From natural SEO to GEO: making your brand visible in AI search engines

SEO is no longer enough: how to make your business visible in ChatGPT, Gemini, and AI engines

Your business can rank first on Google and still never be cited by ChatGPT

This is not unusual. It simply shows that information search no longer relies solely on a list of links ranked by a search engine

A prospect can now ask ChatGPT which agency to choose for migrating a no-code app. They can query Gemini about the best solutions for building a B2B SaaS. They can use Perplexity to compare Bubble, WeWeb, and Next.js, or ask Claude how to connect an AI to their company’s internal data

In each case, the user expects a direct answer. They don’t always want to visit ten sites before forming an opinion

Natural search remains essential, but it must now be supplemented with a new approach: GEO SEO, meaning an agency’s ability to be understood, remembered, cited, and recommended by AI-based answer engines

SEO is not disappearing, but search is evolving

For a long time, the search journey followed a simple logic:

Question, Google results, click, site visit

This journey still exists. Google remains a central channel for discovering a business, comparing solutions, and visiting a site. SEO fundamentals also continue to influence Google’s generative features, such as AI Overviews and AI Mode. Google states that these experiences rely on its index, ranking systems, and document retrieval mechanisms

But a second journey is gaining importance:

Question, generated answer, display of a few sources or brands, optional click

ChatGPT can include citations and a sources panel with its answers. Perplexity displays numbered references to the pages used. Claude can search the web, process multiple sources, and include direct citations. Gemini can also show sources or links related to certain answers

The challenge is no longer just about securing a spot on a results page. It’s also about increasing the chances of appearing in the answer itself

This doesn’t make SEO obsolete. It simply expands the field where digital visibility is determined

What is GEO?

GEO stands for Generative Engine Optimization, or optimization for generative engines

The term refers to the methods used to improve the visibility of a business, brand, or content in the responses generated by artificial intelligence systems

In practice, GEO aims to help AI engines:

• understand the company’s business;
• identify its areas of expertise;
• verify published information;
• find relevant pages;
• use certain content as sources;
• cite the brand;
• recommend the company when the question calls for it.

The concept was notably formalized in an academic paper on visibility optimization in generative engines. The authors describe these engines as systems that retrieve multiple sources before synthesizing an answer. They also highlight their opaque and evolving nature.

This paper deserves precise citation because it coined the term and quantifies what GEO can deliver. It is GEO: Generative Engine Optimization, by Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan and Deshpande. The authors introduce GEO-bench, a query evaluation dataset spanning multiple domains, and measure that the proposed methods increase visibility in generative responses "by up to 40 percent." Two caveats accompany this figure and are rarely mentioned: effectiveness varies significantly by domain, which prevents treating it as a universal promise, and the authors explicitly describe these engines as black boxes over which creators have "no control over when and how" their content is displayed. Publication: the foundational GEO research paper.

GEO is therefore not about adding the word "AI" to every title or repeating a phrase twenty times on a page.

Good AI optimization relies primarily on clarity, precision, consistency, expertise and technical accessibility of information.

GEO vs. SEO: what are the differences between the two approaches?

SEO and GEO share a common foundation but pursue slightly different goals.

The goal of SEO is to improve a page’s ranking in search results and then generate a qualified click.

The goal of GEO is to increase the likelihood that a brand or page is retained, cited or recommended in a generated response.

The resulting visibility also differs. In SEO, users typically see a title, URL and description. In a response engine, information from the site may be rephrased, combined with other sources or used to justify a recommendation.

The metrics also differ. Traditional SEO relies on rankings, impressions, click-through rate, traffic and conversions. AI search adds other measures: citation frequency, brand presence, platforms involved, associated sentiment, sources used and recommendations received.

Brand visibility in AI engines depends less on the company’s website alone. These systems can consult media, comparison sites, directories, forums, reviews, or industry publications.

SEO and AI should not be seen as opposing forces. A well-indexed, fast, and authoritative page already provides a useful foundation. The most effective GEO strategies typically build on strong SEO fundamentals.

Why traditional SEO is no longer enough on its own

Answers can be obtained without visiting a website

Answer engines reduce the number of steps needed to get information

A complex question that once required multiple searches can now receive an initial synthesis in seconds. Google explains that its generative features can run multiple related searches for a single query, then gather the information found. This method is called "query fan-out".

The term comes from Google’s documentation, which describes the technique as "issuing multiple related searches across subtopics and varied data sources" to build the answer. The same document adds a key nuance often missing from discussions on the topic: AI Overviews "are only shown when our systems determine they provide added value over traditional search and, as a result, often do not trigger." Many queries never produce them, and building an entire strategy on the assumption they will is to overestimate the phenomenon. Reference: Google documentation on AI features in search.

