SEO and AI: What are the new rules for visibility? A comprehensive analysis.

Introduction

AI-powered search platforms, such as Google’s AI Mode and AI Overviews, and large language models (LLMs), such as ChatGPT and Perplexity AI, are significantly changing how users search for information. This, in turn, puts pressure on search marketers, forcing them to re-evaluate traditional SEO strategies. However, an ever-growing body of research and articles makes it difficult to distinguish evidence-based information from mere opinions. This analysis summarizes the key findings to provide a clear overview of what actually works in AI-powered search.

You can view the summary infographic here.

Key findings and strategies to improve AI visibility

Studies show that while there is a correlation between AI search engine visibility and organic rankings, it is not absolute. To succeed, it is necessary to understand how AI systems work and adapt content accordingly.

1. What is the overlap between AI mode and organic rankings (Google’s AI Mode)

What is “AI Mode”?

AI Mode is a Google search tool that generates a short and synthesized response directly on the search results page. It differs from traditional search in that it doesn’t just provide links, but tries to answer the query by creating a new, consolidated response compiled from multiple sources.

According to the seoClarity study “Do Better Google Rankings Translate to AI Mode Visibility?” (2025), AI Mode responses and organic rankings only partially overlap.

• Partial overlap: 68% of the analyzed AI Mode query responses contained at least one reference from the top 20 Google organic results. For the top 10 rankings, the overlap was 62% of queries.
• Most sources from outside the top 20: Only 20% of all AI Mode citations (out of 10,853 analyzed citations) came from the top 20 organic rankings. This suggests that a strong organic position alone does not guarantee AI Mode visibility.
• Higher ranking position increases probability, but does not guarantee: Although a higher organic position improves inclusion in AI Mode, it is not a guarantee. Position number one appears in AI Mode 27% of the time, position two 21% of the time, and position three 17% of the time.
• Query Fan-Out: AI Mode uses a process known as “query fan-out”. It splits the original query into multiple sub-questions and pulls information from multiple sources to create a comprehensive and synthesized response. This means that optimizing for a single keyword is no longer enough; success depends on providing broader topical coverage.

2. What is the overlap between AI Overviews and organic rankings (Google)

What is an AI Overview?

Google’s AI Overview is a summary, AI-generated response displayed directly at the top of the search results page. It works similarly to Google’s other AI features, providing a quick and synthesized response to the user’s query.

Another seoClarity study “The Overlap Between AI Overviews and Organic Rankings” (2025) provides an overview of the relationship between AI Overviews and organic results.

• Almost complete overlap with sources: Almost all AI overviews (97%) contain at least one URL from the top 20 search results. Even when narrowing down to the top 10 web results, the overlap remained high – 96%. This suggests that traditional SEO rankings still play a major role in the visibility of AI overviews.
• Multiple sources: AI overviews cite an average of about 5 URLs from the top 20 organic results, often consolidating information from multiple sources. For most keywords, there were three overlapping URLs between AI overview citations and traditional rankings.
• Impact of Top 10 positions: Overlap with the top 10 results occurred 50% of the time; when only three sources were cited, the overlap reached 75%. This shows that being in the top 10 significantly increases your chances of being cited, especially when Google selects a limited number of sources.
• Higher rankings correlate strongly with inclusion in results: A clear correlation was found between a page’s search result ranking position and its likelihood of being cited in an AI overview.
  • Position 1 URLs appear in AI overviews 54% of the time.
  • Position 2: 47%.
  • Position 3: 41%.
  • At position 20, this probability drops to just 7%.
• A significant portion of citations from outside the Top 20: Interestingly, 51% of AI overview citations (out of 4.2 million citations across over 432,000 AI overviews) came from URLs outside the top 20 organic results. This shows that a higher organic position does not fully guarantee inclusion in AI overviews, and other factors such as authority, structure, or topical relevance also play a role.
• Healthcare: Healthcare keywords showed the highest average overlap between AI overviews and search result URLs.

3. What Are the Key Strategies for Improving LLM Search Visibility (Meta-Analysis of 19 Studies)

What is an LLM?

An LLM, or large language model, is an artificial intelligence model trained on vast amounts of textual data. LLMs are capable of understanding, generating, and analyzing human language in a way that resembles human thinking. These models can summarize long texts, answer complex questions, and create new content based on learned knowledge and context.

