# Generative Search: How AI Answers Change the Visibility of Companies

Author: Marin Niehues, Niehues Consulting · Published: 2026-09-24 · Updated: 2026-11-16 · AI Adoption & Operations
Source: https://www.niehuesconsulting.com/blog/generative-search-visibility/

Generative search places a summarized answer above the results list or in its place, and the companies that stay visible are those named as a source within it. For businesses, the guiding question shifts from ranking to citability. The discipline for this is called generative engine optimization, or GEO. A company that appears in the answer of an AI assistant or in Google's AI Overviews shapes a decision-maker's judgment before that person opens a website.

**Visibility in generative search comes from content that an organization substantiates and delivers consistently. It is a delivery task with clear ownership.**

## What is changing in search

Classic search delivered a ranked list and left the selection to the user. Generative search delivers a summarized answer with a small number of named sources. On Google and Bing this answer sits above the results list; in standalone answer engines such as ChatGPT Search or Perplexity it replaces the list. Users read the summary and only click through when they need detail. The answer engine thus decides in advance which companies appear in a market overview and with which statements.

This carries particular weight for B2B decisions. A budget holder looking for a consultancy or a software provider now types whole questions: "Which providers suit an initiative of this size, and what should we look for in the selection?" The answer contains a shortlist and criteria. Whoever is missing from that shortlist has to work their way back into the comparison through other channels, which costs time and trust. Whoever appears goes into the conversation with a head start, because the answer engine has already placed them.

## How far do clicks fall because of AI Overviews?

When an AI summary appears, clicks on classic search results roughly halve. An analysis by the Pew Research Center based on the browsing data of 900 adults in the US shows that users clicked on a classic result in 8 percent of visits with an AI summary, and in 15 percent without one. Links inside the summary itself accounted for only around 1 percent of visits.

**Click behavior with AI summaries**
8 percent | Share of visits with a click on a classic result when an AI summary appears
15 percent | Share of visits with a click on a classic result without an AI summary
1 percent | Share of visits with a click on a link inside the summary
18 percent | Share of the Google searches studied that produced an AI summary

**Exhibit 1: With an AI summary, clicks on classic search results roughly halve**

| Label | Value |
| --- | --- |
| Classic result, without AI summary | 15% |
| Classic result, with AI summary | 8% |
| Link in the AI summary | around 1% |

Source: Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results, 2025; browsing data of 900 adults in the US
Note: Share of visits with a click
Graphic: [SVG](https://www.niehuesconsulting.com/blog/generative-search-visibility/grafik-1.svg) · [CSV](https://www.niehuesconsulting.com/blog/generative-search-visibility/grafik-1.csv) · [JSON](https://www.niehuesconsulting.com/blog/generative-search-visibility/grafik-1.json) · Niehues Consulting, [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/deed.en). Data: Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results, 2025; the source's terms apply.

For management this means that page views alone underestimate the effect of content. A growing part of opinion-forming happens in the answer, without a visit taking place.

## How answer engines select sources

Answer engines combine a language model with retrieval from web pages and indexes. The answer is built from the retrieved texts, and individual passages are shown as sources. Four properties of a source help decide whether it is cited.

The first property is directness: a text whose first sentence answers the question can be taken over as a passage. The second is clarity. Terms are defined and statements backed by figures or criteria. The third is consistency. If the company's own website and articles say the same as directories and trade media, the model places the company with confidence. The fourth is currency: a maintained body of content with a visible date.

Research supports this logic. The paper that coined the term GEO tested text changes on around 10,000 search queries. Content enriched with statistics and quotations gained up to 40 percent in visibility in the generated answers. Simply repeating search terms barely helped.

Technology is the precondition, and content decides the selection: answer engines cite passages that answer a question directly and with evidence. The methods differ between providers and change continuously. A management model therefore starts with the source rather than with individual platform rules.

## What is the difference between SEO and GEO?

SEO optimizes for ranking in a results list, GEO for being named as a source in a generated answer. Both disciplines share the technical foundation. They differ in their objective and in measurement.

**SEO and GEO compared**
Criterion | Classic SEO | GEO | With Niehues Consulting
Objective | Ranking in the top positions | Citation as a source in the answer | Citation with a measurable effect on inquiries
Unit of optimization | Page and keyword | Passage and question | Articles whose first sentence answers the question
Success measure | Rank, clicks, sessions | Presence and accuracy of the citation | Question set in management reporting
Levers | Technology, links, keywords | Direct answers, evidence, consistency across sources | Content model with an editorial rhythm
Ownership | Marketing | Several functions without clear lead | One owner per topic and a central editorial team

Generative search therefore does not replace classic optimization. It shifts the weight from keywords and links to content quality and consistency. The term answer engine optimization, or AEO, is also used frequently. At its core it means the same as GEO.

## How do companies optimize their website for ChatGPT, Perplexity and Google AI Overviews?

Companies optimize for all three systems with the same principles: retrievable pages answer the question in the first sentence with substantiated statements, and the company appears consistently across all sources. Four stages make the task manageable for leadership. They build on one another.

- Stage 1, findability: Content is technically retrievable and cleanly structured. The answer engines' crawlers are not blocked in robots.txt, and clear titles and headings structure every page.
- Stage 2, comprehensibility: Each article answers a specific question in its first sentence and defines its terms.
- Stage 3, verifiability: Statements are backed by concrete figures and thresholds and can be checked.
- Stage 4, recognition: The company appears with the same name and the same positioning wherever a model gathers information.

Without stage 1, every further measure has no effect. Many companies invest in stage 4 through reach campaigns before stage 2 is in place. The result is attention without citation.

