Citations over Rankings: The New Currency of AI Visibility
For twenty years, the game stayed the same. If you wanted to appear at the top of a results page, you optimised keywords, earned backlinks and counted clicks. But that era is coming to an end.
Today, when someone asks chatbots like ChatGPT, Perplexity or even just Google "Which agency should I work with?" they don't get ten blue links. They get one answer, with a few sources behind it. If your brand is not amongst the cited sources, you are simply not getting seen anymore. Gartner, Inc. the American research and advisory firm, expects traditional search volume to drop by a quarter as people shift to AI answer engines.
What is replacing the ranking?
Visibility in an AI answer engine is divided into two categories:
- Mentions. Does the model mention you at all when someone searches for your category?
- Citations: When the model provides an answer, does it draw on your content and attribute it to you?
Mentions and citations are both important, but they have different time horizons. Mentions are what people actually see today (few users actually take the time to check sources). Citations, on the other hand, have a cumulative effect: the sources that models reference today will influence what they mention tomorrow.
This is why we say that the focus has shifted from ranking to citations. Soon, being in first position on Google is going to be worth less than one clear sentence within an answer that everyone reads.
So we ran a test…
At Sergeant, we have analysed what this shift means for brands. Using a reproducible panel-judge methodology, we tested 43 brands across the major AI engines. Our sample spanned DACH market leaders and lesser-known B2B companies from Switzerland, Germany and Austria, as well as Apple as a global benchmark control. Two things stood out:
- 92% of the DACH brands produced at least one hallucination. The models made incorrect statements about them.
- Only three brands were completely accurate. Even Apple, our global control, did not pass the test: the more we verified against a company's own website, the more misinformation surfaced.
Most errors were not the obvious kind. One engine confidently stated a founding year that was off by two years; another assigned an agency to the foodtech industry. Statements a buyer has no way to catch – and no reason to doubt.
In the DACH market, AI engine usage is distributed unevenly. ChatGPT accounts for around two thirds of consumer AI responses, meaning that a single confident mistake there reaches far more buyers than a misranked keyword ever did. The most surprising source, though, was Perplexity. The engine that cites its sources was responsible for 60% of all misinformation flagged in our benchmark. Citations make answers appear trustworthy. However, they do not make them true.
The takeaway is both uncomfortable and useful:
Most brands are already being described by AI, whether they control it or not. And what the engines say about them is often wrong.
How this affects your work
Optimising for an AI answer engine means making your brand quotable and verifiable.
- Provide the clearest and most specific information on the topics you want to be associated with.
- Ensure your facts are consistent wherever a model might check them.
- Secure corroboration from the authoritative sources that models trust.
- Focus on the search engines that actually matter in your market, not just the ones you use yourself.
See where you stand
You can find out what AI is saying about your brand today. Our free Quick Check tool analyses your mentions and citations across major AI engines. In about a minute, it will show you where you are visible and invisible – and where the model is getting you wrong.
