Put Your Brand in the Answer | GEO Product Notes 15: How GEO Changed My Understanding of Brand Growth

How GEO changed my understanding of brand growth

For a long time, I drew brand growth as a line moving forward.

First produce content, then gain exposure. Exposure brings visits; visits become inquiries, leads, and deals. Each layer has its own number. The team’s job is to bring more people into the path and lose as few as possible along the way.

That model is not wrong.

When I began building a GEO product and actually watched AI answer individual user questions, I found that the line is far less smooth than the diagram suggests.

An article being published does not mean AI has seen it. AI seeing it does not mean it understands the brand. A brand being mentioned does not mean it was described correctly. Appearing in an answer does not mean it entered a user’s choice. Even when a monitoring result improves, timing alone cannot assign the change to that article.

Too much sits in between.

In Note 14, I wrote about rebuilding a brand diagnosis: the formula could be correct while the samples answered the wrong question. After that iteration, I looked at “growth” again and realized that the change was not only in the product design.

After working on GEO, I no longer see brand growth only as expanding exposure. I see it as a continuing effort to reduce misunderstanding.

The brand has to move closer to the user questions that matter, make its facts clear, and let public information reinforce itself. The team has to see what AI actually said, identify the gap, act on it, and return to the same question to verify the change.

This process is less dramatic than a rising curve, but it is closer to the growth I believe in now.

01 The starting point changed from “where should we publish?” to “what are users asking?”

When discussing content growth, the easiest place to start used to be channels.

Should we run a newsletter? Add short video? How should Zhihu, search, and industry media be divided? Channels matter because content needs a public place. But since working on GEO, I start there less often.

I would rather ask first: In which user questions should this brand have a reasonable place?

One brand can face very different questions. Someone may be learning the category for the first time. Someone else may be comparing options, concerned about price and risk, or checking a brand’s capabilities, reputation, and fit after hearing its name.

Those questions sit at different distances from the business, so the information the brand should provide is different too.

When the questions are not distinguished, more content does not necessarily mean better direction. A team may publish a series of category explainers without answering the questions people actually use to choose. It may repeat the brand name in a scenario it does not serve simply to create more appearances.

That is why I put the question library at the beginning of the product.

Not because a “question” is more sophisticated than a “keyword,” but because a question contains a person, a situation, and a task. It forces the team to say who is making what decision, why the brand should participate, and which part of the information is missing.

It also changed how I think about the scope of growth.

A brand does not need to appear in every question. If a question has nothing to do with the product, not being mentioned is not a growth failure. Knowing which questions not to compete for lets the team reserve limited resources for the cognitive entry points that matter.

The first step in growth is not occupying more channels. It is choosing the questions worth answering over time.

02 Content is no longer only about amplifying a voice; it reduces the cost of understanding

When content is treated as fuel for traffic, volume becomes the easiest metric to manage.

How many articles went out this month? How many keywords were covered? How many platforms stayed active? These numbers show that the team acted, but not necessarily that the outside world understands the brand more easily.

While working on GEO, I kept seeing another kind of problem.

The brand knows who it is internally, while public pages all tell different versions. A product name has changed but old pages still use the former description. Sales knows the questions customers ask most often, but the website has no clear answer. An internal document is thorough, yet none of its information has been turned into a confirmed, verifiable fact approved for public expression.

AI cannot automatically access the “correct version” in the brand’s head. It can only see publicly reachable information and compress different sources into an answer.

If those sources are vague, contradictory, or stale, publishing more articles may only amplify the confusion.

I therefore look at whether content reduces the cost of one act of understanding: Can a user quickly learn what the product solves, who it is for, and who it is not for? Are capabilities, service scope, and limitations clear? Can a conclusion be checked against a public source? Do different pages keep the same fact consistent?

There is one boundary I am increasingly unwilling to cross: being in the knowledge base does not mean being ready to publish.

Internal material can help the system understand a project, but only confirmed brand facts approved for external expression belong in public content. Unsupported cases, rankings, outcome numbers, and customer testimonials cannot be filled in automatically just because an article would sound more persuasive with them.

This makes content production a little slower, and some drafts less exciting.

Seen another way, it changes brand growth from “continually manufacturing new claims” into “continually accumulating expressions the brand can stand behind.” That may not create a louder voice immediately, but it gives users, the content team, and AI the same brand to work with.

03 I now separate being found, understood, trusted, and chosen

“AI mentioned the brand” is an easy result to share.

It is worth observing. When the user did not name the brand, a natural mention at least means the brand entered the candidate set for that answer.

But the deeper the product goes, the less willing I am to write one appearance down as growth.

The brand may be one item in a long list with no explanation. It may be put in the wrong category or described as suitable for people it does not serve. It may receive a positive-sounding sentence whose capability and price are both wrong.

So I gradually separated four states:

• Found: Did the brand enter the answer to a relevant question?
• Understood: Did AI place the brand in the right category, situation, and capability boundary?
• Trusted: Can the key judgments in the answer be checked against facts and sources?
• Chosen: When the user compares options, does the brand become a candidate for a relevant reason?

There is no automatic bridge between these layers.

Being found is a start, but the brand can stop at “understood.” Getting the facts right does not mean the user is ready to choose. Visits, inquiries, and deals are also shaped by product experience, budget, relationships, and timing.

Separating these layers now feels like a more honest understanding of growth.

The team can act on the current gap: if the brand is not found, check question relevance and public-information coverage; if it is mentioned incorrectly, correct the fact first; if it is accurate but unsupported, inspect the citations; only after it enters a comparison should the team discuss differentiation and reasons to choose.

There is no universal metric, but there is a real next action.

