SEO in the AI Overview era — what actually changed

Google AI Overview is rolled out in most regions, Yandex Neuro is wired into Russian SERPs, and Perplexity and ChatGPT Search are separate traffic sources. The classic "top 10 blue links" shrank: users read the AI summary on top and often do not scroll. Per Search Engine Land 2025 data, the average CTR on the first organic result dropped from ~28% to ~16% in eighteen months.

What AI engines pull into answers

• Structured answers to specific questions — definitions, instructions, lists
• Numbers with a cited source and year
• First-hand observations from a Person with verified expertise
• "X vs Y" comparisons with tables
• Recent material — anything stale gets cut

What AI ignores or downranks

• Fluff at the start of an article — three paragraphs about "in today's modern world"
• Vague authors without E-E-A-T signals (name, photo, links to prior work, organization)
• Duplicate content from other sites
• Keyword stuffing without meaning
• Content with no structure — a wall of paragraphs and no subheads

What to do on the site specifically

• Schema.org Article + Person. Author as a Person with a profile URL and jobTitle. AI Overview cites material with an explicit author more readily
• FAQ schema and blocks. AI parses pages with "question — short answer" pairs and often quotes them directly
• HowTo schema for instructions. Step-by-step answers are a top format for AI pickup
• Publish and modified dates in schema. AI cuts old content — without `dateModified` a fresh piece looks aged
• Sources and links. Every number with a link to its origin. AI scores credibility and will put you as the primary source
• Original data. If you have your own study, survey or case — it is gold. AI cites what is not available elsewhere

Content strategy

• Long-tail beats head terms. "Best CRM" no longer drives traffic — AI answers itself. "amoCRM vs Bitrix24 for dental clinics with 5-10 doctors" still has a chance
• Direct answer in the first paragraph. If AI does not find an answer in the first 200 words, it switches to another source
• Subheads as questions. H2/H3 in the form of "What is X", "How to set up Y" — AI prefers FAQ-shaped structure
• Refresh old posts. Every 3-6 months update top content — rewrite the intro, update numbers, add new sections
• First-hand notes over generic facts. "We tested on 12 clients over six months — here is what we found" is not in the training data and AI will cite it

Metrics to watch now

• Impressions vs Clicks in Search Console. If impressions grow and clicks stagnate — you are now showing in AI Overview without click-through. This is the new normal
• Branded traffic. Direct hits by brand name — the leading indicator that AI knows you and users navigate by remembered brand
• AI traffic source. Analyze referrers from `chat.openai.com`, `perplexity.ai`, `gemini.google.com` — a new channel. In analytics, filter as a separate segment
• Time on page from AI sources. Usually higher — the user arrives pre-warmed

What not to do

• Do not block AI crawlers (`GPTBot`, `ClaudeBot`, `PerplexityBot`). It cuts visibility in AI answers and hurts traffic
• Do not ship piles of unedited AI content — original sources will downrank you and so will search engines
• Do not chase "AI-friendly" tricks without substance — algorithms detect template content quickly

What we do at the studio

On client sites and in our own journal — Schema Article with author as a Person, BreadcrumbList, FAQ blocks on service pages, regular updates to old material, clean H2 structure. Not a silver bullet, but it keeps you in the SERP while competitors react slowly. The best format is a short article with one concrete observation, not a "comprehensive 5000-word guide" that AI digests and never sends a click for.