Where should you publish in Russia to get cited by AI?
According to Kaufman Lab data as of September 2026, only 3 of 11 platforms deliver both reach and citations in AI answers: Habr, a Russian tech community; vc.ru, a Russian business media site; and your own site as the anchor. Dzen, Sostav, Business Secrets and VK get an adapted version of the text. RBC, RB.ru, DTF, Spark and Workspace are worth a post only when you have news.
- Sample
- 11 Russian-language platforms · 2.4 billion combined visits a month, Similarweb, July 2026 · media citations from Medialogia, Q2 2026, and AI citations from three studies of AI answers · 356 sources
- So what
- Reach does not predict citations. Dzen, Yandex's former blog platform, has 1.2 billion visits but gets under 1% of them from search, while Habr, with 31 million, ranks first for citations in Alice AI, Yandex's AI assistant, in the Visiobrand study. Only Habr's corporate blogs pass link equity to outside sites.
Published Sep 2026 · data as of Jul 2026 · built by 25 AI agents: 11 researchers, 11 verifiers, 3 cross-platform audits · checked by a human
Where to publish to get cited: seven steps
- The original on your site. Put the full version on your own site: every other platform will link to it.
- A core of three. Habr, vc.ru and your own site. Taken across all studies, they deliver both reach and citations.
- Play by Habr's rules. The Sandbox approves about 19% of texts and rejects AI-generated ones: write your first piece yourself and take your time.
- Adapt, don't copy. Dzen, Sostav, Business Secrets and VK get a rewritten take on the topic, not the same text.
- Cadence. One or two pieces a month on Habr and vc.ru. Other platforms only when there is real news.
- A link in every piece. Without a link to the original, the post works for the platform, not for you.
- Measure once a month. Ask AI models about your topic and note who they name instead of you. That list is your publishing queue.
Which platform for which goal
- Personal brand — Habr and vc.ru: people go there to read solutions, not news.
- Mass reach — Dzen has the largest audience, but reach there depends on the topic, not your subscriber count.
- One-off news — RBC, RB.ru, DTF, Spark and Workspace, only when you have real news.
Below is the full study: 11 Russian-language platforms compared by reach, media citations and citations in AI answers.
What we learned: reach and citations across 11 platforms
Reach does not predict citations. Dzen has the largest audience, but the feed, not the domain, decides how far an article travels there. Habr, with 31.2 million visits, ranks first for Alice AI citations in the Visiobrand study (6,316 of 317,045 across 150 questions about GEO); vc.ru ranks first in the expert slice of the Habr study (202 citations in 455 answers). Different studies, different leaders, the same core.
The topic picks the platform, not the ranking. With all studies combined, Habr, vc.ru and Sostav.ru, a Russian marketing trade outlet, are on par. Engineering goes to Habr, business case studies to vc.ru, marketing to Sostav.
Your own site is the entity anchor. Without it, the canonical URL and branded queries go to the platforms, and versions of the same text compete with each other.
What we tested: five hypotheses set before data collection
Five criteria and a selection threshold were fixed before data collection. A platform makes the core if it scores at least 4 of 5 on both SEO and GEO, lets a solo author publish for free, and has an audience that matches the lab's topic: AI and process automation. Sostav.ru meets three of the four conditions, but its readers are marketers, so it sits in the second tier.
How we counted: 25 AI agents and a human
A separate agent profiled each platform using the same set of fields. A second agent checked every figure against its source URL. Three cross-platform audits then looked at all eleven at once, from a GEO, an SEO and a marketing angle. A completeness critic sent empty cells back for re-checking; a human wrote the summary and resolved conflicting figures.
