By 2026 voice search is no longer "the future" — it's a visible traffic channel. Per Statista, 35% of Russian users query Alice or Marusya at least once a day; in the US it's 50% on Google Assistant and Siri. Voice search's defining trait: the assistant doesn't show 10 results, it reads one — the "best answer". That one is the optimisation target. This article walks through structuring content so Alice and Google Assistant pick your page as the source, and why featured-snippet strategy is the foundation of all voice search optimisation.
How voice queries differ from text
Voice queries are longer. Average word count: text 2.5 words, voice 6.5. Voice queries are conversational. A user doesn't type "weather Moscow", they say "what's the weather in Moscow today". Voice queries are questions. 60% start with "how", "what", "where", "why", "when". Voice queries are local. 22% include geographic context ("near me", a city name). These four traits dictate content structure: long conversational keywords, direct answers to questions, local anchoring.
Featured snippets as the voice-answer source
Main mechanism: 70% of the time Google Assistant reads the text from the SERP's first featured snippet. Alice uses a similar mechanism in Yandex — the expanded answer or "koldunshik". Translation: to land in a voice answer, first land in the featured snippet for that query. Strategy is unified: write a direct answer to the question in the first 50–60 words after the heading. If your answer is direct, structured and in the top 10, you have a shot at the snippet — and therefore the voice answer. Without it, voice search bypasses you.
Long-tail conversational keywords
If your site is only optimised for short commercial queries ("buy smm panel"), you skip voice search. Voice queries are usually full sentences: "which smm panel to pick for Instagram", "how much does an smm panel cost for an agency". Strategy: collect the full list of long-tail questions, build a dedicated section per question or a dedicated article with a direct answer. Good sources: AnswerThePublic, GSC "Search Queries" filtered to > 5 words, the SERP "People Also Ask" section. Site Metrics Tool auto-tags queries longer than 5 words separately — voice candidates.
Structured data for voice
FAQPage JSON-LD is the single most effective markup type for voice. With an FAQ block on the page plus markup, Google Assistant uses the FAQ answer as the voice answer. Similar for Alice. FAQ structure: 5–10 questions, each answer 40–80 words, direct answer without fluff. Speakable schema (voice-specific) and QAPage also help. Not all assistants support these types, but adding them to JSON-LD takes 30 minutes and the effect can be meaningful for assistants that do parse them. Site Metrics Tool auto-includes FAQPage schema in every blog post when an FAQ block exists.
Local voice search
The most valuable voice segment for local businesses. "Where's the nearest car service", "sushi delivery near me", "barber on Main Street". These queries are rarely typed — almost always voice. Three things land you in voice for local. One — a filled Yandex.Business listing with address, phone, hours, reviews. Two — Google Business Profile (for Google Assistant). Three — LocalBusiness JSON-LD on the site with the same data plus geo coordinates. Without those three, a local business doesn't exist for voice.
Optimisation checklist
- Collect the top 100 potential voice queries via GSC + AnswerThePublic.
- Add a section with a direct 40–60 word answer for each, or create a dedicated article.
- Implement FAQPage JSON-LD on every page with an FAQ block.
- For local businesses, fill Yandex.Business + GBP + LocalBusiness markup.
- Optimise Core Web Vitals — slow sites get skipped by assistants.
- Use simple conversational language in opening paragraphs — assistants must be able to read the answer aloud.
Frequently asked
Can voice search traffic be measured?
Directly — no. In GSC and Analytics voice traffic is indistinguishable from text. Indirectly — track growth on conversational ("how", "what", "where") queries longer than 5 words. Statistically these arrive predominantly via voice.
Are Alice and Google Assistant different optimisations?
About 80% same principles (direct answers, FAQ, local data). Alice additionally leans on Yandex's "koldunshik" — the better you rank on Yandex for a query, the higher the chance Alice reads you out.
Will voice replace text search?
In a 5-year horizon — no. Voice is a supplementary channel. Text remains dominant, especially for work and research queries. Voice grows for everyday and local queries.