Being Cited Is Not Enough: How to Get ChatGPT, Gemini, and Perplexity to Recommend Your Brand

You can appear in ChatGPT and still be almost invisible.

Your website can serve as a source for an answer, receive a link, and even appear repeatedly in citations, while your company name never appears in the text. In other cases, the opposite happens: the model mentions or recommends a brand without linking to its website.

This distinction is becoming one of the most important issues in AI Search Optimization. In the world of ChatGPT, Gemini, Perplexity, and Google AI Mode, asking "Do we appear?" is no longer enough. You need to ask how you appear.

Is your website cited? Is your brand mentioned? Does the system understand who you are? Most importantly, does it recommend you when a user looks for a product, service, or company in your field?

Four different levels of AI visibility

Discussions of AI visibility sometimes combine several different phenomena into a single metric. In practice, research is beginning to show that there are four distinct situations.

Citation occurs when an AI system uses a particular page as a source and includes a link or reference to it.

Brand mention occurs when the company or product name appears within the answer itself, even without a link.

Entity recognition is the stage at which the system understands the brand as a defined entity: a specific company with a field of activity, products, people, an address, and connections to other sources online.

Recommendation is a much more commercially significant result. A user might ask "Which CRM system is suitable for a small business?" or "Which SEO company is recommended?", and the brand appears as one of the options the AI suggests.

For a business, these are entirely different outcomes.

A source listed at the bottom of an answer may receive some exposure. A brand the system itself describes as "a good option for..." occupies a very different position in the customer journey.

Ghost citations: the website appears, the brand disappears

A Semrush study with Kevin Indig examined thousands of domain appearances across different AI systems and found a large gap between citations and brand mentions.

According to the study, a large share of cited sources received no mention of their brand name within the answer. Only a relatively small share of appearances included both a citation and a brand mention.

This phenomenon has been called a ghost citation: the AI used information from the website, but the user barely encountered the identity of the company behind it.

Other studies reached the same broad conclusion, although the percentages vary considerably by platform, query type, and measurement method.

The exact figure therefore matters less than the phenomenon itself: a citation and a brand mention are not the same thing.

ChatGPT and Gemini do not behave the same way

One of the most interesting findings is that the gap is not the same on every platform.

In Semrush's dataset, ChatGPT tended to cite sources at a very high rate, but mentioned brand names at a much lower rate.

Gemini showed almost the opposite pattern: far more brand mentions relative to the number of citations.

It is important not to turn those percentages into a fixed rule. Other studies, based on different queries and samples, produce different numbers. But the strategic conclusion remains:

You cannot treat ChatGPT, Gemini, Perplexity, and Google AI Mode as a single engine.

A brand can perform very well for citations on one platform and poorly for mentions and recommendations on another.

This is also why AI optimization cannot be measured by testing a single prompt in ChatGPT once.

What makes AI recommend a brand?

Here, the research is less conclusive.

Many GEO companies now present "ChatGPT ranking factors," but OpenAI, Google, and Perplexity do not publish a formula explaining why one brand makes a recommendation list and another does not.

There are nevertheless some interesting signals.

One of the most prominent is the brand's presence in external sources.

Correlational studies examined as part of the research found a stronger relationship between the volume of brand mentions across the internet and AI visibility than between AI visibility and the number of backlinks.

This does not mean backlinks have stopped mattering for SEO, nor does it prove that an unlinked mention directly causes ChatGPT to recommend a company.

It does, however, point to an interesting shift: in the world of answer engines, what the web says about a brand may matter at least as much as what the brand says about itself.

Digital PR takes on a new meaning

In traditional SEO, an unlinked company mention was sometimes seen as a missed link opportunity.

In AI search, the picture may be different.

If a brand appears repeatedly in comparison articles, specialist coverage, reviews, forums, guides, industry websites, and "best of" lists, models encounter recurring connections between the brand name and its field.

For example, if a company is mentioned in dozens of independent sources in the context of "accounting software for small businesses," a semantic connection develops online between the entity and the category.

Digital PR, reviews, professional coverage, and brand mentions can therefore move from a small part of off-page work to a central part of AI optimization.