| Lab practice | What it gave this study | Canonical name · source |
|---|---|---|
| One profile per object, same fields | 11 platforms compared on the same columns | Structured codingMiles & Huberman, “Qualitative Data Analysis”, 1994 |
| A second pair of eyes | Every figure is tied to a URL; 4 figures were corrected | Independent double codingKrippendorff, “Content Analysis”, 2004 |
| Two sources for one number | Similarweb next to the media kit and Semrush | Data triangulationDenzin, “The Research Act”, 1978 |
| Threshold before counting | The core criterion was set before data collection | Preregistration of criteriaNosek et al., PNAS, 2018 |
| Log scale for reach | A 4,800× gap fits without squashing the small platforms | The lab's own ruleno external source |
What we found: three charts
| platform | reach | AI cites | own page | link equity | SEO | GEO | total |
|---|---|---|---|---|---|---|---|
| Habr | 3 | 5 | 3 | 3 | 5 | 5 | 24 |
| vc.ru | 3 | 5 | 5 | 1 | 4 | 5 | 23 |
| Sostav.ru | 2 | 5 | 4 | 1 | 4 | 5 | 21 |
| Dzen | 5 | 4 | 4 | 1 | 3 | 4 | 21 |
| VK | 5 | 3 | 5 | 1 | 3 | 4 | 21 |
| DTF | 3 | 2 | 2 | 1 | 3 | 4 | 15 |
| RBC | 4 | — | 1 | 1 | 4 | 3 | 13 |
| Biz Secrets | 2 | — | 3 | 1 | 3 | 3 | 12 |
| Spark.ru | 1 | — | 4 | 1 | 3 | 3 | 12 |
| Workspace.ru | 1 | — | 3 | 1 | 3 | 2 | 10 |
| RB.ru | 1 | — | 1 | 1 | 3 | 3 | 9 |
How far to trust it: limits of the sample
- Sample
- 11 platforms from a list fixed in advance, not pulled from search results. Cnews and other top names in industry rankings are not included.
- Spread
- Similarweb gives estimates, not counts: its figures differ from media kits by up to 2×, and for vc.ru it shows 12.2 million against 7.9 million on Semrush. Where a second source exists, it is given alongside.
- Does not generalize
- All citation studies used marketing and consumer queries, not queries about multi-agent systems.
- Not measured
- No study we found reports AI-answer citations for RBC, Business Secrets, RB.ru or Spark. Visiobrand lists workspace.ru among the 21 domains cited by all nine platforms, but gives no citation count or rank, so it has no score on our scale.
- 1–5 scale
- SEO and GEO come from the summary report. Reach is scored by order of magnitude of visits: under 1 million is 1, each further order of magnitude adds a point, 1 billion and up is 5. AI citations are scored by rank across three studies: top places get 5, a single mention gets 2, no data is shown as a dash. Own page runs from “free, no pre-moderation” (5) to “paid only or through editors” (1). Link equity runs from dofollow (5) to nofollow or a redirect (1).
What to do: three publishing tiers
One or two pieces a month on Habr and vc.ru, with the original on your own site. Dzen, Sostav, Business Secrets and VK get one adapted version per topic. RBC, RB.ru, DTF, Spark and Workspace only when you have news.
The barrier to entry on Habr is measurable: the Sandbox approves about 19% of texts and rejects AI-generated ones.
Common questions about the platforms
What is the platform map for AI case studies?
The platform map for AI case studies covers 11 Russian-language sites where an expert can publish a piece about artificial intelligence, free or paid, and compares them by reach, citations and link equity. Kaufman Lab compiled it from July 2026 data. It does not measure the return on any single article and is no substitute for testing your own topic.
Where should I publish AI articles so ChatGPT, Alice and Perplexity cite them?
On the Russian-language internet the answer comes down to Yandex, Russia's largest search engine. Alice AI takes up to 90% of its sources from the Yandex top 10 and repeats the same list in 67% of cases, so you need a page that ranks in Yandex's top results on a strong domain. Western models take 68% of their citations from third-party domains, not from the brand's own site. Across three studies, Habr, vc.ru and Sostav.ru are cited most often; for your own site, they serve as sources of mentions.
Habr or vc.ru: which suits a technical deep dive, and which a business case?
A technical deep dive goes to Habr: 64% of its audience are mid-level or senior developers, and GEO and SEO both score 5 of 5, but the Sandbox approves about 19% of texts. A business case goes to vc.ru: an audience of decision makers, a free blog without pre-moderation, and 1st place for Alice AI citations in the expert slice. Don't post the same text to both: vc.ru bans duplicates within the site, and Habr expects technical depth.
Do links from Habr, vc.ru and Dzen pass link equity to a site?
