Responses generated by language models are establishing themselves as new reputation territory for brands. Traditional media, search engines and social networks are now joined by a fourth space in which companies are described, evaluated, recommended or ignored before the user even visits a corporate website.

In this new environment, the media retain a particularly high capacity for influence. Its presence can multiply the mentions of a brand by 1.78 in the responses of artificial intelligence models, according to an analysis prepared by Good Rebels and Asesores based on four sector studies on banking, insurance, hotels and telecommunications.

The research has examined more than 40,000 responses generated by ChatGPT, Gemini, Perplexity and AI Overview from more than 10,000 instructions per model. Its objective has been to identify what type of sources are involved in the construction of the responses and to what extent they contribute to a brand appearing associated with certain attributes.

The analysis thus introduces the concept of algorithmic reputation: the image of a company that artificial intelligence systems build from the sources they find, interpret and combine.

This is a less visible space than the previous ones and in which communication teams do not yet have metrics as standardized as those used in search engines, media or social networks. “Brands have been measuring their reputation in the media, search engines and networks for years. Now there is a fourth territory that is opaque, that does not provide keywords or volumes and that is already making decisions about its brand before the user visits any website”explained Iván Hernández, Partner and Director of Search, SEO, GEO and Media at Good Rebels. “What we have done with this analysis is to turn that opacity into actionable data: know which sources build a brand’s narrative in AI, with what volume, in what models and under what attributes. And, most importantly, what can be done to influence it,” has added.

Models do not use fonts in the same way

One of the main conclusions of the study is that there is no single valid strategy for all platforms. Each model gives different weight to corporate websites, comparators, editorial media and other digital sources.

• Perplexity is the system that cites corporate pages and comparators with the greatest intensity. These sources appear in more than 78% of their responses, according to data provided by Good Rebels and Asesores.
• AI Overview maintains a logic closer to traditional organic positioning and the criteria that have historically conditioned visibility in search engines. Gemini, for its part, is more selective with editorial sources.
• ChatGPT barely offers direct web addresses compared to other tools, but it responds especially sensitively to earned media presence. 24.7% of their responses included at least one reference to the media.

These differences prevent the application of a GEO strategy – the optimization of presence in generative engines – in a homogeneous way. Planning must consider the usage rate of each tool, the audience it reaches and the type of sources with which it constructs its responses.
The consequence is that work on generative visibility cannot be developed in isolation from SEO or corporate communication. You must coordinate your own content, editorial presence, activity on social networks, paid media, comparators and other authority spaces.

Between six and eight sources to exceed 60% of mentions in banking

The research also identifies a diversity threshold beyond which the probability of a brand appearing in the responses increases significantly.
In the banking sector, having between six and eight different sources cited raises the mention rate to a range between 61% and 68%. In telecommunications, the interval considered optimal is between five and seven sources, while in insurance it ranges between four and eight.

The architecture that offers the best results includes a corporate website, between three and four comparators specialized in the category and one or two editorial references in quality media. This composition reproduces, to a large extent, the traditional balance of an integrated communication strategy: own assets to establish the story, comparison platforms to provide functional information and earned media to incorporate external authority.

The problem for many organizations would not, therefore, be a complete absence of digital presence, but rather poor distribution. Brands may be present on irrelevant channels, with poorly structured content or using inconsistent expressions that make it difficult for models to correctly recognize and attribute their identity.

In the case of specialized comparators, between 70% and 90% of the responses contain brand mentions. This high presence makes them a basic condition to compete in generative visibility, but at the same time limits their ability to provide differentiation. If practically all the relevant companies in a category are present in the same comparators, appearing in them avoids being left out of the answer, but it is not enough to build an advantage over competitors.

Editorial content, for its part, plays a different role. The media can collect context, strategy, positioning, innovation, reputation or third-party evaluations that do not appear in the same depth in a corporate file or in a comparative table. According to the analysis, this narrative capacity increases the volume of brand mentions and contributes to building authority in the models.
In the case of banking, editorial presence offers a multiplier of 1.78, the highest of the four sectors studied. This means that references in the media can lead to a brand being mentioned almost twice as often as when the response is supported by other types of sources.

Media is also the category with the greatest authority in insurance, where it reaches a multiplier of 1.64. In the hotel sector, the impact drops to 1.38, although editorial content continues to be among the three main sources involved in the construction of the responses. And in telecommunications, the multiplier is 1.05. Even though it is the lowest of the sectors analyzed, references in the media have the same weight as corporate websites and comparators: 97% of the responses include a brand mention when the quote comes from any of these three categories.

The bank grants the greatest authority to editorial content

The weight of the media in banking responds to the need to interpret products, conditions, solvency, service, innovation and reputation within a sector in which trust plays a central role. Thus, the models are not limited to locating rates or features. When they answer questions about recommended entities, quality of service, digital capacity or positioning, they turn to sources that can provide context and external validation.

Corporate information allows an entity to define its own attributes, but the appearance of those same values ​​in the media offers attribution made by third parties. This distance from the original sender helps to reinforce the authority that the model grants to the content.
The research thus indicates an important change for public relations. Editorial impact no longer influences only the people who read a news story, but also the systems that will use that information to respond to future queries.

“The PR has always known that credibility is not bought: it is built, and it is built through third parties who attribute it”said Carlos Hergueta, Partner and Director of Business Development of Advisors. “What changes now is that this credibility must also be built to be read by machines and combined with other sources. AI models do not read statements: they read reviews, cite media and learn from what others say about a brand. That is exactly what PR has been able to do for decades”he explained.

Integrated communication recovers strategic value

The analysis questions the possibility of addressing algorithmic reputation as a technical discipline separate from the rest of communication. Generative optimization depends on elements from public relations, SEO, content, paid media, social networks and asset management.

The website is still necessary to establish official and structured information. Comparators provide functional data and allow entry into evaluation processes. The media provide external legitimacy and the ability to establish attributes. Networks and other channels contribute to expanding the conversation and the recurrence of these messages.

So the novelty lies not so much in the appearance of new sources as in the fact that all of them now feed a system that synthesizes and responds in its own name. This forces companies to think less about isolated impacts and more about a signaling architecture. A news story, a product sheet, a review or a social publication can fulfill different functions, but their value increases when they contribute to the same brand interpretation.

The research concludes that algorithmic reputation does not replace previous territories. It integrates and reorders them. The media, search engines and networks are still relevant, but now they also function as raw material for artificial intelligence responses.