How Artificial Intelligence Is Changing Search Results and SEO

Artificial intelligence is rapidly moving into the core of search engines. Yandex is developing Neuro, a service that generates quick answers with a neural network, Google has introduced AI Overviews and an experimental AI Mode for users in the United States, and Perplexity has offered LLM-powered search since 2022. For website owners, and especially for news and media publishers, these changes raise a central question: how much traffic will remain when a machine-written answer appears above the organic results? This article summarizes the emerging research on AI search and its impact on SEO, and adds context for the Russian-language market.

The most visible change in Google results is the AI Overview block, a zero-click answer placed above traditional listings. In Russian-language search it is labeled 'Обзор от ИИ' and has quickly become one of the most prominent features for informational queries. Yandex's equivalent, Neuro, now appears on most informational searches, often as the first and effectively 'zero' result. Both systems cite source sites, but the links are less prominent in Google, where users must expand panels to see them, while Yandex displays site names more clearly.

Several analyses attempt to quantify the effect of AI answers. An Axios analysis using Similarweb data of 500 top US news sites found AI bot queries rising to 5.5 million while search referrals to publishers dropped about 15%. Ahrefs compared 300,000 keywords with and without AI Overviews and found the CTR of the first organic result fell by roughly 34.5%. Amsive analyzed 700,000 keywords across finance, education, SaaS, healthcare and pet industries, finding AI Overviews in about 15.5% of cases, with organic CTR down about 20% and featured snippet CTR down about 37%. Seer Interactive observed lower organic CTR when AI Overviews appeared, while paid CTR declined less. Each of these studies measures organic clicks rather than total traffic, which limits direct conclusions for publishers.

Research focused on news queries is particularly concerning. Press Gazette analyzed 3,300 high-value queries for leading publishers and found AI-generated summaries appearing in nearly a quarter (23.7%) of cases, pushing the first organic publisher result down by roughly one screen (about 980 pixels). The effect varies by topic: health-related queries triggered AI Overviews in 55% of cases, while branded searches for a specific publication were never replaced. Authoritas found that 97.3% of AI Overview responses shared at least one domain with the top ten organic results, a sharp change from a year earlier when almost all generative sources came from outside top-ranking domains. Publishers therefore see their content paraphrased and summarized above the very links that would have sent users to them.

Not all data points in the same direction. Semrush reported that organic traffic to several large publishers, including ProPublica, BuzzFeed, Rolling Stone, Vanity Fair and New Republic, grew between May 2024 and February 2025, suggesting that the overall impact may be uneven. Tools such as Parse.ly and Chartbeat have not shown a clear aggregate decline attributable to AI search, although Google does not separate AI-triggered clicks from ordinary clicks, making attribution difficult. Some claims circulating in the industry, such as a 54% drop in clicks for repair-related queries or a 50% traffic decline in a specialized niche, are unsupported by published data and should be treated with skepticism. Even before AI Overviews, a SparkToro study found that only about 29% of US searches led to a click on a website from Google results.

Beyond traffic, AI search raises concerns about accuracy and sourcing. The Tow Center for Digital Journalism tested eight generative search tools with content from 20 news publishers and found frequent citation errors; Google's Gemini linked the correct original article in only 16 of 261 cases (about 6%). Generative models produce answers that sound plausible based on patterns rather than grounding them in specific sources. In one informal test, several AI assistants summarized a nonexistent article by drawing on other outlets that had not covered the topic. TollBit's analysis of AI chatbots, which is methodologically debatable, reported that chatbots generate far fewer referrals than Google Search, while scraping publisher content hundreds of millions of times.

In Runet, Yandex's Neuro has become a central element of search. Launched in early 2024 in Yandex Browser and the Alice app, it became part of the main search in October 2024 and by May 2025 could no longer be switched off. By April 2025, Neuro usage had grown 1.5 times and reached about 10.5 million daily users. According to Yandex, the share of clicks from Neuro answers to source sites has remained unchanged, which suggests that users who want details still follow links. Both Yandex and Google warn that AI answers may contain errors, and users can rate responses to help train the models. Yandex allows follow-up questions and a dedicated Neuro tab, while Google's overviews appear or disappear depending on query sensitivity.

For sites promoted in Runet, the priority is to adapt rather than retreat. First, focus optimization efforts on Yandex, where source links from Neuro are more visible and clickable. Second, build informational semantics: map the questions users ask in your niche, and create complete, well-structured texts written in clear language. Include key phrases naturally in meta tags, headings and body content. Third, implement structured data markup so search systems can accurately extract facts, definitions, FAQs and product details. Fourth, manage reputation and brand queries, such as 'company name + reviews', by publishing authoritative pages and encouraging genuine customer feedback so that AI answers draw accurate information about your business. Large news publishers may need to rethink their traffic strategy more fundamentally.

The evidence on AI search is incomplete and sometimes contradictory, but the direction of change is clear: AI-generated answers are pushing organic listings down and reducing clicks for many informational queries, while also repackaging publisher content. Publishers who depend on search traffic face the greatest risk. Site owners who respond with clear, expert, structured and accurately attributed content, and who optimize for the search systems their audiences actually use, will be best positioned to benefit from the AI-driven search era.