Brand Book
The facts RankX AI holds about your business, in six tabs. What each one feeds, and why a profile can be complete and still not ready to generate from.
Brand Book is where you tell RankX AI what your business actually is, and it is
the single input behind more of the product than any other screen. Your
generated llms.txt is a projection of it. The
answer accuracy check judges what
assistants say against it. Content Studio writes from it. Several
AI Readiness checks look for facts that only
exist here.
Brand Book lives at Settings, then Brand Book, and it is scoped to one Website. It is unmetered: nothing on it calls a provider or spends credits, apart from the one control that says so.
The six tabs
| Tab | What it holds |
|---|---|
| Overview | The essential facts: brand name, industry, short description, primary service, target audience, business location, primary language. Also your ideal-customer segments, the Website's own details, and your review sources |
| Products & Services | Everything you sell, with an optional structured price per service |
| Voice & Style | How your business writes: tone, reading level, the words you use and the ones you avoid |
| Knowledge & Proof | Approved claims, proof points, case studies, testimonials, FAQs, awards, accreditations, and the two negative types below |
| Authors & Experts | The real people your content is attributed to |
| Competitors | The rivals RankX AI measures you against |
Competitors used to be its own screen and Brand Book used to be its own section of the sidebar. Both now live here, in the settings shell, because both are configuration rather than measurement.
Two numbers, and they measure different things
Brand Book shows a profile completeness percentage and a generation readiness state. They are deliberately separate, and a profile can be 100% complete and still not ready to generate.
Profile completeness counts eight essential fields: brand name, industry, short description, primary service, target audience, business location, primary language, and whether you have listed at least one product or service. It counts a draft service. It never rewards an optional or advanced field, so the percentage cannot be inflated by filling in things nothing reads.
Generation readiness asks a harder question: is there at least one service or knowledge item that Content Studio could actually publish from? That needs the item to be approved, public and not confidential, which is a stricter test than merely existing.
Somebody who fills in every Overview field and approves nothing has a complete profile and an empty generation pipeline. Showing one number would have hidden exactly that case.
Approved, public and not confidential
This is the rule that decides whether a Brand Book row ever leaves your account, and it applies identically to every consumer.
A service or knowledge item reaches your generated llms.txt, an article draft,
or the ground truth behind an accuracy verdict only when it is approved,
public and not marked confidential. The same predicate governs all three
paths, deliberately: a looser local copy in any one of them would be a leak, and
a public file reaches every crawler at once and cannot be taken back.
Two knowledge types are the exception, and they run the other way:
- Restricted claim, something that must not be said about you without qualification.
- Prohibited phrase, something that must never be said about you at all.
Neither is ever published. Both are inputs to the accuracy check, where a prohibited phrase appearing in an assistant's answer is the worst verdict available.
Prices are structured, not prose
A price in Brand Book is an amount, a currency and a unit, rather than a sentence. The unit set is one-off, hour, day, month, year, per user per month, and project.
The structure is load-bearing rather than tidiness. The same figure is rendered
into your llms.txt and read back by the accuracy checker's price parser
when it decides whether an assistant contradicted you. A price stored in a form
that parser cannot read would silently downgrade a "contradicts your own facts"
verdict to "says something you have never published", telling you that you have
published nothing when you have. Every currency the picker offers is proven to
survive that round trip.
The practical consequence for you: publish a price on at least one service. It is the single highest-value field in Brand Book, because stating a price at all is one of the checks AI Readiness scores, and it is the fact buyers ask assistants about most.
What reads Brand Book
| Consumer | What it takes |
|---|---|
Generated llms.txt | Approved, public services and knowledge, rendered as a description of what the business is |
| Answer accuracy | The same facts, used as ground truth for the seven verdicts |
| AI Readiness | Pricing, integrations and services, for the checks that ask whether a buyer question has an answer anywhere |
| Content Studio | Services, knowledge, voice and authors, when drafting |
| AI Visibility | The Competitors tab, which is the rival set every share-of-answers number is computed against |
The generated llms.txt is worth one caution. RankX AI will not write a
correction into it. If an assistant is describing your business wrongly, the
temptation is to publish a sentence denying it, and that publishes the false
sentence machine-readably on your own domain, where it becomes the best-cited
source for it. Corrections belong in the positive assertion: state the true
thing plainly, in the words a buyer would use, first.
Who can edit it
Brand Book is not staff-only, and that is deliberate. It describes the client's own business, so an Agency Client looking at their own workspace can open it and edit it. Several of the Website band's other sections are closed to them, and Settings says which and why.
Removing rows is the one narrower permission. The delete routes are staff-only, and whether the delete controls render is resolved from the database role rather than from the session token, which can stay stale for up to a token lifetime after somebody is demoted. It fails closed: a role that cannot be read is not treated as staff. So the control never appears for somebody whose delete request would be refused.
One control on the page is metered: Suggest with AI on the Competitors tab runs competitor discovery at its own published rate. Everything else on Brand Book is free, including on free access after a trial, where Brand Book is one of the three surfaces that stay open.
Where to go next
- Settings, for the rest of the shell and what each section holds.
- Answer accuracy, the check Brand Book supplies the truth for.
- AI Readiness, for the checks that read your published facts.
- Content briefs, for what Brand Book does when an article is drafted.
Last updated
Settings
Every section of RankX AI's settings, grouped by what it belongs to. The three scopes, what a client user never sees, and where each one is documented.
Plans and limits
Every published RankX AI plan limit, on both tracks, generated from the same source the pricing page renders. What each limit caps, and what it does not.