The click does not disappear, but it sometimes becomes secondary. Users may already have a definition, comparison, or initial shortlist of providers before visiting a site.

A strong Google ranking does not guarantee a citation

Generative systems do not automatically pull from the top ten results for the exact query wording.

A study published by Ahrefs in August 2025 compared 15,000 questions asked to multiple assistants with Google results. On average, only 12% of URLs cited by ChatGPT, Gemini, and Copilot appeared in the top ten Google results for the same query. This figure describes a specific sample and is not a universal rule, but it confirms that the two forms of visibility do not fully overlap.

Another Ahrefs analysis, published in March 2026 based on four million URLs cited in AI Overviews, found that about 38% of cited pages also appeared in the top ten result blocks for the studied query. Again, the methodology and scope must be considered.

A strong ranking remains an advantage. It is simply no longer a sufficient guarantee for inclusion in a generative search.

Brand presence across the web matters more

An AI engine may encounter a company on its official website, but also in an external article, a professional directory, a study, a video, a client testimonial, or a specialized discussion.

A brand described consistently across all sources is easier to understand. Conversely, a vague or contradictory positioning creates uncertainty.

An agency that presents itself alternately as a no-code agency, design studio, consulting firm, and general development company risks sending a less clear signal than a business with consistent expertise, case studies, and terminology.

Searches are becoming conversational

Users no longer limit themselves to short queries like "Bubble agency Paris" or "SaaS development".

They may ask:

Which agency can take over an application built with Bubble and migrate it to Next.js?

How can I connect an AI to a small business’s internal data without exposing confidential documents?

Should I choose Make or n8n to automate critical workflows across multiple business systems?

These questions combine a need, context, and constraints. Generic content has little chance of answering them properly

How do AI engines select their sources

There is no guaranteed acceptance testing tobe cited by ChatGPTorbe recommended by Gemini.

The algorithms are proprietary. They evolve regularly. Triggering a search, formulating intermediate queries, selecting pages, and generating the final response can vary by platform, country, conversation context, and date

However, some facts are established

Google states that its generative features rely on its index and can perform multiple related searches. OpenAI specifies that a public site can appear in ChatGPT Search, provided its search bot can access content intended for summaries and citations. Anthropic and Perplexity also document dedicated search bots

Based on available documentation and observations of these platforms, several signals are worth addressing:

• precise alignment with the question
• clarity of the answer
• consistency of the information
• demonstrated expertise
• presence of verifiable evidence
• quality of cited sources
• freshness of the information
• brand reputation
• mentions on third-party sites
• technical accessibility of the page

Recent academic work also suggests distinguishing between the selection of a citation and its actual influence on the generated response. A page may be cited without its content playing a major role in the final synthesis. Measuring GEO remains more complex than a simple link count

How to optimize your site for GEO

Answer your prospects’ real questions

Content optimized for AI starts with a deep understanding of commercial inquiries

It’s not enough to target “web development.” You need to address questions closer to real decision-making:

• How do you take over an application built with an AI generation tool?
• When should you migrate a no-code application to code?
• What architecture should you choose for a B2B SaaS?
• How do you secure an assistant connected to internal data?
• What are the hidden costs of a critical automation?

Each page should provide an identifiable, well-reasoned answer tailored to a specific situation.

Structure information without artificially splitting content

Titles should clearly state what the reader will learn. The first sentences of a section should provide a comprehensible answer without requiring the reader to finish the page.

Lists, short definitions, step-by-step guides and self-contained paragraphs are useful when they genuinely improve readability.

However, you shouldn’t turn every page into a series of fragments designed solely for bots. Google states that no specific “block-based” structure is required for its generative features. The priority remains useful content for people.

Publish hard-to-replicate expertise

AI engines can already summarize general information found on hundreds of sites.

To stand out, a company must provide what generic content lacks:

• explained technical choices;
• lessons learned;
• observed limitations of a tool;
• data from real projects;
• argued comparisons;
• internal methods;
• detailed case studies;
• assumed professional opinions.

Google itself recommends creating useful, original content based on firsthand experience rather than mere summaries of existing information.

Develop high-intent content

The most strategic formats are often those that support a decision.

