Is Search Dead? No – but the rules of the game for business have already changed

AEO and GEO instead of SEO alone: ​​the new reality of search engine optimization.

Search is no longer what the market has become accustomed to over the past ten years. Users increasingly receive a ready-made answer directly in the interface without having to go to the website. Yandex uses Neuro Search to solve this problem, Google uses AI Overviews, and ChatGPT, Gemini, and Perplexity have long since made the answer their core product, not just an intermediary step before clicking.

How to get into search results in Gemini, ChatGPT, Perplexity, and other neural networks

This is a game-changer for businesses: now it’s important not only to appear in search results, but also to be part of the answer the user sees first. According to Google, AI Overviews has grown from over 1 billion monthly users in 2024 to 2 billion monthly users in 2025. Gemini App exceeded 750 million monthly active users by February 2026, ChatGPT serves over 700 million weekly users, and Perplexity is already building a search infrastructure for 200 million daily queries.

In practice, this means a simple thing: you now need to compete not only for search rankings, but also for the right to be selected by a neural network as a source, phrase, or brand to mention in a response.

Why SEO Alone Is No Longer Enough

Classical SEO hasn’t disappeared. But it’s no longer sufficient if a business’s goal isn’t simply to increase visibility but to maintain contact with its audience. Previously, the path was relatively linear: a user entered a query, saw a list of links, went to a website, and viewed an offer. Today, this route often ends at the platform level. The answer has already been generated, a short summary has already been displayed, and the website remains behind the scenes.

This is confirmed by the numbers. According to Bain research, approximately 80% of users rely on zero-click results for at least 40% of their searches, and approximately 60% of queries end without clicking on an external resource. Bain also estimates a 15-25% decline in organic traffic for brands. Data from Datos and SparkToro shows the same trend: in the US, the share of Google users who clicked on organic results fell from 44.2% to 40.3% over the year, while the share of zero-click scenarios increased from 24.4% to 27.2%.

It’s important not to go to extremes, however. Google, for its part, claims that it continues to send billions of clicks to websites per day and that AI-powered search features drive higher-quality conversions for complex and insightful queries. So the question isn’t whether SEO is dead. The question is whether SEO alone can solve the same problem it once did. And here the answer is no.

Therefore, two more areas are now being added to classic SEO.

AEO, or Answer Engine Optimization, is the optimization of content for systems that immediately answer user questions. This includes quick answers in search, voice assistants, blocks with ready-made explanations, and other answer-first formats. GEO, or Generative Engine Optimization, is optimization for generative systems that collect responses from multiple sources and deliver them as a coherent text. This is the domain of ChatGPT, Gemini, Perplexity, and other AI services.

In practice, the working strategy looks like this: SEO is responsible for indexability, technical foundation, reach, and search presence. AEO increases the chance of ranking in a short, ready-made answer. GEO ensures that your brand, your wording, and your pages appear in generative search results. Not instead of SEO, but on top of it.

For click.ru users, some SEO tasks can be automated through a special module. This is a convenient option for teams who need to manage promotion in a single interface, see traffic and keyword analytics, and quickly implement basic optimization scenarios.

What neural networks consider good content

Large language models don’t read pages the way humans do. They value structure, extractability of facts, source reliability, clear logic, and overall thematic clarity of the material. Simply put, AI isn’t looking for just text. It’s looking for usable knowledge.

1. Relevance

Data freshness is more influential today than many people might think. If nothing has been published about a company in a while, if articles aren’t updated, and the website is full of outdated figures and irrelevant examples, the chances of getting an AI response are reduced. Neural networks are more likely to extract information where they see a live, updated context.

This is especially noticeable against the backdrop of growing AI traffic. Adobe recorded that AI referrals in retail increased 35-fold between July 2024 and May 2025. In travel and banking, the growth was also measured in the tens of times. This means that users actually visit websites after interacting with generative interfaces, rather than simply testing them out of curiosity.

What helps maintain relevance:

• Regular release of new materials;
• Updating old articles with updated dates and fresh data;
• Guest posting on external platforms;
• Constant brand visibility.

If you have an article on end-to-end analytics for 2023, and a competitor’s similar material was updated in 2026 and supplemented with new integrations, market changes, and practical cases, AI will almost always choose the more recent and relevant source.

2. Specifics over generalities

Neural networks work better with what can be extracted from the text without guesswork. Therefore, phrases like “convenient service,” “high speed,” “affordable price,” and “good analytics” are of little use. Measurable characteristics are much more powerful. For example:

• not “fast delivery,” but “delivery in Moscow in 2 hours”;
• not “clear interface,” but “new users complete onboarding in 7 minutes”;
• not “high advertising effectiveness,” but “CPA decreased by 18%, and ROAS increased from 320% to 410%.”

This type of presentation is convenient for a neural network: the numbers can be extracted, compared, integrated into a summary, and used in argumentation.

3. Depth of elaboration

Superficial text is easy to recognize. It has few cause-and-effect relationships, almost no conclusions, no limitations, and no context. Such material may take up space on a website, but rarely serves as the basis for an AI response.

Powerful material is structured differently. It explains not only what happened, but also why. Not just “the company increased conversions,” but “the company redesigned landing pages, narrowed the semantics, strengthened the FAQ section, and added case studies with numbers, after which the share of branded queries increased, and organic conversion improved.” This is the kind of logic that is beneficial to both humans and algorithms.

