GEO in ChatGPT, Perplexity, and AI Overviews: how to reclaim visibility

Update, August 2026. Google has clarified its Search Central documentation for AI features and rejected several tactics this article previously recommended: designing content for chunks, llms.txt files, and markup created specifically for models. We have updated the relevant sections and marked them below. Google has also announced the removal of expandable FAQ results from search starting August 2026. The basis for AI source selection remains the same: the search index and content quality. Details in the llms.txt analysis.

GEO (Generative Engine Optimization): what is visibility in AI engines?

GEO, or Generative Engine Optimization, is a strategy for optimizing a company's digital presence aimed at having its content cited by generative language models: ChatGPT, Perplexity AI, Google AI Overviews, and Claude. Unlike traditional SEO, the goal is not a URL ranking position, but being the source from which AI draws knowledge when building answers for users.

A company invisible to these models is invisible to a growing group of customers who begin their buying journey with a question to AI, not with Google.

Why traditional SEO is no longer enough

For more than two decades, the visibility game followed a single rule: page 1 on Google meant market dominance. That model is slowly becoming obsolete. Analyses from early 2025 indicate that a significant portion of Google queries end with an AI Overview answer, without any link being clicked. Perplexity AI handles over 100 million queries per month, and ChatGPT has become the de facto first place where professionals look for product recommendations, service providers, and expert answers to B2B questions.

For SMB companies, this means one thing: position 3 on Google no longer guarantees reaching the customer who typed their question into ChatGPT. Presence in generative engines allows companies to recapture the attention of customers who are already bypassing the classic search engine.

This is not a technical problem to be fixed once. It is a structural change in how customers search for information, and it requires a strategic response: systematically building topical authority that AI models can cite.

How do generative search engines work? How do they differ from Google?

A classic search engine indexes pages and ranks URLs based on hundreds of signals. A generative engine works differently: it retrieves content from the web, processes it through a language model, and synthesizes an answer, often without showing a list of links at all.

The key mechanisms that determine whose content makes it into AI answers:

• Citation of topical authorities. AI prefers sources that cover a given topic repeatedly and consistently, rather than generalist pages.
• Structured data and semantic markup. Schema.org organizes the description of a page, though Google notes that structured data is not a prerequisite for appearing in AI answers. Its greatest real value is for products and local businesses.
• External citations and unlinked mentions. Perplexity and Google AI Overviews analyze how frequently a company's name appears in the context of a given topic across the entire web, not just on its own site.
• Content quality and depth. Short, superficial articles are skipped. AI selects content that defines a topic, containing data, comparisons, and answers to questions.
• E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness). Google AI Overviews inherits trust signals from classic rankings, but reinforces them with the factor of experience and authorship.

SEO and GEO are not competitors, but complementary layers of the same visibility strategy.

ChatGPT, Perplexity, and AI Overviews: how does each choose its sources?

ChatGPT and the Retrieval-Augmented Generation mechanism

ChatGPT in Search mode uses the RAG mechanism, or Retrieval-Augmented Generation. This means the model reaches into current web resources to supplement its answer with the latest data. Content must meet three conditions to enter this mechanism:

1. Indexability by Bing. ChatGPT Search relies on the Bing index. A page must be technically sound and visible to Microsoft's crawlers.
2. Answers to questions, not just product descriptions. Content that defines a term, explains "how X works" or "when to choose Y" is cited many times more often than sales pages.
3. Consistency of the site as an expert source. A single article is not enough. ChatGPT prefers domains where a topic appears multiple times, from different perspectives and levels of detail.

This logic leads to a straightforward conclusion: build topical clusters, not individual articles. Interconnected content signals to language models that a given domain is an authority in its field.

Perplexity AI: visibility confirmed by citation

Perplexity works differently from ChatGPT: it always shows the sources its answer comes from. This makes it particularly valuable from a GEO perspective, because being cited by Perplexity means direct brand exposure at the moment a B2B customer makes a decision.

Three factors that correlate most strongly with Perplexity citation:

• Content freshness. Perplexity strongly favors current materials. Regular publications (a minimum of 2-4 articles per month on core topics) increase the chance of being cited.
• Precise answers to specific questions. Questions in the format "what is the cost of X," "how long does Y take," "how do I compare A and B" generate a large volume of Perplexity queries. Content that answers these questions directly is prioritized.
• Presence in trade media and aggregators. A mention of the company in an independent trade article acts as a trust signal for the Perplexity algorithm.

Google AI Overviews: when GEO beats SEO

Google AI Overviews (formerly SGE) is the most complex GEO environment, because it combines signals from classic rankings with a generative layer. Research conducted by the Brightedge team in 2024 shows that 52% of pages appearing in AI Overviews are not in the top 10 classic results for the same query. This means GEO opens a visibility window for brands that previously lost the position war to larger players.

Key AI Overviews signals:

• Content that answers the question directly. What determines source selection is presence in the index and answer quality, not the markup itself. Google is also retiring expandable FAQ results from search starting August 2026.
• Definitional and comparative content. Sections containing term definitions, comparison tables, and lists of pros and cons are extracted by Google in the process of building an AI Overview.
• Domain authority in a niche. Google AI favors specialized domains over generalists. Narrow topical specialization is an advantage, not a limitation.

There is no better proof that the rules of visibility have changed: a niche page with good content architecture can outpace an industry leader exactly where the customer makes their decision.

How to build a GEO strategy: from audit to monitoring

Visibility in generative engines does not come from a one-time optimization. It is a continuous, methodical process that effective companies carry out in four stages.