A technical company can publish comparisons, migration guides, “alternative to” pages, case studies, or “how to choose” content

Scroll, for example, has an article explaining the difference between a web app and a website. This type of topic answers a clear question while helping the reader qualify their project

Strengthen brand consistency

The company’s name, positioning, services, and proof points must remain consistent across the website, social profiles, directories, and external publications

“About” pages, expert bios, location pages, service pages, and case studies must use compatible information

The goal is not to repeat the same text everywhere. It’s about making the company’s identity understandable and verifiable

Secure mentions on third-party sources

GEO isn’t just about optimizing your own domain

Press relations, partnerships, expert contributions, podcasts, joint studies, reviews, and testimonials all contribute to overall visibility

These mentions must remain legitimate. Google advises against tactics that artificially create signals or citations solely to influence generative results

Work on technical fundamentals

An inaccessible page cannot become a reliable source

Key aspects to check include

• indexability
• robots.txt rules
• CDN or firewall protections
• server-side rendering when the site relies heavily on JavaScript
• performance
• internal linking
• availability of important content in text form
• relevant structured data
• consistency between structured data and visible content
• accessibility of menus, buttons, and forms

OpenAI recommends not blocking OAI-SearchBot for pages you want to appear in ChatGPT summaries and citations. Anthropic states that blocking Claude-SearchBot may reduce a site’s visibility in its search results. Perplexity also recommends allowing PerplexityBot to appear in its index.

Structured data can help engines understand certain information and enable rich results. However, it does not guarantee any citation. Google specifies that there is no special Schema.org markup for AI Overviews or AI Mode.

The same caution applies to the llms.txt file. It may be considered in some contexts, but it is not a universal shortcut to AI search visibility. Google explicitly states it does not use it for its search features. Scroll details the role and limitations of the llms.txt file.

Which content is most likely to be used by AI?

Response engines often look for passages that can support a claim or directly answer a question.

The most useful formats include clear definitions, argued comparisons, sourced statistics, detailed guides, procedures, original studies, experience reports, case studies and regularly updated expertise pages.

For Scroll, several topics show strong potential:

• Bubble or Next.js: how to choose?
• How to take over an application generated with Lovable?
• Make or n8n for critical workflows?
• How to migrate a no-code application to code?
• How to choose the technical architecture for a SaaS?
• How to connect an AI to a company’s internal data?

The latter topic allows creating links with content dedicated to RAG and its enterprise use cases, then with a more technical guide on LangChain and RAG architectures.

The goal is not to publish the highest number of pages. It is to cover topics where the company has real expertise and can provide better answers than existing content.

How to measure GEO performance?

GEO measurement remains less standardized than SEO.

A company can track:

• brand presence in responses;
• number and position of citations;
• platforms where it appears;
• pages used as sources;
• sentiment associated with the brand;
• questions for which it is recommended;
• traffic from assistants;
• brand searches;
• conversions from these visits.

Public rankings also provide an initial benchmark: Topya, for example, measures brand share of voice in ChatGPT, Claude, and Gemini responses, category by category.

OpenAI automatically adds a utm_source=chatgpt.com parameter to some recommendation links, making it easier to identify this traffic in analytics tools.

Since June 2026, Google has also been testing dedicated reports in Search Console for impressions in its generative features. The rollout still covered a subset of sites at the time of the announcement.

For other platforms, the most reliable method is to create a list of strategic questions and regularly repeat tests with different phrasings.

You need to track brand presence, but also competitors, cited sources, and how responses evolve. A single test is not enough, as results can vary.

Should you invest more in GEO than in SEO?

There is no one-size-fits-all allocation for every business.

Investment levels depend on the industry, prospect behavior, sales cycle, website maturity, and the role of recommendations in choosing a provider.

A B2B company with a long sales cycle should prioritize visibility on ChatGPT, Gemini, Claude, and Perplexity. Prospects often use multiple sources before contacting a vendor.

This does not mean abruptly reducing SEO efforts.

Natural search remains essential for exploration, indexing, understanding, and discovery. It also continues to feed some generative search systems.

For most businesses, the best approach is to evolve the existing strategy so the same content serves both traditional results and response engines.

Why start now?

GEO is not yet an exact science. That’s precisely why you should avoid opportunistic actions and focus on building lasting assets.

Clear editorial expertise, a consistent brand, structured content, verifiable evidence and a presence on trusted sources remain valuable, regardless of how any single tool evolves.

These elements take time. A credible case study, industry reputation or specialized content corpus can’t be built in a few weeks.

Starting now doesn’t mean following a trend. It means gradually adapting your strategy to new search journeys.

Being visible is no longer enough, you also need to be retained

SEO still largely determines who is visible. GEO is starting to determine who is retained, cited and recommended.

Companies don’t have to choose between the two. They need to build a coherentGEO SEO strategy, with a solid website, genuinely expert content, a clear brand presence and regular performance tracking.

This also makes the topic broader than simple editorial optimization. Visibility on ChatGPT or Gemini depends on content, but also on web development, data structuring, technical accessibility and the ability to integrate AI into a comprehensive digital strategy.

Want to know if your company is visible on Google, ChatGPT, Gemini, Claude or Perplexity? Scroll can analyze your current presence and identify the content, technical optimizations and authority signals to develop.