The short version
- Start with the decisions a writer needs to make, then collect the evidence behind them.
- Separate customer observations, search signals, approved facts and working hypotheses.
- Test the reference on a real brief and use review corrections to keep it useful.
What is a Brand Brain, and when do you need one?
A Brand Brain is a maintained reference that connects what a business knows with the decisions its content team needs to make. It records the audience, their questions, the evidence available, the brand’s position and the boundaries of what it can credibly say. Its purpose is to help a writer produce a useful brief and an accurate draft without rediscovering the business on every assignment.
You need this kind of reference when the same gaps keep returning during review. Writers choose the wrong audience. Product claims need correcting. Articles sound interchangeable with competitors. An expert rewrites the introduction because the draft has missed the real question. Adding another tone-of-voice adjective will not resolve those problems.
We use Brand Brain as the name for this part of our editorial strategy work. It is not a standard software category, a ranking factor or a requirement to purchase a particular tool. You can build the reference in documents or a database. The important design choice is whether someone can follow an editorial decision back to its evidence.
If your briefs are already accurate and reviewers mostly make small stylistic changes, a new knowledge system may not be your priority. If the team is repeatedly correcting the premise, start by examining what information the writer had before drafting.
Find the missing context before collecting more documents
Take a recently reviewed article and classify the substantive corrections. Which changed a fact? Which changed the intended reader? Which added an exception? Which challenged the usefulness of the whole piece? This is an editorial diagnostic exercise, not a performance score for the writer.
Trace each correction back to the brief. If the answer existed but was hard to find, improve retrieval. If two sources disagreed, establish which one the business stands behind. If nobody had supplied the answer, record a research question. Do not treat all three situations as a need for a longer brief.
Use this review to set the initial scope. A company whose articles repeatedly confuse implementation with onboarding needs clearer audience and task definitions. A company whose product descriptions drift needs approved specifications and an owner for updates. Different failures require different reference material.
The table below is a suggested diagnostic. Its value is in helping the editor choose the next action, rather than generating another document the team has to maintain.
| Repeated correction | What to add or clarify |
|---|---|
| This is for the wrong reader | Buying situation, task and prior knowledge. |
| We cannot promise that | Source, approved wording and claim boundaries. |
| We already answered this | Existing article, its purpose and the new piece’s distinct question. |
| This does not sound like us | An edited example with an explanation of the choice. |
| The expert has to rewrite it | Expert reasoning, exceptions and evidence needed before drafting. |
How do you find audience questions you can trust?
Start with material produced while people were trying to solve a real problem: questions in a sales conversation, a support exchange, a research interview or search queries associated with your website. Keep the context attached. A prospect asking about migration before a purchase and a customer struggling with setup after purchase may use similar words but need different content.
GOV.UK’s user-research guidance recommends learning who users are, what they are trying to do and the problems they encounter. It also distinguishes evidence about actual users from assumptions. That is a useful discipline for editorial research, although the guidance itself concerns service design rather than content marketing. GOV.UK: learning about users and their needs
Ask the subject expert for a specific recent situation. What was the person trying to decide? What did they misunderstand? Which detail changed the answer? What evidence did the expert use? Preserve the difference between the buyer’s recorded words and the expert’s interpretation of them.
Existing articles are another useful input, but their topics show what the previous editorial team chose to cover. They do not establish what customers asked for. Competitor pages have the same limitation. Use both to find questions worth investigating and coverage that may already exist, then seek evidence for the underlying need.
When direct customer evidence is unavailable, start with the strongest available signal and name its limits. An unanswered question can become a research task. It should not silently become a confident audience insight.
Separate what you observed from what you think it means
A useful reference lets the writer distinguish four kinds of material: an observed statement or behaviour, a verified business fact, an interpretation of evidence and an untested hypothesis. These categories can sit together, but they should not look equally certain.
