How source selection works in the new search model

Table of contents

1. Introduction: a shift in the search model
2. Search is no longer a list of results
3. How AI Overviews work at a system level
4. Ranking vs. source selection
5. Why most content is not selected
6. Key factors behind source selection
7. The role of traditional SEO
8. What this means for content
9. Summary

1. Introduction: a shift in the search model

Most discussions around AI Overviews start with the question of how to “optimize” for them, but this framing is simply misleading – the real issue is not optimization for a new feature, but a deep change in how search itself works.

According to McKinsey & Company, search is evolving from a gateway to a list of links into a system that provides direct answers – this shift moves a significant part of the decision-making process before the user even visits a website.

In this context, visibility is no longer enough – what matters is whether your content becomes part of the answer.

2. Search is no longer a list of results

In the traditional model, users navigated through multiple results and built their own understanding by comparing sources. In the AI-driven model, that process is compressed, because the answer is generated directly in the search interface by combining information from multiple sources.

Documentation from Google indicates that AI-powered features build on top of existing search systems, adding a generative layer that synthesizes information.

This changes the role of a webpage which is no longer the destination, but one of the inputs used to construct the response.

3. How AI Overviews work at a system level

AI Overviews are not a separate system detached from search, they operate as an additional layer on top of the existing ranking infrastructure.

The process can be simplified into three stages:

1. The system identifies a set of candidate sources using traditional ranking signals
2. It evaluates their content in terms of usefulness for answering the query
3. It generates a response by combining relevant pieces of information

This means that ranking still matters, but it is no longer sufficient on its own.

4. Ranking vs. source selection

Ranking and source selection are fundamentally different processes.

Ranking determines which pages are most relevant for a query. Source selection determines which pieces of content can be used to construct an answer.

As a result, a page can:

• rank highly but not be used as a source
• not be in the top positions but still be included in the answer

The distinction exists because AI systems do not need entire pages. They need specific, extractable pieces of information.

5. Why most content is not selected

Most content is not designed to function as a source of information. It is designed to capture traffic.

Common issues include:

• lack of clear, direct answers
• overly long and unfocused structure
• repetition of existing information
• absence of original insight

In the traditional model, users filtered content themselves and extracted what they needed, but in the AI model, that step is handled by the system – only content that is immediately useful is considered.

6. Key factors behind source selection

Based on available documentation and observed patterns, several factors consistently influence selection:

• Completeness – content that covers a topic in a coherent and self-contained way is more likely to be used.
• Structure – information needs to be presented clearly and in a way that can be easily extracted and reused.
• Credibility – while traditional authority signals still matter, consistency and topical expertise become more important.
• Query-level relevance – the system prioritizes content that directly answers a specific question rather than broadly covering a topic.

So, together these factors define a new standard where content is evaluated not only for ranking, but for its usability as part of an answer.

7. The role of traditional SEO

SEO remains relevant, but its role changes. Of course ranking still determines which pages enter the pool of potential sources. Without visibility, inclusion is unlikely.

However, ranking alone does not guarantee selection – this positions SEO as a foundational layer rather than the end goal. On top of it sits a second layer focused on how content performs within generative systems.

8. What this means for content

The shift in search behavior requires a shift in content strategy.

Content needs to:

• provide direct and unambiguous answers,
• reduce unnecessary introduction and repetition,
• be structured around specific questions,
• include original insights, data, or experience,
• be formatted in a way that allows extraction and reuse.

In practical terms, this means moving from writing for a reader who explores content to writing for a system that selects and composes it.

9. Summary

AI Overviews do not just change how results are displayed, but they change how content is selected.

Visibility is no longer defined by position in a list, but by inclusion in an answer.

In this model, advantage shifts toward content that is not only relevant, but also usable as a source. The central question is no longer how to rank, but how to create content that systems choose to rely on.