Playbook · GROW
Search has split in two. We optimize for both.
Classic SEO gets you ranked. GEO - generative engine optimization - gets you quoted: by ChatGPT, Perplexity, Google AI Overviews and Алиса. This is the method we run on client sites, with a worked example and live numbers below.
48%
of Google searches now top out with an AI answer
up from 34.5% in Dec 2025
58.5%
of searches end without any click
visibility ≠ clicks anymore
12%
of ChatGPT-cited URLs also rank Google top-10
AI visibility is a separate game
46.5M
people see Алиса AI answers monthly
fed almost only by Yandex top-5
Industry studies, June 2026 - re-verified quarterly.
Eight rules, every page
01
Answer first
The direct answer sits in the first 2-3 sentences of the page and of every section. 44% of AI citations come from the first third of a page - engines never wait for the conclusion.
02
Tables over prose
Structured data - comparison tables, step lists - is 28-40% more likely to be cited than paragraphs. Every money page carries at least one.
03
FAQ + schema
Human-phrased questions with concise answers, marked up as FAQPage JSON-LD. Roughly +40% weight in ChatGPT source selection.
04
Atomic paragraphs
Each paragraph answers one question and survives being quoted alone. Perplexity lifts blocks verbatim - we write blocks worth lifting.
05
Real, verified numbers
Original stats raise AI visibility 30-40%; unsupported claims get skipped. Every rate, law and price is verified against current sources before publishing - never from memory.
06
Freshness with substance
AI engines weight recency in days, not months. Monthly refresh pass on top pages - real fact updates, not cosmetic date bumps (engines detect those).
07
One page per intent
Every query cluster gets a dedicated page that matches its intent exactly. A service page ranking #15 for an informational query is a missing article, not a ranking problem.
08
Topical depth
Google's 2026 updates reward sites that go deep in one subject over broad shallow ones. We build the full stack per niche: services, guides, glossary, FAQ - and refuse off-topic content.
Each engine picks sources differently
| Engine | What gets you in | What gets you picked |
|---|---|---|
| Google AI Overviews / AI Mode | Normal top-10 rankings | Answer-first blocks, tables, freshness |
| ChatGPT search | Bing index + domain authority | FAQ schema, citations, first-third answers, recency |
| Perplexity | Its own crawl | Literal extractable Q&A blocks, ~30-day freshness |
| Яндекс Алиса AI | Yandex top-5 positions | Answer-first, tables, Article + FAQ schema |
Trust Way - freight forwarding, Kazakhstan
A young domain in a competitive B2B niche. The full method applied: topical depth (39 articles + a 40-term customs glossary), answer-first structure, FAQ schema, verified regulatory facts, monthly freshness. Numbers below are real Search Console data, June 2026.
9.0 → 3.0
Position on the head commercial query
«логистические компании казахстана», ~2 months
+105%
Search clicks, 28d vs previous 28d
Google Search Console, June 2026
1,750
AI-driven impressions in 7 days on ONE glossary page
Google AI Mode fan-out citing it as a source
+427%
Search impressions, 28d vs previous
content cluster + technical fixes compounding
The GEO signal worth knowing
When Google's AI Mode researches a question, it fans out into dozens of synthetic sub-queries and pulls sources for each. In Search Console this looks like a swarm of long, oddly specific queries with zero clicks - most teams read it as noise. It's the opposite: the AI is citing your page. One Trust Way glossary page drew 1,750 such impressions in a week. We then add the how-to sections and FAQs that turn that citation visibility into human clicks.
Not a project - a loop
Weekly
Scoreboard refresh: GA4 + Search Console deltas per site, health checks, content-change review against the playbook rules.
Monthly
Freshness pass: every number, rate and date on the top-10 pages re-verified against current sources and updated with substance.
Quarterly
Method re-verification: the AI-search landscape moves monthly; the playbook's own claims get the same fact-check we give client content.