Almost never. Personal articles on Habr use rel=nofollow, vc.ru and DTF route links through a redirect blocked in robots.txt, Dzen uses nofollow or a redirect, and VK uses the away.php redirect. Only Habr corporate blogs pass direct equity, from 222,000 RUB for six months (about $2,600). These platforms are worth it for mentions and citations, not for link equity.
Can you publish AI-assisted text on Habr and Dzen?
Habr bans fully or partly generated texts, except illustrations and a small share of long technical articles; Sandbox submissions are checked with GigaCheck, Sber's AI-text detector. Dzen bans content created without the author's involvement, and the penalty is that posts are shown to subscribers only. vc.ru has no specific rules. Since 2023 Spark has rejected duplicate and rewritten text. The safe setup: the original on your own site, and on the platforms an adapted version the author worked on.
How much does publishing on RBC, Habr and vc.ru cost?
RBC: there is no free publishing; RBC Companies is paid and open only to sole proprietors and companies, while Pro and Trends accept pitches to their editors. Habr: a personal post is free, a corporate blog costs from 222,000 RUB to 1,220,000 RUB for six months (about $2,600–14,200). vc.ru: a personal blog is free, Pro starts at 32,900 RUB a month (about $385), a sponsored article at 500,000 RUB (about $5,800). Dzen, Sostav, Business Secrets, VK, Spark and Workspace offer a free page with limits on volume and uniqueness.
How many days until a new article shows up in AI answers?
In the Visiobrand study, Russian platforms pick up a new page in 7 days on average, Perplexity in 13 and ChatGPT in 18. Once a page is in Alice AI's pool, it tends to stay there: Alice repeats the same source list in 67% of cases, against 32% for ChatGPT. These figures come from marketing and GEO queries; no one has measured this for multi-agent systems.
- Similarweb, domain pages for dzen.ru, vk.com, vk.ru, rbc.ru, companies.rbc.ru, habr.com, dtf.ru, vc.ru, secrets.tbank.ru, sostav.ru, spark.ru, rb.ru, workspace.ru; snapshot of July 2026. similarweb.com
- Медиалогия. Рейтинги «Федеральные интернет-ресурсы, июль 2026», «IT и телеком СМИ, II кв. 2026», «Финансовые СМИ, II кв. 2026». [Medialogia, a Russian media monitoring company: rankings of federal online media, July 2026; IT and telecom media, Q2 2026; financial media, Q2 2026] mlg.ru
- Visiobrand. Как нейросети выбирают источники: 317 045 цитат, 9 платформ. 20.08.2026. [How AI models choose sources: 317,045 citations across 9 platforms] visiobrand.io
- Хабр. Кого цитирует Алиса AI: 455 ответов, 2 976 цитат. 19.08.2026. [Habr: whom Alice AI cites, 455 answers and 2,976 citations] habr.com/ru/articles/1072206
- Хабр. Правила публикаций, редакция 08.07.2026; отчёт Песочницы за 2025 год; прайс корпоративных блогов. [Habr publishing rules as revised on 08.07.2026, the 2025 Sandbox report, corporate blog price list] habr.com/ru/docs/help/rules
- SpyWords через SEOnews. В нейропоиск Яндекса чаще всего попадают Дзен, Википедия и Ответы Mail. 11.07.2025. [SpyWords via SEOnews: Dzen, Wikipedia and Otvety Mail are the most frequent sources in Yandex's AI search] seonews.ru
- vc.ru. Правила и тарифы Pro. [vc.ru rules and Pro pricing] vc.ru/rules · vc.ru/ads
- Дзен. Требования к контенту и правила о спаме. [Dzen content requirements and spam rules] dzen.ru/help
- Aggarwal P., Murahari V., Rajpurohit T., Kalyan A., Narasimhan K., Deshpande A. GEO: Generative Engine Optimization. KDD, 2024. arXiv:2311.09735
- Nosek B. A., Ebersole C. R., DeHaven A. C., Mellor D. T. The preregistration revolution. PNAS, 115(11), 2018. doi:10.1073/pnas.1708274114
- Krippendorff K. Content Analysis: An Introduction to Its Methodology. Sage, 2004.
- Denzin N. K. The Research Act: A Theoretical Introduction to Sociological Methods. McGraw-Hill, 1978.