The most well-known examples of LLMs are

ChatGPT and Perplexity AI. They are widely used in both search engines and other applications that require natural language understanding and generation.

“Optimizing for AI Search Engines: A Meta-Analysis of 19 Research Studies” highlights 10 core strategies that are evidence-based and linked to improved LLM visibility. These strategies are based on 19 independent studies and 6 real-life case studies, covering over 10,000 LLM-generated responses.

Table 1: Summary of strategies for improving LLM visibility

Strategy	Studies indicating	Impact credibility score (1-10)	Required resources	Timeframe (for implementation)
Structured content (FAQ/Schema markup)	5	9	SEO, content, front-end developer (for Schema)	1-4 weeks
In-depth content	4	10	Content team, experts, editorial	2-6 months
Brand mentions and digital PR	5	8	PR, outreach, SEO	1-3 months
Knowledge graph and brand presence	2	7	SEO, Wiki contributors, communication	1-2 months
E-E-A-T and trustworthiness	2	10	Content strategy, legal, SEO, design	3-6 months
User-generated content and community	2	8	Community manager, brand, support	Ongoing
Content freshness and timeliness	2	4	Content management, SEO	Ongoing
Fast indexing and Bing optimization	1	7	SEO and development	1-4 weeks
Customer reviews and reputation signals	2	6	SEO, content, product development	Ongoing
Prompt injection (experimental tactic)	1	3	SEO	1-4 weeks

A more detailed overview of core strategies

Correlation coefficient calculation principle: The correlation coefficient (ranging from 0 to 1) indicates how closely each tactic is linked to the improvement of LLM visibility. It is calculated by combining the results of two key factors – the number of supporting studies (on a scale of 0-0.5) and the impact credibility score (on a scale of 0-0.5).

1. Structured content (FAQ, lists, summaries):
   
   • Correlation: 0.81.
   • Why it matters: LLMs prefer easily digestible and well-structured content, which helps AI systems quickly extract and present relevant information. Schema markup, lists, and FAQ blocks significantly improve engagement.
   • Tactical actions: Use literal, question-style headings (e.g., “How does X work?”), break down long explanations into bullet points, add FAQ sections to the end of product or service pages, and implement schema types like FAQPage and HowTo.
2. In-depth, authoritative content:
   
   • Correlation: 0.79.
   • Why it matters: Search-augmented models (e.g., ChatGPT, Perplexity) prefer content that fully answers the query. This is especially true for long, fact-dense, and clearly structured content.
   • Tactical actions: Write long-form content covering multiple subtopics, use expert authors or reviewers, keep the tone confident, objective, and factual, and include definitions, use cases, and examples.
3. Brand mentions and digital PR:
   
   • Correlation: 0.76.
   • Why it matters: Frequent brand mentions from high-authority sources act as trust signals for LLMs. LLMs tend to reinforce the visibility of brands already mentioned in multiple sources.
   • Tactical actions: Appear in “Top X” lists, offer unique data or stories to journalists, and maintain brand activity on LinkedIn, Medium, and trusted industry publications.
4. E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) and trust signals:
   
   • Correlation: 0.64.
   • Why it matters: E-E-A-T signals correlate with visibility in LLM-generated responses. LLMs prefer sources with verified authorship, biographies, and structured trust signals.
   • Tactical actions: Add expert author markup (Schema markup) and biographies to articles, include a detailed “About Us” page and editorial policy, and use citations and external sources to support claims.
5. User-generated content and participation in forums and groups:
   
   • Correlation: 0.54.
   • Why it matters: Mentions on platforms like Reddit, Quora, and niche forums correlate with LLM inclusions. In the case of Estonia, local forums should be considered here. Brands frequently mentioned in informal online discussions tend to appear more positively and frequently in LLM outputs.
   • Tactical actions: Participate in community FAQs, encourage organic Reddit mentions (e.g., AMAs), spread brand discussions in niche forums.
6. Knowledge Graph and brand presence:
   
   • Correlation: 0.49.
   • Why it matters: Presence in structured databases like Wikipedia and Wikidata correlates with LLM-generated brand engagement. Entity recognition plays a central role in grounding responses.
   • Tactical actions: Create or update your Wikipedia page with citations, submit entries to Wikidata or product knowledge graphs, use schema markup to define your organization or entity.
7. Content freshness and timeliness:
   
