Why AI Recommends One Brand and Forgets Another

Ask ChatGPT what the best tool in a category is twice in a row. The list of brands may change between one answer and the next, even if nothing has changed in the real world between the two questions. This is not a system failure; it is the result of how these models decide what to include in a response.

Where the Answer Comes From

Every answer from a generative AI model combines, in varying proportions depending on the model and the question, two different sources of knowledge. The first is training: the model learned, during this process, statistical associations between brands, categories, and contexts from a massive volume of text. Brands that are frequently mentioned in clear contexts tend to be more strongly associated with their category.

The second source is real-time search, also known as grounding. Various models search the web at the moment they generate the response, rather than relying solely on what they learned during training. Each one decides this differently: Gemini analyzes the prompt and decides whether it’s worth searching on Google (Google AI for Developers), ChatGPT can trigger a search automatically or by user activation (OpenAI Help Center), Claude decides based on the content of the prompt (Anthropic), and Perplexity searches by default in practically every response (Perplexity Help Center).

Not Every Brand is Recognized the Same Way

Before mentioning a brand, the model needs to recognize that name as referring to a specific entity, not a common word or another company with the same name. Brands with generic names face an extra challenge in this recognition, even if they have a significant market presence. Structured data that explicitly declares what an organization is, such as the schema for Organization, helps compatible mechanisms resolve this ambiguity more accurately (Google Search Central).

Why the Same Question Doesn’t Always Bring the Same Answer

Generative AI models are not deterministic. A study by SparkToro in partnership with Gumshoe.ai, with 2,961 prompt executions in ChatGPT, Claude, and Google’s AI, found less than a 1 in 100 chance that two executions of the same prompt would yield exactly the same list of brands (SparkToro).

In practice, this means that a single answer shows no pattern at all. What indicates something real is observing multiple executions of the same prompt over time, not an isolated answer.

The Complete Mechanism in One Guide

These three factors—training, real-time search, and entity recognition—explain why a brand appears in one response and disappears in the next, and why two brands of the same size can have very different results. MencionAI monitors these signals, prompt by prompt, across multiple models simultaneously, to show where each brand appears and where it still does not. We also put together a comprehensive guide detailing each of these mechanisms, with the official links for each model and what can be done from that: How Do AIs Choose Which Brands to Recommend?

If you want to see how this works for your specific brand, just schedule a demo.

Pedro, CMO of MencionAI

Frequently Asked Questions

Why does the same question in AI bring different brands on different days? Because generative AI models are not deterministic. A study by SparkToro and Gumshoe.ai found less than a 1 in 100 chance that two executions of the same prompt would yield exactly the same list of brands, which is why an isolated answer shows no pattern.

Does paying for ads help my brand appear in an AI response? Not directly. The mechanisms that decide this are training and real-time search; neither involves an ad auction like in a traditional search engine. What matters is the presence and clarity of content in the sources the model uses.

Does my brand need to be well-positioned on Google to appear in AI? It helps, but it doesn’t guarantee anything on its own. It also depends on whether the model recognizes the brand name as a specific entity, which is harder when the name is generic or similar to another company’s.

Where can I find the complete explanation of these mechanisms? In the guide How Do AIs Choose Which Brands to Recommend?, which details each source of knowledge, the search behavior of each model, and the role of structured data.