Still, it is important not to overinterpret the findings. Existing studies mainly show correlation, not proof that a particular mention directly causes a recommendation.

What about the content on your own website?

The evidence here is stronger.

The GEO study by researchers from Princeton and other institutions, presented at KDD 2024, tested content changes and their impact on visibility in generative engines under controlled conditions.

Techniques that showed improvement included:

• Incorporating data and statistics
• Quoting experts and sources
• Referencing credible sources
• Improving clarity and readability
• Adding original information rather than simply rewriting material already online

Keyword stuffing, by contrast, did not help and even harmed some metrics.

Precision matters: the study mainly examined the likelihood of citation and prominence within an answer, not whether the system would choose to recommend the brand.

Good content, in other words, can help make a website a better source. It does not necessarily turn the company into a recommended brand automatically.

Entity optimization: does AI understand who you are?

Before a system can recommend a brand, it needs to understand it as a clear entity.

This is where entity optimization comes in.

A company name written in different ways, conflicting information about its activities, outdated profiles, or inconsistencies between its website and external sources can make it harder for search systems to build a clear picture.

It therefore helps to keep details consistent, including:

The company name, field of activity, address, founders, products, official profiles, and social media pages.

Organization structured data and sameAs can help search engines understand these connections.

But there is a particularly important point here.

Schema is not a "promote me in AI" button

One recurring claim in GEO is that adding schema will make a website appear more often in AI.

At present, there is not enough evidence to support that claim.

Google explicitly states in its documentation on AI features that no special schema, AI-specific file, or new markup is required to be eligible to appear in AI Overviews and AI Mode.

Some studies found a positive correlation between structured data and visibility in AI systems, but another controlled experiment included in the research found no significant increase in citations after JSON-LD was added.

The reasonable conclusion for now is that schema still matters.

Primarily, however, it is a way to help systems understand the website and entity in an organized manner, not a proven ranking factor or a guarantee of a ChatGPT citation.

Instead of asking "Are we in ChatGPT?", measure four things

The move to AI search also requires a change in measurement.

Instead of occasionally asking ChatGPT a question to see whether your company appears, you can build a fixed set of real queries potential customers might ask.

For example:

"Which X system is recommended for a small business?"

"What are the best companies in Y?"

"What solutions are available for Z?"

You can then track the following separately:

Citation rate: How many answers list your website as a source.

Brand mention rate: How many answers include your brand name in the text.

Recommendation rate: How many answers actually present your brand as a recommended option.

Share of voice: How often you appear compared with competitors across the same query set.

Run the checks on several platforms, not just ChatGPT.

Tools such as Ahrefs Brand Radar, Profound, and other AI visibility platforms are already trying to turn these metrics into structured monitoring systems, but even a simple spreadsheet with 20 to 50 fixed prompts can be a good starting point.

How do you improve your chances of being recommended?

Based on existing research, the work can be divided into three layers.

First, make sure your website is accessible to AI search systems at all. Check robots.txt, indexing, readable HTML, and whether important content depends entirely on JavaScript.

Second, make your website a source worth citing. Add original information, data, statistics, sources, expert input, and clear answers to real user questions.

Third, and probably the newest part, build your brand's presence beyond your website.

The goal is not to generate hundreds of artificial mentions. In fact, Google already warns against attempts to create inauthentic mentions solely for AI.

The goal is for the brand to appear naturally where people and professional sources discuss the category: reviews, comparisons, articles, communities, research, guides, and professional coverage.

SEO is not disappearing, but the unit of measurement is changing

In SEO, we have become used to thinking in terms of rankings, clicks, and links.

In AI search, the journey is more complex.

A brand can be a source without being mentioned. It can be mentioned without a link. It can be well known to the model and still fail to make the recommendation list for a particular question.

The new question is therefore not simply:

"Does ChatGPT know us?"

But:

"In what contexts does it know us, in which sources does it encounter us, and what does it say about us when a customer asks for a recommendation?"

That is the difference between optimizing for citations and building real visibility in AI engines.

In a world where more searches end with an answer and no website visit, that difference may matter much more than moving up one more place in search results.