TRADITIONAL SEARCH IS FADING: HOW BUSINESSES CAN ADAPT AND GROW IN 2026

Statistics from 2025 confirm this tectonic shift. Demand for chatbots has surged dramatically. In the Russian-speaking internet alone, ChatGPT receives over 1.17 million search queries per month, while globally this figure is approaching 47 billion.

Why are users moving from search bars to prompts?

• Speed and structure. People turn to AI for ready-made answers. There’s no need to navigate websites: everything comes in a conversational format.
• Ecosystem integration. Search can happen via voice, images, or documents, with concise and up-to-date summaries as the result.
• Professional adoption. In science, education, and finance, AI-based search already accounts for over 40% of searches among active users.

Despite the trend, Google remains the absolute leader, holding around 90% of the global search market, with approximately 13.6 billion queries per day. However, for the first time in nine years, this share has begun to decline. To retain users, Google and Yandex have introduced AI-generated answers, allowing even conservative audiences to experience the convenience of generative responses.

At the same time, a new habit of information consumption is forming. According to research by the APUM marketing agency, impulse.guru, and the Setka platform, 85% of Russians still use traditional search engines, but 29% already actively use LLMs (ChatGPT, YandexGPT) specifically as search tools. Among users aged 18–24, this figure reaches 47%. This is a clear signal for businesses: your future customers and young professionals are no longer searching for you in a search bar, but in a neural network chat window.

Is SEO Dead? Long Live AEO and GEO

This trend has made marketers uneasy: “Why did we invest in SEO if people no longer click links?” The answer is simple: links haven’t disappeared, but the way they’re distributed has changed. Your task now is not just to rank in the top 10, but to be included in a specific AI-generated answer.

New approaches are replacing the classics:

1. AEO (Answer Engine Optimization). Its goal is to ensure that chatbots (ChatGPT, Perplexity, Copilot) pull information directly from your content.

• Focus: clear Q&A pages, fact-based content with dates and sources, structured data markup, author expertise (E-E-A-T*).

2. GEO (Generative Engine Optimization). Its goal is to prepare company data so AI can easily find, parse, and legally cite it.

• Focus: machine-readable data (JSON-LD, APIs), licensed datasets, white papers, stable URLs.

*E-E-A-T (stands for Experience, Expertise, Authoritativeness, Trustworthiness) is a concept describing how well a source and its author demonstrate real experience, expertise, authority, and reliability. The more facts, confirmations, and transparent links between entities (companies, people, projects), the higher the level of trust from search engines and AI models.

Visibility in LLMs depends heavily on publishing content on relevant platforms such as Habr, VC.ru, Medium, your own blog, or Telegram channels.

How does it work? If your company specializes in AI agents and publishes an in-depth article on virtual assistants, an LLM builds a knowledge chain: “This company writes in detail → they have expertise → I can cite them as experts.”

Publish case studies, share experience, and don’t be afraid to show how things work internally. Models learn from text, and the more high-quality material you provide, the higher your chances of being included in AI answers.

At Doubletapp, we closely track these changes and see confirmation in our own analytics. Yandex.Metrica shows steady growth in traffic coming to our site from ChatGPT.

In December, 12 users came to the Mini Apps landing page from ChatGPT, and 5 of them completed target actions. The traffic doesn’t just exist — it drives meaningful engagement.

Moreover, this directly affects reputation in AI-powered results. For example, for the query “top AI developers in Yekaterinburg”, Perplexity ranks Doubletapp first, building the list based on our digital footprint across Russia. This proves that neural networks can already identify market leaders using available data.

But content alone is no longer enough. To lock in these results and become a “source of truth” for AI, you need to speak to machines in their own language.

How LLMs Work: Customer Ratings and On-Site Markup

To understand how to appear in AI-generated answers, you need to start thinking like a neural network. Unlike humans, an LLM doesn’t “read” a website — it analyzes it in search of clear, trustworthy signals. Classic SEO copy overloaded with keywords often registers as mere “noise.” What the model actually needs are facts and evidence.

This is where two key elements come to the forefront: customer ratings (social proof) and technical markup (the machine’s language).

Customer Ratings: Social Proof for LLMs

For an LLM, a rating isn’t just decorative stars. It’s a direct, measurable, and hard-to-fake trust signal.

Models like Google AI Overviews or Perplexity process data at massive scale. A claim on your landing page like “We are market leaders” means very little to them. But “4.9/5 rating based on 1,000 reviews” is a concrete, verifiable fact that can be used to construct an answer.

LLMs go beyond simply reading the score. They analyze the semantics of reviews. If dozens of reviews mention “a strong Android development team” or “on-time delivery,” the model internalizes this as a confirmed attribute. Later, when asked “reliable mobile app development vendor,” it is highly likely to include Doubletapp in its response.

Ratings and reviews on external, authoritative platforms are especially valuable — such as Clutch, Behance, Google Maps, or Habr. In Doubletapp’s case, industry awards provided additional proof of expertise, including the Tagline Awards. Content published on the Tagline domain is well indexed and frequently appears in top search and AI-generated result blocks. When an LLM registers that a company has won “Mobile Development” at a major industry competition, it more strongly associates that company with mobile app expertise. For algorithms, this signal carries more weight than dozens of generic mentions.

Website Markup: Translating from Human Language to AI Language

If ratings represent what others say about us, markup is how we speak to the neural network. It’s special code (structured data) added to pages to explicitly explain content to machines.

How it works:
A page might say: “Doubletapp, Yekaterinburg.” A human understands this is an office location. An LLM may not. By using LocalBusiness schema, we explicitly state: “This is an organization. Here is its name, address, phone number, and rating.” This makes the data machine-readable.