GOV.UK’s research-analysis method separates observations from findings and subsequent actions. For content teams, the practical adaptation is to retain the original observation beside the editorial inference. A reviewer can then challenge the interpretation without losing the source. GOV.UK: separating observations, findings and actions
Consider this illustrative B2B example. A prospect asks whether a system can import historical records. The observation is a question about imports. A possible interpretation is that migration effort matters to this evaluation. The proposed content action is a migration guide. You still do not know that migration is the main reason the account might buy, or that every prospect has the same concern.
Store a source reference, date, context and owner with each important record. For a claim that changes frequently, add a recheck trigger. If a source is confidential, keep a restricted original and a usable approved summary. Giving writers context does not require exposing every customer conversation to everyone.
| Record type | What the writer can do with it |
|---|---|
| Observed question | Describe that question in its documented context. |
| Verified product fact | Use the supported claim within its stated scope. |
| Editorial interpretation | Use it to shape an angle, with its uncertainty visible. |
| Working hypothesis | Research or validate it before treating it as established. |
Use search data without pretending it explains the whole buyer
Search Console can help identify queries associated with a site’s search appearances and clicks. Google describes its performance report as a view of how the site performs in Google Search. It is evidence about that search interaction, not a transcript of the visitor’s buying process. Google Search Console: what the performance report measures
In the Brand Brain, record the query, relevant page, date range and audience hypothesis. Then ask what the query leaves unanswered. Someone searching for implementation costs might be comparing vendors, building a budget or checking an existing invoice. The phrase alone does not settle their role or stage.
Group query variations around a decision the article could help with. A page about a cost estimate should explain what changes the estimate, what is included, what is excluded and where the reader needs a specific quote. Repeating several versions of the query does not add that information.
Check the existing library before proposing a new page. If a current article already answers the same question for the same reader, the useful action may be a revision. If the new need is distinct, write down the boundary and the internal link between the two pieces. This is how the reference starts shaping a roadmap rather than accumulating keywords.
What should the first usable version contain?
Organise the reference around editorial decisions. A folder full of source documents makes useful evidence available; it does not necessarily make the evidence usable. A writer should be able to answer who the article serves, which question it owns, what supports the answer and who can resolve a gap.
Keep audience situations separate where they change the argument. For a property business, a homebuyer evaluating everyday use and an investor assessing an investment have different decisions to make. The Brand Brain should preserve that distinction in examples, evidence and the next step, rather than merging both into a broad property-interested persona.
Voice guidance needs examples of judgement. Show an overconfident draft sentence beside the approved revision, then explain what changed. Perhaps the revision names an assumption, replaces a sweeping promise or puts the useful answer first. A small set of explained edits gives a writer more to apply than a long list of adjectives.
Assign responsibility for factual changes separately from editorial maintenance. An editor can maintain the reference structure; the product or subject owner may need to confirm the underlying fact. Without that distinction, an orderly document can still contain unapproved claims.
| Reference area | Decision it should support |
|---|---|
| Audience and evidence | Who needs this, and how do we know? |
| Offer and claim library | What can we say, prove and qualify? |
| Expertise and sources | Who can answer the difficult part? |
| Voice examples | How do we explain this in the brand’s own manner? |
| Content ownership | Which page answers which question? |
| Gaps and updates | What needs checking before this is assigned? |
Turn a research record into a brief
A brief should use the relevant slice of the Brand Brain. Pasting the whole reference into the assignment makes the writer responsible for deciding which facts and audience assumptions matter. Make those decisions explicit before commissioning the draft.
Here is an illustrative brief for a software migration article. The intended reader is an operations lead evaluating a move from an existing system. The decision is what to confirm before agreeing to migrate historical data. The article should help them prepare questions and identify dependencies, not promise a migration timetable the vendor has not verified.
The evidence package would include current import documentation, an approved explanation from the implementation owner and any relevant restrictions on supported data. The brief should list unanswered questions such as how attachments, custom fields or historical permissions are handled. If those details cannot be verified, the writer should not replace them with a generic reassurance.