Diagnosis: the AI visibility starting point

Before building, you measure the starting point. The diagnosis covers the current state of brand citations in ChatGPT and Perplexity, a technical audit of the site for AI model interpretability, a map of queries the brand should be answering, and an analysis of topical gaps relative to competitors. This is the stage that most often surprises: most companies do not know how they are perceived by AI, or are not perceived at all.

Content architecture and topical authority

Based on the diagnosis, topical clusters are designed: groups of related content that together build authority in a specific niche. Each cluster contains a pillar article (2,000-4,000 words), several satellite articles (800-1,500 words), and content answering questions in conversational mode (FAQ, how-to, definitions). The structure reflects how customers actually ask AI about a given domain.

Production and distribution

Content does not just go on the website, it is syndicated to trade media, LinkedIn, partner portals, and aggregators. Every external mention of the brand in the context of a given topic is a signal for AI algorithms. The production pipeline combines a strategic framework with expert editing and multichannel distribution.

Monitoring and iteration

Citation rate is measured: how often content is cited by AI. Unlinked brand mentions are monitored. Content that is not being cited is rewritten or expanded. The pipeline is alive, not a one-time effort.

GEO rewards patience and consistency, not one-off campaigns.

Practical steps: what exactly to do to appear in AI

Below are the actions that make up an effective GEO pipeline:

1. Mark up content with schema.org. FAQ schema, Article schema, Organization schema, BreadcrumbList. This is the minimum every site should have in 2026.
2. Write in definitional mode. Every article should contain a paragraph that defines the topic in a self-contained and citable way (40-60 words, without external context).
3. Answer conversational questions. "How much does X cost," "when is it worth investing in Y," "what are the differences between A and B." These formats are prioritized by RAG models.
4. Build external mentions. Guest articles, interviews, presence in industry rankings, citations in independent reviews. Every mention is a signal for AI.
5. Keep content up to date. Update dates matter. Perplexity and Google AI prefer content from the last 12 months for queries on dynamic topics.
6. Optimize for Bing. Submit your site to Bing Webmaster Tools and check indexation. Many companies ignore Bing, yet it is precisely this index that ChatGPT Search uses.

Each of these actions has limited effect on its own. Together they build an architecture that AI models learn to recognize as a credible source.

A company's own knowledge as the foundation of GEO

GEO does not operate in isolation from a company's existing content strategy. The strongest signals for AI algorithms come from brands that transform their unique expert knowledge into specific, citable content. Case studies, data from their own projects, methodologies, client results: these are materials that generic articles do not offer, and it is precisely this kind of content that AI models are increasingly prioritizing as a premium source.

Companies that collect and systematize this knowledge, then transform it into topical clusters, build something more than search visibility. They become a recognizable topical authority for AI models: an entity that exists in the model's knowledge, not just in search results.

The most valuable asset in a GEO strategy is proprietary data and opinions that no competitor can copy with a single prompt.

Results and timeline: what to realistically expect from GEO

GEO is an investment with a 6-18 month horizon. The first citations in Perplexity for niche queries often appear 8-12 weeks after launching the pipeline. Google AI Overviews requires a longer buildup of authority, typically 4-6 months of consistent publication.

A study conducted by Surfer SEO and Brightedge (2024) shows that content optimized for GEO is cited by AI 37-65% more often than content written solely for classic SEO, at the same Google ranking position.

For B2B companies, the most important metric is not site traffic, but the quality and funnel stage of queries generated by AI. A customer who arrives at a page after being cited by ChatGPT is at a later decision stage than an organic user, which translates directly into conversion.

If you want to know how your company is perceived by ChatGPT, Perplexity, and Google AI, and what to do to turn that visibility into a measurable lead pipeline, schedule a free audit at modulla. Based on the diagnosis, you will learn how to assess your starting point and plan a GEO architecture to reclaim traffic in your industry.

FAQ: the most common questions about GEO and AI visibility

What exactly is the difference between GEO and SEO?

SEO (Search Engine Optimization) optimizes a page for position in classic Google search results: the goal is a link click. GEO (Generative Engine Optimization) optimizes content for citation by generative AI models such as ChatGPT, Perplexity, or Google AI Overviews: the goal is to appear as a knowledge source in the answer generated by AI. Both strategies should run in parallel, because generative engines partly inherit trust signals from classic rankings.

How long does it take to appear in ChatGPT or Perplexity results?

Perplexity AI can begin citing content 8-12 weeks after publishing high-quality, optimized content, provided the site is properly indexed and the content answers specific conversational questions. Google AI Overviews requires a longer buildup of topical authority: the typical horizon is 4-6 months of a consistent strategy. ChatGPT Search (based on Bing) has a similar horizon to Perplexity, but requires additional optimization for the Bing index.

Can a small company compete in GEO with major brands?

Yes, and this is one of the fundamental advantages of GEO over classic SEO. Generative AI models prioritize topical authority and depth of knowledge, not budget size or link count. A company specialized in a narrow niche that consistently publishes expert content answering its customers' questions can be cited by ChatGPT and Perplexity more often than a large player with generalist content. Niche specialization is a competitive advantage in GEO, not a limitation.

Which industries will benefit most from a GEO strategy?

The greatest benefits from GEO will go to B2B service companies, where clients ask AI expert questions before making a purchasing decision: "what solution to choose for X," "how much does Y cost," "how do companies that do Z work." This applies in particular to: consulting and advisory, marketing and technology (including AI), premium e-commerce, legal and financial services, commercial real estate, and creative and production agencies. In each of these industries, B2B customers increasingly start their buying journey with a question to AI rather than Google.