04 The value of publishing is not only channel count; it is the relationship among sources

I used to understand distribution mainly as coverage.

Put the same article on more platforms and the brand gains more entry points. That still has value, but AI answers made me care about another question: What shared version of the brand are these sources telling?

The official site can carry stable product facts. A help center can explain usage and boundaries. Industry media may offer a third-party view. Community discussion can preserve opinions closer to actual use. Their roles are different and should not be reduced to a sum of page counts.

If every channel copies the same broad brand introduction, the number of pages grows while the information does not.

When official pages make the facts clear and external content confirms important judgments from its own perspective, the brand gradually builds a more complete set of public sources. The brand cannot control whether AI cites one particular page. It can control what it contributes, which facts are checkable, and whether important questions have suitable public answers.

This also changed what “published” means to me.

Generating a draft is not publication. A platform returning “submitted successfully” does not necessarily mean there is a reachable public link. Only the actual public version, channel, time, and link tell later monitoring what to observe.

Publishing is not the period at the end of content work. It is a confirmed action left in the external world.

From this perspective, channel growth is not simply adding accounts and article counts. It is building a set of public sources that do not contradict one another, can be checked, and genuinely answer user questions.

05 Growth is no longer a result screenshot; it is a falsifiable judgment loop

The most addictive part of GEO is a beautiful answer.

Ask a question, see the brand appear in a good position, save a screenshot, and it feels as if the whole process has produced a result.

But the answer may change when the same question is asked on another platform, account, date, or web-access state. One answer is real evidence, but it does not automatically represent a stable state.

That is why I have come to care more about baselines, collection context, and repeated observation.

The team first preserves what happened before the action, then records what it plans to do, why, and what change it expects. An article being complete or a page being public proves only that an action happened. New evidence from a comparable question and environment is needed before discussing improvement.

Even when the result changes, restraint matters.

An article is published first and mentions rise later. That is a time sequence, not necessarily a causal relationship. A model update, a third-party page, outside discussion, or collection conditions may have changed the answer.

That is why, in BeanInsight, I define a GEO opportunity as a falsifiable improvement hypothesis rather than a task that must work. An opportunity can be confirmed, deferred, or rejected. Once an action is complete, it enters “awaiting verification” instead of being declared improved. If a retest misses its stated criterion, the path should preserve a failure or unverifiable result.

This state model does not resemble a traditional growth report.

It offers another kind of value: the team does not have to explain every action as a success. An unverified judgment can still prevent the next round from taking one wrong path.

I now prefer to think of growth as learning speed, not only result speed.

06 Brand growth moves from a content-team task to a shared habit

Once I started working from questions, facts, sources, and retests, I saw that GEO could not belong only to the content team.

Sales is closest to the questions real customers ask. Product knows capability boundaries and version changes. Delivery knows which scenarios hold and which promises cannot be made. Brand and content turn that information into public expression. Operations uses publishing and monitoring to see how the outside answer understands it.

If the information stays in chats and personal experience, writers have to guess again each time, and the brand drifts with each author.

The product’s role is not to make every judgment for everyone. It is to give those judgments a shared home.

Questions enter the library. Brand materials enter the knowledge system. Publishable facts are confirmed separately. Articles retain the evidence they used. Publication records preserve external results. Monitoring returns to the same question. GEO opportunities turn observations into the next action.

It sounds like a product flow. The hard part is the habit.

Will the team confirm the question before writing, check facts before publishing, open the original answer even after a good result, and admit that an effect is still awaiting verification after a task is complete?

These actions do not directly create a beautiful growth curve. They gradually reduce the loss in brand expression: what sales knows no longer disappears before it reaches content, product updates do not remain trapped in old pages, and monitoring errors have a path back into materials and public content for correction.

I have started to believe that many brand-growth problems are not only a matter of insufficient effort. Context is lost as information moves through the team and into the outside world.

GEO made those breaks visible to me.

07 How I define brand growth today

If I had to describe brand growth in one sentence now, I would not say only “make more people see the brand.”

I would say:

Brand growth means becoming easier to find in the user questions that matter, correctly understood, supported by trustworthy information, and chosen for a relevant reason; while the team can see deviations, take action, and keep learning from new evidence.

It is a long sentence, not a good poster headline.

But it reminds me of a few things.

Growth is not seeking appearances everywhere; it is entering relevant questions. It is not making AI say uniformly positive things; it is reducing error and ambiguity first. It is not filling every channel with content; it is building public facts that reinforce one another. It is not treating task completion as improved performance; it is returning to the original question to verify. And it is not assigning every outcome to one department; it is keeping brand facts current across the team.

It also keeps one part unresolved.

GEO can help a team observe how a brand appears in AI answers, shorten the distance from a gap to an action, and make later comparison more grounded. It cannot promise that a piece of content will be cited, control how external models change, or turn one mention into business growth without visit, inquiry, and deal data.

These boundaries do not weaken GEO’s value. They put it in a more credible place.

Before building this product, I cared more about sending the brand to more places.

Now I care more about what people see when the brand arrives, why the information is understood that way, whether the team can correct it from evidence, and whether the next encounter with the same question becomes a little clearer.

“Putting the brand in the answer” therefore has another meaning.

It is not only putting the brand name into an AI response. It is making the brand a choice with clear boundaries, public support, and a reason to belong in the answer.

With Note 15, the series is close to its final note.

Next I want to return to the product itself: if GEO is not one content project or a handful of separate tools, what should it become? My current answer is a growth operating system rather than a collection of tools.

But “operating system” cannot be a larger product slogan. It has to answer which contexts should connect, which judgments remain with people, which results require external evidence, and how the product can help a team avoid ineffective work over time.