In AI search, a brand should appear in the results, not just at the top of the search results

In ChatGPT, a brand may lose out to a competitor even if it has more backlinks and a higher domain authority. The reason is simple: in the new search experience, users often make their decision directly based on the model’s response, without ever visiting the company’s website.

According to data from Profound cited by the author of the study, 92.8% of conversations in ChatGPT end without a click. At the same time, AI traffic is still low: according to Ahrefs, ChatGPT accounts for about 0.19% of all link clicks, compared to 42% for Google. The paradox is that the rare visitors coming from ChatGPT are significantly more valuable: they have a conversion rate of about 7.1% and a visit value 4.4 times higher than average.

This doesn’t mean that links are no longer needed. It means they only work in situations where the answer doesn’t fully resolve the user’s task: when comparing prices, reading reviews, studying documentation, or making a purchase with a long decision-making cycle. A query like “which delivery service should I choose?” might end with a brand name. A query about the exact delivery cost, terms, and availability will require a visit to the website.

Link weight is not the same as presence in the answer

In the case described, the client had more links and donor sites than the competitor. However, twenty test queries in ChatGPT yielded not a single mention of the client. The competitor, on the other hand, had detailed editorial reviews explaining how it differed from other services.

This is an important distinction between search results and the model’s response. A directory or aggregator may provide a link to a brand but offer almost no substantive context. It is precisely this context that helps the model understand in which response—and alongside which alternatives—it is appropriate to mention the brand.

That said, it’s not wise to draw too strong a conclusion from a single client case study. It illustrates a possible mechanism but does not prove a universal ranking rule for ChatGPT. Especially since models use different sources and generate responses in different ways.

Ranking first on Google also doesn’t guarantee a mention in ChatGPT. Regular search results and generative responses have different selection mechanisms, so a website’s ranking is only an indirect indicator of brand visibility in AI systems.

What to Measure Instead of a Single Ranking Position

It’s more practical to supplement SEO with separate monitoring of model responses:

• the share of brand mentions among competitors for a predefined set of real search queries;
• the order in which the model lists companies;
• the wording and tone of the mention—neutral, positive, or limiting the choice;
• the sources the model cites or likely uses to describe the brand;
• differences between ChatGPT, Perplexity, Alisa AI, and GigaChat.

The last point is particularly important. Perplexity relies more heavily on cited materials, Alisa AI incorporates fresh Yandex search results, and other systems use their own combinations of sources. A mention in one model does not guarantee a mention in another.

The practical takeaway for content teams is also changing. Instead of mass link building, it’s sometimes wiser to prepare a few in-depth articles on platforms that actually fit the context of the models’ responses. But this isn’t a “secret way to buy a mention”: the result depends on the quality of the source, its accessibility to the model, and the specific query.

SEO is gradually gaining another metric—not just “did users click through to the site,” but also “did the model mention the brand at the moment of selection”. Clicks will remain important for complex decisions, but for brief recommendations, the main asset is an established, well-described reputation. This should be verified not based on gut feelings, but through regular test prompts and with an understanding of the limitations of such measurements.

Source: All Articles / Machine Learning / Habr