on every run using that voice.

Words about making content

Brief: What you want written, in your own words, typed into the generate screen. It can be a sentence or several paragraphs. Alongside it you can optionally set a target word count, a target paragraph count and a focus keyphrase.
Run (generation job): One execution of the writing pipeline: brief in, finished piece out. A run moves through queued, running, then either done or failed. Every run is recorded, whether or not it succeeded — see Reporting.
Repurpose: Re-author an existing, approved piece into a different format — turning a blog into a social post, say. It reads the finished piece rather than the original brief, so it carries the edits you made. It is a fresh run and produces a new piece. Not switched on yet (#887).
Improve: Take text you already have, paste it in, and have it re-written in one of your captured voices. Where generating starts from an idea, improving starts from someone else's draft.
Feature image: The image chosen to lead a piece. Images are generated in a set and you pick one; its layout controls how it sits relative to the text. Covered in Images.

Words about quality

Read this before the definitions below

None of these scores block anything. BrandKind will let you publish a piece with the worst score it can produce. They exist to tell you where a draft stands, not to hold a gate.

Brand fit: How closely the finished piece matches the voice it was supposed to be written in, scored from 0 to 10 by a model that reads both. It answers “does this sound like us?” and nothing else — a piece can be excellent writing and score badly because it does not sound like you.
Detection score: How strongly an AI-detection tool believes the text was machine-written, on a 0–100 scale where higher means more suspicious. BrandKind records it before and after humanising so you can see what the pass achieved.
Human score: The detection score inverted — 100 − detection score — so that, like every other number in the product, higher is better. It measures how likely the text is to survive a detector, which is not the same as how well written it is. A clumsy sentence can score highly.
Reading age: The approximate age at which a reader could follow the text, computed from sentence and word structure. It is shown for information only. There is no target, no pass mark, and it never affects the publish check — your audience decides what the right reading age is, and BrandKind does not know your audience.
Accessibility flags: Structural problems that make content hard to use with assistive technology — missing image descriptions, headings that skip levels, link text that says “click here”. Flagged, never enforced.
Publish check: The single verdict that aggregates the above. It reports the worst status among the checks that actually ran, and a check that could not run never drags the verdict down. Thresholds and how to read it are in The editor.

Words about publishing

Destination: A saved place to send finished content — an address on your CMS, plus the wire format it expects and any credential it needs. Set up in Destinations.
Delivery: One attempt to send one piece to one destination, with its own record: status, the HTTP code that came back, how many attempts have been made, and when the next retry is due. Statuses are queued, sent and failed.
Share link: A read-only public URL for a single piece, for showing a draft to someone without an account. Anyone holding the link can read it; nobody can edit through it. This is separate from a piece's visibility inside the workspace.
Content job: A standing instruction to generate on a repeating schedule — a brief, a voice, a format and a cadence — which runs whether or not anyone is signed in. Covered in Scheduling.

Words about people and cost

Workspace role: What a person may do inside one workspace: viewer, member, manager or admin, in increasing order of power. The same person can hold different roles in different workspaces. Exactly what each can do is in Team and access.
Platform role: A separate, account-wide flag that grants RoodtSquared support staff access to the admin console. It is unrelated to your workspace role and no customer holds it.
Credits: The unit BrandKind bills in. Actions that call the AI pipeline consume credits; reading, editing and exporting do not. What costs what is in Billing and credits.
Audit log: A read-only record of significant actions taken in the workspace — who invited whom, who changed a role, who deleted what. Visible to managers and admins.