Getting Your Brand Recommended by AI Is the Wrong Goal. What Is the Right One?

AI search visibility has created a new reflex. A platform's citation rate moves up or down, and by the next day it is being treated as a strategic rule. I have seen the same reflex in the commentary around Reddit.

Companies and consultants can look at a graph like this and jump quickly to a new recommendation: be more visible on Reddit, get people talking about your brand, improve the odds that AI recommends you. We have seen this reflex many times in SEO.

I would start with a different question:

Who are we doing this work for, and who should benefit from it?

When a visibility target distorts behaviour

SEO—and now GEO—keeps producing the same cycle. A single visibility example seen on a platform or in a model response quickly becomes a prescription for everyone. The conditions, business, and context that produced the example are often left behind.

None of these conclusions is automatically true. A channel that works in one case may not fit your business. More importantly, when you reduce a channel's value to its citation rate, you can lose the reason you were there in the first place.

Reddit is a good example. If there is a community where your work or expertise is genuinely discussed, being on Reddit can make a great deal of sense. You can answer people's questions, share your experience, and listen in order to understand a problem better. That value does not require ChatGPT to recommend you 3% of the time instead of 1%.

In fact, if you join Reddit only for AI visibility, you miss the platform's real value. The community may see you not as a contributor, but as a company that arrived to distribute itself. Your answer starts serving your own site rather than the person who asked the question. A short-term visibility experiment can weaken long-term trust.

This does not mean you should stay inactive on Reddit. If your work belongs on Reddit, be on Reddit—not so AI will recommend you more often, but because the people and conversations there genuinely matter to your work.

Do not forget who you are creating content for

Content must make sense to people first and to distribution channels second. A text that does not help people make a decision, solve a problem, see a risk, or understand a subject better does not become valuable merely because a search engine can index it more easily.

I have been publishing websites for twenty-five years. Some were company sites; others were entirely content-focused. I have experienced their SEO, their conversion, the ones that shone for a while and then disappeared, and the ones that succeeded. You do not need to experience everything personally; after being part of this ecosystem for so long, you also observe what happens around you. Making technical improvements so search engines can crawl and understand a site more easily is one thing. Filling a page with unnatural content so search engines will index it more is an entirely different thing.

Good information architecture, clear headings, accessible HTML, a fast page, accurate internal links, and structured data that clearly explains what a page is about are technically meaningful choices. They also help the user. You help a search engine understand the work without damaging the reader's experience.

By contrast, adding a long FAQ section nobody needs; answering the same question repeatedly in different wording; or creating unnatural headings just to enter particular queries is something else. At that point, you are targeting a measurement system's loophole rather than technical comprehensibility.

The two behaviours can look similar from the outside: both may contain headings, schema markup, or question-and-answer formats. The difference lies in why they were added. If a user genuinely asks a question and answering it strengthens the page's purpose, it is a good content decision. If the question exists only to send a signal to a search engine, the format may be correct while the content serves the wrong purpose.

AI citations are not a success metric

Being cited by AI can be a useful outcome. It can be tracked as a signal that your content answered a question and that a model included it in its response. But a signal and an outcome are not the same thing.

A citation does not show that the user clicked the link. It does not show that the person read your content, trusted you, worked with you, or achieved a better result. Likewise, not receiving a citation does not prove that your content is worthless. The context of the answer, the sources selected by the model, the wording of the query, and the measurement tool's scope can all change the result.

I explored this distinction in my review of AI visibility tools and my analysis of Bing's Citation Share data. You can measure what happens in selected prompts or on a particular search surface. You cannot claim that these numbers alone represent real user decisions, preferences, or commercial impact.

I do not look at a dashboard by asking “How many times were we recommended?” I ask a broader set of questions:

• Did the content genuinely make someone's decision easier?
• Did people merely find the content, or did they use it?
• Did the content explain a real strength of our work or product?
• Did being present on this channel create value for us, the community, and the reader?

Sometimes the answer is yes, and visibility follows naturally. Sometimes the answer is no; even if the graph rises, the work is heading in the wrong direction.

Being natural does not mean being passive

“Be natural” does not mean never plan or measure anything. You should think about which questions you answer, which communities you join, which technical barriers you remove, and which business result you expect.

But the starting point of the plan should not be getting a model to recommend us. The starting point should be a real problem. If there is a place where people are already talking and where we can make a meaningful contribution, we should be there. If our content explains a subject, makes a decision easier, or honestly conveys an experience, we should publish it. If a major newspaper takes interest, that can be a welcome distribution and trust outcome; we should not shape the text solely for that possibility.

This approach can look slower. You do not fill every channel at once, turn every metric into a target, or apply every new “optimise for AI” recommendation. In return, you can see more clearly what you are producing, who you are producing it for, and why you are present there.

The same question appeared in my content-pruning work: a post does not need to keep living simply because it remains on a site. In my content-pruning review, I looked at whether a page answered a real search intent or provided value to the reader. The same standard should apply to new content. Producing a new page merely because it might become visible is no different from keeping an old page merely because it is still indexed.

Benefit first, distribution second

The real question is: Does our work create value on that platform in a natural way?

If the answer is yes, be there. If the answer is no, do not force yourself onto the platform so AI will recommend you more often. The same principle applies to SEO, GEO, press visibility, and whatever distribution channel appears next.

Help search engines understand your work. Create content people can use. Contribute in the communities where you genuinely belong. Do these things not for a visibility score, but because they are the natural consequence of doing the work well.

Good work is sometimes found, sometimes recommended, and sometimes accumulates quietly for a long time. Its value is not determined only by who recommends it.

Further Reading

• CXL's Promptwatch comparison A short-term prompt comparison that shows how quickly citation measurements can move. It is a signal to inspect, not a rule for channel strategy.
• Tinuiti's AI-citation analysis A look at citation share across platforms and categories; useful for seeing why period and measurement scope matter.
• Search Engine Land's assessment A useful account of how the same data can support different strategic conclusions, including the distinction between Reddit discussions and brand profiles.