Alisa, ChatGPT and Perplexity Cite Only 34 of 309 Article Platforms
For a neural network to recommend your product, it has to be mentioned on the pages it reads. A marketing course author compiled 309 free publishing platforms, asked Alisa, ChatGPT, Perplexity and DeepSeek 67 questions, and collected 316 answers. Only 34 of the 309 platforms appeared among the sources, and the remaining 275 did not appear in a single answer. That is why it makes sense to measure your niche separately for each model before writing any articles.
AI-processed from Habr AI; edited by Hamidun NewsWhen a person asks Alisa or ChatGPT "recommend a neural networks course for a marketer," the model assembles its answer from other people's pages: roundups, reviews, catalogs. So to make it into the answer, you have to be on the pages the model reads. This is what GEO, or generative engine optimization, refers to.
The author of a marketing course decided to test in practice where it is worth publishing. She compiled 309 platforms where an article can be placed for free (178 Russian-language and 131 English-language), asked four models 67 questions, and got 316 answers. It turned out that the neural networks cited only 34 of the 309 platforms.
The other 275 never appeared among the sources. She described the results on Habr.
Why one article cannot be spread across all the platforms
The scheme of "write it on a blog and adapt it for the platforms" broke down at the first edit. RB.RU states directly in its rules: if the text was sent to someone, posted on a blog, or published openly, the column will not be accepted.
Claude Code checked the rules of all 178 Russian-language platforms against the author's technical brief. The quotes of the most important platforms were verified by hand, and only what is written in plain text was taken into account:
- 19 platforms require exclusivity, among them Cossa, RB.RU, Lifehacker, Executive.ru, SEOnews, PPC.world, Tproger, iXBT Live and the T-Bank blogs;
- 30 run texts through anti-plagiarism checks, with a threshold of 75% on Habr, 85% on Konnekt, and 90% on the 1PS blog;
- 9 allow you to transfer your own text, including VC.ru and DTF; Dzen asks for a cross-link;
- 50 say nothing about uniqueness, and 70 have no author rules at all.
Honest cross-posting works on roughly ten of the 178 platforms. Another thirty or so will require rewriting the text almost entirely.
What the check across four models showed
For the check, the author used the GEORank service: it asks neural networks the questions of clients and saves the answers together with the list of sources. There were 67 questions. Of these, 55 general ones were generated by the service itself, for example "where to sign up for AI marketing courses and how much they cost." The author added another 12 narrow and brand-specific questions by hand. They were run through Alisa, ChatGPT, Perplexity and DeepSeek via API, so personal chat history did not affect the answers. Narrow questions were asked twice, because the answers fluctuate. Of the 316 answers, 267 contained source links.
The main conclusion: almost nine-tenths of the list can be set aside for this task. The models nevertheless read different sites:
- Alisa cited Dzen 16 times; the other three models never did. All 9 answers that cited Pikabu also came from Alisa, and Timeweb Community received 25 mentions from Alisa out of 26;
- Perplexity cites Habr, VC.ru and Sostav more often than the others: Sostav appeared in 8 of its answers, and in 3 of Alisa's;
- ChatGPT reads the widest range: on general questions it cited 319 different sites, Alisa 153, and Perplexity 117;
- DeepSeek rarely provides links: sources appeared in only 30 of its 79 answers, so tracking it through sources is pointless.
Therefore, "getting published on VC" is a bet on specific models. If your clients ask Alisa, Dzen and Pikabu will matter more than your reputation in professional circles.
Roundups and catalogs matter more than expert articles
In the AI marketing course niche, the source pages were almost always roundups such as "Top 7 Online Neural Network Courses for Marketing" on Timeweb Community or "TOP 8 Online Courses on Using AI in Marketing" on Cossa. The models retell them as a ready-made answer. So instead of an expert article, it turned out to be more useful to get into existing roundups or to create your own honest one, where the course sits alongside its competitors.
Course catalogs are a separate channel. The Checkroi catalog appeared in 30 answers, Habr Career in 12. Adding a course card there is enough.
The course's own website appeared among the sources for brand questions in 21 of 32 answers, and the author or course was named in 28 of 32. On narrow questions without the course name, it was never named in any of the 64. The neural networks know the course by name but do not recommend it to anyone themselves. External platforms are what should close this gap.
To prioritize, the author calculated a score: the number of answers that include a platform, divided by the entry threshold. Self-publishing scores 1, pre-moderation 2, editorial review 3, plus 1 for exclusivity. The leaders are VC.ru (36), Habr (26.5), DTF and Dzen (16 each). In the end she kept six platforms and two catalogs: one adapted article for VC.ru and Dzen, a separate roundup for Timeweb Community, a text for Sostav, an exclusive for Cossa, short articles on TenChat, and cards in Checkroi and Habr Career.
Limitations and what to do yourself
The sample is small: 316 answers, one niche, collected 7–9 November 2026. Platform rules change, and there are no results "after publication" yet: the author promises a repeat check in 2–4 weeks. It is also worth remembering that the text was written by the course author and at the same time promotes the service she used.
The ranking from the article does not transfer to other niches: in loans or legal services, the models will cite other sites. But the check can be repeated in an evening. Formulate your clients' questions, run them through several models via API, match the sources against your list of platforms, and open the pages themselves. The format (roundup, review or catalog) often matters more than the name of the platform.
- Nov 21, 202622 Claude Agents Spent Six Weeks on Black Hole Problems — and Solved Two
- Nov 21, 2026Google to split Gemini chat history from model training, so you can opt out without losing your history
- Nov 21, 2026Apple's macOS 27.3 beta 2 will let you limit AI agents' access to Mac files
- Nov 21, 2026Apple Is Building an AI Agent for Mac While Making It Harder for Outsiders to Get In: 14 Clicks Instead of 4
Key Questions Answered
Why one article cannot be spread across all the platforms?
The scheme of "write it on a blog and adapt it for the platforms" broke down at the first edit. RB.RU states directly in its rules: if the text was sent to someone, posted on a blog, or published openly, the column will not be accepted. Claude Code checked the rules of all 178 Russian-language platforms against the author's technical brief.
What the check across four models showed?
For the check, the author used the GEORank service: it asks neural networks the questions of clients and saves the answers together with the list of sources. There were 67 questions. Of these, 55 general ones were generated by the service itself, for example "where to sign up for AI marketing courses and how much they cost." The author added another 12 narrow and brand-specific questions by hand.
What are the key details on Roundups and catalogs matter more than expert articles?
In the AI marketing course niche, the source pages were almost always roundups such as "Top 7 Online Neural Network Courses for Marketing" on Timeweb Community or "TOP 8 Online Courses on Using AI in Marketing" on Cossa. The models retell them as a ready-made answer. So instead of an expert article, it turned out to be more useful to get into existing roundups or to create your own honest one, where the course sits alongside its competitors. Course catalogs are a separate channel.
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.