Neural networks are best suited for:

• detailed instructions;
• practical guides;
• case studies with numbers;
• research and analytics;
• reviews and comparisons;
• testimonials;
• FAQ sections;
• data-driven ratings and selections.

For example, an article “How to choose a laptop for programming under $1,200” will be stronger if it includes a table of models, RAM parameters, battery life, processor type, budget limitations, and comments on use cases. This is a ready-made basis for an answer along the lines of “here are the 3 best options and why.”

4. Information Verification

Neural networks are significantly more likely to use materials that are supported by external confirmation. This could include official websites, industry research, analytical reports, open databases, or the company’s own methodology.

Google explicitly states that it uses structured data to understand page content and information about the world, and Perplexity explains in its help section that it collects responses from web sources in real time. This means that texts without verifiable data are gradually outperformed by materials with transparent sources.

5. Author and Expertise

Today, it’s not enough to simply publish text on behalf of a brand. It’s preferable for the material to be signed by a specific expert. The author must have a profession, specialization, experience, and a clear area of ​​expertise.

Example of a working pitch:

“The material was prepared by Ivan Petrov, Head of Performance Marketing, 12 years in digital, specializing in e-commerce and lead generation.”

This increases trust for people. This is also an important signal for the algorithm: it’s not looking at an anonymous text, but at a publication with provenance and responsibility for its content.

6. Thematic Specialization of the Resource

The narrower and clearer a website’s specialization, the higher its credibility as a source. Resources that cover everything tend to lose out to sites with a clearly defined thematic framework. If a company systematically publishes materials only in its subject area, the algorithm is more likely to consider it a specialized source.

Why the entire digital environment, not just a website, is important

Previously, you could focus on your own domain and think that was enough. Today, neural networks evaluate the broader context. They look at where else the brand appears, who mentions it, how it’s described, and how consistently key statements about the company are repeated.

To increase your chances of being included in AI responses, it’s helpful to:

• publish in industry media;
• present yourself as an expert;
• be included in studies and reviews;
• obtain high-quality links and mentions;
• participate in discussions on forums and communities;
• maintain professional activity on social media.

Here, quality and consistency are key, not quantity for quantity’s sake. If you call your product an “online advertising management platform” on your website, are described in the media as an “ecosystem for performance teams,” and are listed in catalogs as an “analytics service,” it’s more difficult for the algorithm to assemble a coherent image. The more consistent the wording, the higher the chance that it will be retained in AI responses.

Why page structure has become critical

Neural networks find it easier to work with text that is pre-organized into clear semantic blocks.

What is especially helpful:

• comparative tables;
• numbered instructions;
• short paragraphs with a single idea;
• subheadings in the form of questions;
• separate FAQ blocks;
• clear conclusions after sections.

Let’s say you’re writing a material “How to choose a CRM for your sales department.” In one version, this would be 12 screens of continuous text. In another version, you’d add a table of 5 CRMs, a “Who is each suitable for” block, a “Common Implementation Mistakes” section, and an FAQ. The second version will almost always be stronger for both the user and the AI.

How to make a website understandable for neural networks

Good copy isn’t enough if the website itself is difficult to interpret. The algorithm needs to quickly determine the page type and its purpose: is it an article, a case study, a service page, an FAQ, a product card, a company description, or something else.

Entities

Modern search engines and AI models work not only with words but also with entities. An entity is a specific real-world object: a person, a company, a product, a place, a technology, a concept.

Google describes the Knowledge Graph as a database with billions of facts about people, places, and objects. This is one of the foundations of modern entity-based search, not just keyword-based search.

If a user enters the query “Which laptop is best for a designer under $1,500,” the system sees not a set of keys, but a cluster of entities:

• laptop;
• design;
• budget under $1,500.

Then the algorithm searches for pages where this cluster is logically and concretely developed.

Triplets

Another useful principle is triplets. A simple construction: subject -> action -> object.

For neural networks, this is a natural form of storing and retrieving knowledge. If the same pattern is consistently repeated on a website, in the media, in reviews, interviews, and on social media, it begins to be perceived as a stable fact.

Therefore, it is important for a brand to consciously formulate basic statements about itself.

For example:

• Service X automates the distribution of advertising budgets across channels;
• Platform Y simplifies the sales department’s work with incoming leads.

The more often these statements appear in various high-quality sources, the higher the likelihood that they will be included in AI responses.

Schema.org

Structured data remains one of the most understandable ways to explain to algorithms what exactly is on a page. Google explicitly states that it uses structured data to understand page content and to richer display results, and Schema.org remains the primary vocabulary for such descriptions.

When a website has correct markup, it is easier for the algorithm to understand:

• Who the author is;
• What kind of page this is;
• Where the product is and what its price is;
• Where the Q&A section is;
• How the instructions are structured;
• Where the company is and how to contact them.

The bare minimum for most sites:

• Organization or LocalBusiness;
• Article;
• Person;
• Product and Offer;
• FAQPage;
• BreadcrumbList.

You can easily check this using Schema.org Validator and Google Rich Results Test.

llms.txt

The llms.txt file is located in the root of the site and contains a list of priority pages with short descriptions. Essentially, it’s an additional content map for language models. There’s an important caveat here: llms.txt is not yet a mandatory standard, but rather a developing proposal. However, the idea itself has already been formulated and is being actively discussed as a way to simplify LLM navigation through important sections of the site.