Structured data based on Schema.org belongs to stage 1. It helps search engines recognize the author and date of a page unambiguously. Google states no additional technical requirements for AI Overviews beyond normal indexability; structured data is therefore a basis for clarity and not a lever for citation in its own right. The llms.txt file is a proposal from the developer community that gives language models an overview of a site's key content. It is not a binding standard, and an effect on citations has not been demonstrated. It costs little and ranks behind stages 2 and 3.

**Checklist for the executive team**
- Does the first sentence of each article answer the title question?
- Are statements backed by figures, thresholds or criteria?
- Does every page describe the company and its services in the same way?
- Who owns the maintenance of content, and at what rhythm?

## From marketing task to delivery task

Generative visibility touches several functions at once. Marketing and the business units supply content, and IT is responsible for technical delivery. In many organizations it is not settled who approves content and who supplies expertise. The result is outdated pages and contradictory statements that answer engines reproduce unchanged.

A robust delivery model settles ownership first. One owner per topic provides the expertise, and a central editorial team secures the form and currency of the articles. The executive team sets the positioning. A fixed rhythm follows: articles are reviewed before publication and checked for currency every quarter.

## Which tools measure visibility in AI answers?

Measurement uses a fixed set of questions that the company regularly puts to several answer engines, manually or with specialized tools. Established SEO platforms now offer dedicated modules for this, and there are also standalone tools for monitoring AI answers. To start with, a set of 30 to 50 questions that decision-makers in your market actually ask is sufficient. Three measures show progress:

- Presence: In what share of answers is the company named?
- Position: Is it among the first mentions?
- Accuracy: Is its service and positioning described correctly?

Because answers fluctuate from run to run, the trend over several months is what counts. Answer engines with web retrieval can cite new content once it is indexed, often within a few weeks. Knowledge stored in the model itself only changes with a new training cut. Plan three to six months, therefore, before presence changes reliably.

The values feed into management reporting. How such measures fit into steering is described in [AI metrics for management reporting](/blog/ai-kpis-management-reporting/). The role of content and personalization in marketing is covered in [AI in marketing](/blog/ai-in-marketing/). How generative visibility fits into the overall management of AI is set out in the article [AI management: what executive leadership must steer](/blog/ai-management-what-executives-must-steer/).

Three misconceptions delay the effect. The first treats visibility as the result of technical tricks. The second treats the topic as a marketing project with an end date, although content and models change continuously; the task therefore needs an owner in regular operation. The third concerns measurement: whoever only counts page views misses the part of the effect that occurs without a click.

## Niehues Consulting as a partner for generative visibility

[Niehues Consulting](/) helps organizations turn this initiative into a delivery task with clear ownership, using the Niehues Delivery Model with the phases Understand, Build and Introduce. In Understand, a picture emerges of how answer engines portray the company today, guided by the question: what is the baseline, and how do we measure success? In Build, the content model and editorial process are set up, followed directly by measurement. Introduce anchors roles and processes in the business units so that the content keeps being maintained after the project. Here the question is: who owns operation from day one?

Marin Niehues steers such initiatives as Delivery Manager or Product Owner. He knows the delivery perspective from projects for VC-funded scale-ups through to DAX 40 groups, with assignments in 20+ countries on six continents. The entry point is a [free initial consultation of 30 minutes](/#kontakt), in which we clarify where your company stands in AI answers today.

**What needs deciding now**
Generative search makes visibility a question of content quality. Set a question set with baselines and name one owner per topic. Companies that settle ownership and rhythm early are named in answers.

**Sources**
- Pew Research Center: Google users are less likely to click on links when an AI summary appears in the results, 2025
- Aggarwal et al.: GEO: Generative Engine Optimization, KDD, 2024

## Frequently asked questions

### What is generative search?

Generative search is the answering of search queries by a language model that retrieves content from the web and condenses it into a coherent answer. Individual sources are named or linked. Examples are Google's AI Overviews and standalone answer engines such as Perplexity.

### What is GEO (generative engine optimization)?

GEO is the practice of aligning content so that it is named as a source in AI-generated answers. The term comes from a research paper showing that substantiated content with figures and quotations appears considerably more often. GEO complements classic search engine optimization.

### What is the difference between SEO and GEO?

SEO aims at ranking in the results list, GEO at being named in the generated answer. SEO measures rank and clicks, GEO the presence and accuracy of the citation. The technical foundation is the same.

### Do AI Overviews reduce clicks to websites?

Yes, when an AI summary appears, clicks on classic results roughly halve, from 15 to 8 percent of visits. Links inside the summary account for around 1 percent. Page views alone therefore underestimate the effect of content.

### How does a company get mentioned in AI answers?

A company is named when its content answers a question directly and with evidence and appears consistently across several sources. Clear first sentences with concrete figures and criteria are decisive. Consistent positioning on all channels is added to this.

### Does Schema.org structured data help with AI answers?

Structured data based on Schema.org makes the author and date of a page unambiguous. It is a technical foundation, but not a lever for citation in its own right. Selection is driven by the content of the passage.

### What is llms.txt, and do you need it?

An llms.txt is a text file that gives language models an overview of a website's key content. It is a proposal from the developer community and not a binding standard. An effect on citations has not been demonstrated, so it has low priority.

### Which tools track visibility in ChatGPT and other AI answers?

Established SEO platforms offer dedicated modules, and there are also specialized tools for monitoring AI answers. The basis is always a fixed set of questions put regularly to several answer engines. The trend over several months is what is evaluated.

### Does generative search replace classic search engine optimization?

No, it complements it. Technical findability remains a precondition, because answer engines need retrievable content. Added to this is aligning content with direct, substantiated answers.
