New Research Reveals Why Only 11% of Brands Break Into AI Answers

When the platforms were asked directly to describe a company, they accurately recognised 96% of the brands tested. This suggests that most established businesses already leave enough information across their websites and the wider web for AI systems to form a broadly correct understanding of who they are and what they do.

This is an important distinction because many businesses assess their AI visibility using only branded questions. They ask ChatGPT to describe the company, receive a correct summary and conclude that their visibility is strong. The study shows why that test can produce false confidence. Recognition confirms that the system can identify a brand when prompted, but it does not show whether the brand will be introduced to someone who does not already know its name.

Despite the high recognition rate, 89% of the brands never appeared in the category-level AI answers measured by the researchers. In other words, only around 11% broke into at least one of the answers customers could encounter while researching a type of product, service or provider.

This reveals a large gap between understanding and selection. An AI platform may hold an accurate description of a company but still exclude it when deciding which brands are relevant enough to mention. For businesses, the commercial value lies much closer to selection. Customers rarely discover a new supplier by asking an AI platform to describe a company they have never heard of.

A branded prompt gives the AI system the answer within the question. When a user asks, “What does this company do?”, the platform only needs to retrieve and summarise information attached to that name. The test can identify inaccuracies, outdated details and recognition problems, but it cannot demonstrate whether the company is competitive within its market.

A stronger assessment uses unbranded customer questions. These might ask for suitable providers, comparisons between approaches, local recommendations or the best option for a particular need. The answers show whether the business is visible before the customer has formed a shortlist, which competitors are selected and what evidence appears to influence the recommendation.

The study found that referring domains and third-party web mentions were the two signals most strongly associated with AI mentions. Third-party mentions included indexed pages that named the brand outside its own domain, while referring domains showed the breadth of websites linking back to it. Higher-quality referring domains added some additional relationship, but broad third-party visibility appeared to be the stronger signal.

These are relationships rather than proven ranking factors. The study does not demonstrate that gaining a certain number of links or mentions will cause an AI platform to recommend a company. Strong businesses may naturally attract more coverage because they are already prominent. Nevertheless, the findings support the idea that AI systems look beyond what a business says about itself and seek evidence from the wider web.

The researchers analysed 49,391 citations across 5,830 AI-generated answers. Of those citations, 99.99% pointed to third-party websites rather than the domains belonging to the brands being discussed. Only four of the 150 brands included in this part of the analysis received at least one citation to their own website.

The pattern appeared across Google AI Mode, ChatGPT, Gemini, Perplexity, Google AI Overviews and Copilot. This does not mean company websites are unimportant. First-party pages still provide essential information that can help systems understand a brand. However, the visible citation attached to an answer may be a directory, publisher, comparison page, review platform or other external source instead.

A company can describe itself as trusted, innovative or market-leading, but those claims become more persuasive when credible external sources support them. Reviews, awards, industry memberships, expert commentary, case studies, interviews and editorial coverage can create a more complete evidence trail around the organisation.

The aim should not be to generate artificial mentions across low-quality websites. AI visibility depends on whether a brand appears in sources that are relevant and credible within its particular market. A legal firm may be influenced by different sources from a software company, healthcare provider or retailer. The study found that citation patterns varied by industry, reinforcing the need for a sector-specific approach.

Victorious also found that AI systems used different types of sources at different stages of the buyer journey. Early problem-awareness questions generated a different citation ecosystem from category-research prompts, and the likelihood of brands being named changed accordingly. Someone asking how to recognise a problem may receive an educational answer supported by articles, professional guidance or public information.

Someone asking which provider can solve that problem may receive names drawn from directories, reviews, comparisons or other commercial sources. Businesses therefore need visibility across the full journey rather than concentrating solely on direct recommendation prompts.

The low number of own-domain citations might make businesses question whether publishing on their own websites is worthwhile. That would be the wrong conclusion. Educational content can help define the topics, problems and areas of expertise associated with a brand. It also creates material that journalists, partners and industry websites can reference, extending the company’s footprint beyond its own domain.

The key is to publish content that adds something useful. Original research, expert explanations, customer evidence, clear frameworks and specialist insights are more likely to earn attention than generic articles repeating common advice. First-party content becomes especially valuable when it generates further discussion, links and mentions elsewhere.

Businesses should begin by identifying the external sources already appearing in AI answers across their most important customer prompts. These sources may include national publications, trade websites, local directories, specialist databases, review platforms, professional bodies or comparison pages. The most influential mix will differ according to the market and the type of question.

Once those sources are understood, businesses can assess where their information is absent, inaccurate or weaker than competitors. The response might involve strengthening directory profiles, collecting credible reviews, contributing expert commentary, developing partnerships or producing research that relevant publishers have a reason to reference. The objective is evidence and relevance, not mention volume for its own sake.

The research included eight AI experiences because a company’s visibility can vary significantly between them. Each platform has different retrieval systems, data access, citation behaviour and response styles. A business may be mentioned frequently by one model, recognised but rarely selected by another and represented inaccurately somewhere else.

Testing only ChatGPT or only Google AI Overviews can therefore produce an incomplete picture. Businesses should identify the platforms their customers are most likely to use and monitor a consistent group of commercially relevant prompts across them. This makes it possible to compare visibility, competitors, citations and changes over time rather than relying on occasional manual searches.

Recognition remains an important foundation. An AI platform cannot confidently recommend a business it misunderstands or cannot identify. However, once basic accuracy has been established, the more valuable question is how often the company is selected across the searches that influence customer decisions.

Businesses should track their share of mentions, prominence within answers, recommendation context and the sources supporting each response. They should also separate citation visibility from brand visibility. A company’s article might be cited without the company being recommended, while the company might be recommended using a citation to an unrelated third-party source.

The study does not provide a universal formula for earning AI recommendations. It examines a defined group of brands and identifies relationships within that dataset. The results should therefore guide further testing rather than being treated as a list of confirmed platform ranking factors.

Its central lesson is still significant. A business cannot judge AI visibility by asking a model whether it knows the company. The real test is whether the business appears when a potential customer asks for help, compares options or requests a recommendation. Achieving that visibility requires clear first-party information combined with credible evidence across the wider web.

Most of the brands in the study had already passed the recognition test. AI systems could describe them accurately, yet almost nine in ten remained absent when the conversation moved to category-level discovery. That is the gap businesses now need to understand.

The next stage of search visibility is not simply about making a company identifiable. It is about giving AI systems enough relevant, consistent and externally supported evidence to select that company when it matters. Businesses that measure this distinction will be better placed to understand why competitors appear and where their own visibility strategy needs to improve.