Specify what makes the piece useful: a preparation checklist, a comparison between supported and unsupported inputs, and questions to ask before committing. Also specify what it should not do, such as imply that every migration can happen without interruption. The next step can be a requirements discussion rather than a premature request to buy.
Ask a writer who did not assemble the research to explain the brief back. If they cannot identify the central question, strongest evidence and unresolved dependency, improve the assignment before adding more prose.
What this looked like in the TVProperty preparation
The TVProperty workspace separates brand facts, offerings, audience profiles, pain points, voice, search topics and the roles of individual websites. That structure supports decisions about both the audience and where an article belongs. It is the documented foundation behind our Brand Brain case study.
The source material also distinguishes provisional audience assumptions from confirmed evidence. That distinction matters as much as the contents of the reference: a plausible objection does not become an attributed buyer quote merely because it appears in an editorial file.
The topic queue connects a candidate article to an audience, a demand signal, an angle and a check against existing coverage. The production handoff then adds article and publishing requirements. These are different stages with different purposes: the reference supplies context, the queue chooses the assignment, and the brief defines what the writer must deliver.
This example describes preparation documented in the project files. It does not establish a measured reduction in revisions or a ranking improvement. The useful lesson is the traceable connection between source material and production decisions, which a team can inspect before any performance claim is available.
Can you use the Brand Brain with AI?
Yes, as reference material for tasks such as finding relevant records, proposing an outline or identifying unanswered questions. Require the output to point back to the records it used. Then check those records and whether they support the exact claim in the draft.
NIST’s Generative AI Profile identifies confabulation as confidently stated false or erroneous content. A supplied reference does not turn generated wording into verified evidence. The editorial check still needs to establish that a citation exists, applies to the subject and has not been stretched beyond what it says. NIST: Generative AI Profile, confabulation risk
For a practical acceptance test, give the workflow a question whose answer is present, one that needs combining two records, and one the reference cannot answer. The third matters: a useful draft should expose the missing information rather than manufacture a reassuring sentence. This is a suggested editorial test, not a guarantee of model reliability.
Keep search expectations grounded too. Google’s guidance for AI features says established SEO practices remain relevant and there are no additional technical requirements specifically for inclusion. A Brand Brain can improve the evidence and clarity available to a writer; it does not guarantee an AI citation. Judge the reference first by the quality of the decisions and content it enables. Google: AI features and your website
How do you know whether it is helping?
Compare the same kinds of assignments before and after the reference is used. Record substantial factual corrections, questions sent back to experts, revisions caused by the wrong audience and delays waiting for missing evidence. These are suggested operational measures; their purpose is to find where the reference still fails the team.
Do not turn fewer comments into the sole success measure. Review can become quieter because the brief improved or because nobody checked closely. Inspect whether the published or prepared piece answers its intended question and whether its central claims can still be traced to their sources.
Feed repeated corrections back into the relevant record. When an approved specification changes, flag dependent briefs and articles for review. Retire superseded guidance visibly so the team is not choosing between two apparently current answers.
Start with the next real assignment. Build enough context to brief it well, test the result, then extend the reference where the next decision requires it. A Brand Brain earns its place in the workflow when the team uses it to resolve uncertainty, not when the folder looks complete.
Practical questions before you start
Is a Brand Brain the same as a brand book? They overlap, but their jobs differ. A brand book may define identity and voice. This reference also connects audience evidence, factual claims, source ownership and content decisions. Keep an existing brand book as an input rather than rebuilding it unnecessarily.
What if you have no customer interviews? Use available evidence such as search data, support questions and approved expertise, with its limitations labelled. Record the unanswered audience questions and plan direct research. Do not invent customer language to make the reference feel complete.
How much should you build before drafting? Enough for a writer to identify the reader, the decision, the evidence and the unresolved questions for one assignment. There is no universal page count. The first brief provides a more useful test than the size of the document.
Who should own it? Give editorial maintenance a named owner and factual approval to the appropriate subject experts. Make it clear who can propose a change, who confirms it and how affected content is identified.
