Intellms: from zero AI visibility to 22 citations in one publishing cycle

Intellms develops custom enterprise IT solutions: internal LLM models, ERP, CRM, and web services. It works in the Russian market. The product is complex and the sales cycle is long. This is a niche where the choice of contractor is formed during market research, not at the last moment.

The starting point was as inconvenient as it gets:

• 0 mentions in ChatGPT, Perplexity, and Yandex answers. The brand did not exist in AI results.
• Young blog pages with no accumulated authority signals.
• The top for most relevant queries was held by mature sites such as Bitrix24 and Cleverence. Technically, AI engines did not even find the site as a potential source. There was no acquisition channel from AI search at all.

We collected 160 queries that potential clients actually put to AI engines on the topic of custom development. We analyzed who AI engines cite now and why.

The conclusion was obvious: with zero authority signals, breaking through head-on on broad general queries is impossible, because the top is already taken by mature domains with a history. We decided to enter through our own narrow expert pages on exact topics instead of competing on general queries with players who have been in the niche for years.

We prioritized topics by the formula business importance × probability of entering results. Following Ben Hunt's Ladder, the pilot deliberately covered the top of the funnel, the awareness and evaluation stages, rather than comparison of specific contractors.

Five materials were published on the Intellms site, built around questions people actually ask AI engines:

1. Developing an internal LLM model for a company.
2. Custom ERP for large business.
3. Custom web development and digital transformation.
4. An AI assistant with a company knowledge base.
5. Single-tenant vs multi-tenant: choosing an architecture.

In parallel we placed rewrites of the same materials on VC.ru, RBC, and Dzen as an additional authority signal for the main domain. We tracked citations for each of the 160 queries in each AI engine separately.

• From May 20 to July 17, 2026, Intellms mentions in AI answers grew from 0 to 22.
• The key detail: 100% of these mentions lead directly to the intellms.ru domain, not to third-party articles about the company. This is the most valuable type of visibility there is. A direct citation means the brand controls the source. It cannot be lost if another site suddenly removes a piece or changes its content.

Five published pages delivered 14 of the 22 citations, so the content worked almost immediately, without a long ramp-up.

The leader is the piece on the internal LLM model. Three different AI engines cite it. That is a signal that the topic hit the structure of real user questions precisely.

All growth in the first cycle came from Russian-language AI engines (Yandex GPT, Yandex AIO) and Perplexity. ChatGPT, Claude, and Gemini do not cite Intellms at all yet.

We expected that. Global models take longer to index young pages and rely more on accumulated authority and UGC signals, which a fresh domain by definition does not have yet. Russian-language AI engines react faster, so the first visible result came from there.

The hypothesis the pilot started with was confirmed: a well-chosen narrow topic enters AI results 1 to 2 months after publication. That is exactly what happened with the Intellms pilot materials.

The next step is the bottom of the funnel. The top (awareness, evaluation) already gives a measurable result. Ahead are queries for direct comparison and choosing a specific contractor, where the purchase decision is actually made. In parallel we adopted a plan to expand external signals (a knowledge base and additional rewrites on authoritative sites), to bring global LLMs in line with the logic that already works for Russian-language AI engines.