AI Search – how it works
AI Search is a new way of searching in which the user no longer browses dozens of links, but instead asks a question and gets a ready-made answer generated by an AI system. More and more often, these are questions like who do you recommend, what should I choose, or how do I do it, and the AI creates a shortlist, a comparison, or a step-by-step guide. In practice, this means that visibility is no longer only about fighting for a position in the TOP10, but about making sure your brand appears in the answer as a recommendation, an example, or a cited source.
This is not just theory. The zero-click trend, meaning searches without a click to a website, is already very strong: in SparkToro’s 2024 study, out of 1,000 searches, only 360 in the US and 374 in the EU ended with a click to the open web.
On top of that, there are AI Overviews and AI Mode: in an analysis of a large sample of over 10 million keywords, Semrush indicated that the share of queries with AI Overviews stabilized at around 16%.
The conclusion is simple: if some users do not click, then the classic SEO equals traffic approach becomes insufficient. That is why GEO and AI Search are emerging: actions designed to increase the chances that a brand will be included in AI-generated answers.
What AI Search really changes and why some companies are losing CTR
In traditional SEO, the user saw a list of results and chose what to click. In AI Search, the answer appears on the spot, and the click becomes optional. This affects CTR, and we have specific data to prove it.
Seer Interactive, as reported by Search Engine Land, showed that for queries with AI Overviews, organic CTR dropped from 1.76% to 0.61%, about minus 61%, while paid CTR fell from 19.7% to 6.34%, about minus 68%. At the same time, brands cited in AI Overviews saw on average 35% more organic clicks and 91% more paid clicks compared with brands that were not cited. This is the key point: in the AI era, it matters less to be in the results and more to be in the sources behind the answer.
AI Search behind the scenes: where AI gets the answer from and how it chooses brands
What is often poorly understood is that AI does not invent recommendations out of thin air. Most systems work in a hybrid way:
1) First, intent understanding
The model interprets the question: does the user want a recommendation who do you recommend, a comparison what is better, an instruction how to do it, or a list of criteria how to choose.
2) Then retrieval – pulling the sources
In many AI tools, the Retrieval-Augmented Generation mechanism, or RAG, is used: before the AI writes anything, it searches for and retrieves fragments of content from documents, websites, or databases. That is why indexability, accessibility, and content structure are still fundamental.
And here is a technical detail that makes a difference: in RAG, content is usually divided into chunks. For example, in RAG documentation and implementations, you often see settings such as a chunk size of around 800 tokens with an overlap of around 400 tokens to preserve context. Poorly split content means a worse chance of matching the right answer.
In practice, implementation tools often start with simple settings such as 1,000 characters plus a 200-character overlapso that the meaning of a sentence or definition is not cut off.
3) Reranking and quality filtering – the lesser-known stage
In modern systems, results may go through an additional filtering process called reranking. In descriptions of Perplexity’s ranking mechanisms, for example, there is mention of multi-layer reranking L3 for entity-based queries such as companies, people, or concepts. That means AI does not just take the first available results, but reevaluates them and may reject the entire set if the quality or relevance is too weak. For brands, this sends a clear signal: it is not enough to be mentioned somewhere – the mention has to be meaningful, in the right context, and within content that matches the user’s intent.
4) Answer synthesis plus citations
At the end, the AI creates the answer, sometimes adding citations to the sources. If a brand has:
- a consistent description of its offer,
- content that is easy to cite, such as FAQs, definitions, and checklists,
- confirmation in trustworthy places,
then the chances of being included increase.
5) Entity resolution – AI connects the dots
AI thinks in terms of entities. A brand is an entity: name, offer, location, reviews, relationships. If the data is inconsistent, for example different NAP data or different descriptions, AI has trouble recognizing it unambiguously. This is one of the most common and at the same time most underestimated reasons why AI and Google do not include a brand in their answers.
And one more thing: search engines are changing all the time. Google states that in 2023 it ran more than 700,000 experiments, which resulted in more than 4,000 improvements in Search. That is why AI Search is not a one-time action, but an iterative process.
What AI positioning means in practice
At FunkyMedia, we do not sell the promise of position number one in ChatGPT, because AI does not work like a traditional TOP10 ranking. We do what actually increases the probability that a brand will be recommended or cited:
- we organize information on the website so it is AI Ready,
- we build topical authority,
- we strengthen brand confirmation across the web through brand mentions and citations,
- and then we measure, iterate, and close the gaps.
Our approach is based on two pillars:
AI First – first we check how AI describes the category and what the user expects, and only then do we plan the actions.
AI Ready – we prepare the website and communication in a clear and unambiguous way, so that the offer is easy to summarize, compare with others, and trust.
What the AI Search process looks like at FunkyMedia
From the outside, it looks like SEO plus content. On the inside, it is a coherent system: brand clarity plus topical authority plus confirmation across the web.
We start with an AI Search audit. We look at how AI sees your company and your category. We look for gaps: is the brand unambiguous as an entity, is the offer too generic, are there decision-closing pieces of content such as how to choose, how much it costs, or when it makes sense, are there trust-building elements such as process, standards, and case studies, is company data consistent, and is the website technically ready for indexation.
Then we make the site AI Ready. We refine service pages and the offer so they are concrete: scope, process, target audience, what we do not do, results, and answers to objections. We add the formats AI likes to cite: FAQs, definitions, checklists, comparisons, a short answer plus an extended version.
Next, we build topical authority. We create content clusters that cover the topic from A to Z: from education, through comparisons, to the buying decision. This makes the brand feel obvious within the category, because it has a complete set of answers, not just a single article.
At the same time, we work off-site through brand mentions and confirmation signals. In AI Search, authority does not end on the website. We strengthen the brand’s presence in places that build credibility: publications, profiles, mentions in thematic contexts, consistent NAP data, as well as reputation signals such as reviews. This often unlocks the visibility ceiling.
Finally, there is monitoring and iteration. We check where the brand appears, where it does not, what arguments AI uses, whether there are any distortions, and we strengthen what actually affects the answers.
What the start of cooperation looks like
In the first weeks, we organize the foundations: audit, priorities, quick fixes, brand data consistency, and the first AI Ready elements. In the second stage, we refine the offer and create content designed for answers such as FAQs, comparisons, and choice scenarios, along with the first topical clusters. In the third stage, we scale: more clusters, off-site authority, and data-driven iterations.
FAQ – the most common questions about AI Search
Can you permanently get into AI answers?
AI is not a ranking like the TOP10. What you can do is increase the probability of recommendations and citations through entity consistency, AI Ready content, topical authority, trustworthy external sources, and data-driven iteration.
Do AI Overviews reduce traffic?
For some queries, yes. That is why being cited in answers is becoming more important. The data shows CTR declines, but also an advantage for brands that are cited.
Does AI Search replace SEO?
No. It is another layer. SEO, including technical setup, indexation, content, and authority, is the foundation on which AI visibility is built.
Are brand mentions necessary?
In AI Search, they help a lot because they act as external confirmation of the brand’s existence and credibility. Without them, many companies hit a visibility ceiling.
What do you measure in AI Search?
We monitor whether the brand appears in answers to key category questions, the context and arguments used, and at the same time we track classic KPIs: traffic, inquiries and leads, and conversions.