1. Answer Structuring (AEO): Markup like FAQPage or HowTo is a goldmine. An LLM sees a clear Question → Answer structure and can lift it directly into its response. For example, when asked “How does onboarding work at the company?”, the model is far more likely to use a properly marked-up block from our site than to extract meaning from competitors’ unstructured text.
2. Entity Relationships (E-E-A-T): Structured data allows us to connect entities. For example, this Case Study → done for this client (Organization) → written by this author (Person). This builds a rich factual graph for the LLM and increases author credibility.

Many still think Schema.org exists only for “pretty stars” in Google. In the GenAI era, its role has become foundational. Without markup, your site is unstructured text that the model must parse — risking hallucinations. Schema.org turns content into a database (JSON-LD). It’s an instruction to AI: “Don’t guess. Here is the exact data.”

Code Example: How to Explain to AI That You’re an Expert

Instead of forcing the model to parse HTML, we provide a clean data object:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Тренды AEO и GEO оптимизации в 2025 году",
  "publisher": {
    "@type": "Organization",
    "name": "Doubletapp",
    "url": "https://doubletapp.ru",
    "founder": {
      "@type": "Person",
      "name": "Сергей Анчутин",
      "jobTitle": "CEO",
      "sameAs": [
        "https://vc.ru/u/sergey-anchutin",
        "https://www.linkedin.com/in/sergey-anchutin"
      ]
    },
    "award": [
      "Tagline Awards 2024: Gold in Mobile Development"
    ],
    "sameAs": [
      "https://tagline.ru/awards/winners/",
      "https://clutch.co/profile/doubletapp"
    ]
  },
  "mentions": [
    {
      "@type": "Thing",
      "name": "Generative AI",
      "sameAs": "https://en.wikipedia.org/wiki/Generative_artificial_intelligence"
    }
  ]
}
</script>

In this example, we not only name the author but also link to verified profiles (sameAs), confirming their real-world existence. We also connect the article to the Generative AI global entity via Wikipedia, clarifying the context.

How Simple Mechanisms Push Businesses to the Top

To sum up, the rules of the game have changed. Simply “being online” is no longer enough. The new goal is to become the most reliable source of truth for AI.

Customer ratings and technical markup aren’t magic — they’re a direct conversation with LLMs. You’re telling the machine: “Here’s proof of our value from real people, and here are the facts in a format you understand.” You’re not trying to manipulate the algorithm — you’re becoming its partner. By investing in real reviews and implementing JSON-LD, you transform your business from “just another link” into a ready-made authoritative answer.

For an agency like Doubletapp, the goal is AEO — provide a precise answer and capture a lead. But for large platforms, the rules change. This is where GEO begins.

The difference is fundamental:

• AEO is a sniper tactic: hit the snippet, answer the question.
• GEO is a knowledge-base strategy: its goal is to become the source models learn from.

For giants like Avito, Sber, or Ozon, semantic markup evolves from SEO routine into an API for neural networks.

How do market leaders do it?

1. Marketplaces & Classifieds (Avito, Wildberries)
   With millions of dynamic pages, deep markup using Product, Offer, and ItemCondition turns HTML chaos into a structured dataset. When a user asks ChatGPT “Average price of a used MacBook Pro M1”, the model doesn’t hallucinate but relies on data fed through markup.
2. Fintech (Sber, T-Bank)
   Here the YMYL (Your Money or Your Life) factor applies. Accuracy is critical. Using FinancialProduct schema and linking to Corporation entities via the Knowledge Graph helps protect brand integrity. If an LLM is asked “Sber credit card terms,” it should pull data from the official marked-up page, rather than a forum.

Two Search Loops: Why AEO and RAG Are One Ecosystem

From an engineering perspective, search splits into two loops:

1. External Loop (AEO / GEO): This is how the world sees you. It’s about global models — ChatGPT, Google Gemini, YandexGPT. Its goal: appear in answers and attract customers.
2. Internal Loop (RAG): This is how your company operates internally. It’s about local models, corporate knowledge bases. Its goal: help employees instantly find documents, policies, or code.

The core problem is the same in both loops: hallucinations caused by unstructured data. Poor data externally means lost customers. Poor data internally means incorrect answers, slow RAG systems, and high costs.

At Doubletapp, we see the full picture. By applying GEO principles, we solve both loops simultaneously. Structured data becomes ideal fuel for external visibility and internal vectorization.

That’s why we go beyond consulting. Our case studies include building custom RAG systems and search pipelines where data is structured once and reused everywhere for both customer acquisition and internal process automation.

For example, we’ve implemented an AI Contractor Assessor for industrial companies: it processes massive tender documentation and cuts analysis time to 5–7 minutes, even when file structure is complex, which makes the process more efficient and transparent.

Another case study is an HR AI screening system: it automatically filters applications, removes irrelevant candidates, and improves hiring efficiency by 40%, saving a significant amount of time at the initial screening stage.

The era of clicking through dozens of links is ending. In the new reality, the winners aren’t those who shout the loudest, but those whose data is most understandable to machines. Neural networks look for facts, not slogans. The choice is simple: you either structure your knowledge and become a ready-made answer or remain informational noise for algorithms.

AEO/GEO isn’t just the next version of SEO. It’s the creation of a single source of truth for your business. By investing in clean data and structured markup today, you solve two problems at once: capturing AI-driven demand externally, attracting customers, and building a perfect foundation for internal RAG systems, improving operational efficiency.