   • Correlation: 0.34.
   • Why it matters: Up-to-date content is linked to visibility in LLM-generated responses. Platforms like ChatGPT and Bing Chat tend to highlight recently published or updated content, especially for time-sensitive or trending topics.
   • Tactical actions: Publish content referencing new events, tools, or publications, update and republish old blog posts, add timestamped sections like “As of April 2025…”.
8. Fast indexing and Bing optimization:
   
   • Correlation: 0.42.
   • Why it matters: Most AI search tools, including ChatGPT, Copilot, and DuckDuckGo AI, rely on the Bing index to surface and cite web content. If your site is not indexed quickly or fully in Bing, it may remain invisible to LLMs.
   • Tactical actions: Submit new URLs directly via Bing Webmaster Tools, check and fix Bing-specific indexing issues, prioritize fast-loading, mobile-friendly pages, monitor Bing rankings.
9. Customer reviews and reputation signals:
   
   • Correlation: 0.44.
   • Why it matters: Public reviews and online brand reputation, especially on aggregator sites like Trustpilot, Reddit, or G2, can influence whether LLMs consider your brand trustworthy or relevant.
   • Tactical actions: Monitor and improve presence on review platforms (e.g., G2, Capterra, Trustpilot), encourage satisfied customers to leave detailed, keyword-rich reviews, and monitor mentions on Reddit and community forums.
10. Prompt injection (experimental “command injection” tactic):
    
    • Correlation: 0.17.
    • Why it matters: Some researchers have shown that ChatGPT or Bing responses can be influenced by injecting hidden instructions or suggestive phrases into a webpage. However, this tactic is highly experimental, ethically questionable, and unlikely to remain effective in the long run as models improve.
    • Tactical actions (with caution): Add hidden HTML comments or off-screen text with targeted prompts, and use unindexed pages for testing. Do not use this tactic on production websites or mission-critical pages.

AI Mode vs. AI Overviews

Synthesized responses are the new rules. Here are the key differences between these two main AI features.

Google’s AI Mode

• Osaline kattuvus: vaid 68% päringutest sisaldab top 20 orgaanilisest tulemusest pärit viiteid.
• Enamik väliseid allikaid: 80% viidetest on väljastpoolt top 20 orgaanilist tulemust.
• Päringu laialijagamine: jagab päringu alamküsimusteks ja loob vastuse mitmest allikast.

Google’s AI Overviews (AIOs)

• Kõrge kattuvus: 97% AI-ülevaadetest sisaldab viiteid top 20 orgaanilistest tulemustest.
• Mitu allikat: tsiteerib keskmiselt 5 URL-i top 20-st.
• Tugev seos järjestusega: Positsioon 1-20 omab suurimat tõenäosust olla kaasatud.

Key findings from the analysis

• Structure and depth are the foundation of LLM visibility. The strongest and most consistent signal across all studies is clear content formatting, FAQs, schema markup, bullet points, along with long and editorially correct content. LLMs prefer content that fully answers the query and is easily machine-readable and citable.
• External signals and brand mentions amplify engagement, but only with structure. Mentions in PR, roundups, and high-authority media help LLMs “remember” brands, especially when reinforced by structured citations like Wikipedia, Wikidata, or schema-based databases. These mentions alone are not enough, but they increase effectiveness along with well-optimized content.
• Trustworthiness indicators and user attitudes reinforce, but rarely drive citations. Author bios, expert credentials, review platforms like G2, and Reddit discussions act as secondary trust signals. They help LLMs assess legitimacy, especially for commercial queries, but carry no weight without strong content foundations.
• Freshness and Bing indexing act as eligibility filters, not ranking signals. Up-to-date website updates improve your chances of appearing for trend-related queries, and Bing indexing is a prerequisite for the website to be considered at all. However, neither freshness nor visibility in Bing compensates for poor content structure or weak content depth.
• Experimental tactics, such as “prompt injection”, are not viable in the long term. Some tests show that LLMs can be manipulated with hidden prompts, but this method is fragile, not brand-safe, and likely to be patched. Sustainable visibility comes from strategy, not shortcuts.

Summary

Achieving visibility in the evolving AI search landscape requires a shift beyond traditional SEO. While organic rankings remain important, especially for AI Overviews, the emphasis is now on structured, comprehensive, and authoritative content that clearly answers user queries and demonstrates brand trustworthiness across various online touchpoints. Adapting your content strategy to these new paradigms is essential for continued visibility in AI-driven searches.

Frequently Asked Questions (FAQ)