Intellms case study: a software developer's site caught the niche leader in AI citations, 71 vs 72 for Bitrix24

Intellms wanted to appear in AI answers to queries about ERP development, LLM solutions and corporate services. The main goal was for AI assistants to recommend the company at the "who should we hire" step, because that is where a lead forms.

At the start of this phase the project had three problems:

• the site was young and had no history in AI answers;
• the funnel was lopsided: the top (Awareness) was covered, but the bottom (Decision) had only 5% coverage;
• ChatGPT and Gemini had never cited the site.

Parameter	Value
Client	Intellms
Site	intellms.ru
Niche	Custom corporate software: ERP, LLM solutions, web services
Market	🇷🇺 Russia
Period	August to October 2026, measurements on 18 Aug and 26 Oct
Measurement	AI Visibility, run 723: 425 queries on 5 engines. The funnel is compared on 160 queries

A buyer asks an AI assistant something like "who can build an in-house LLM for us end to end". The assistant composes an answer and cites intellms.ru as a source. The reader clicks through to a case study or a service page and enters the sales funnel.

As of 26 October, the site is cited in answers to 36 of the 160 funnel queries. In November we plan to widen that coverage.

The work ran in three stages.

• Analytics and strategy. We built a query pool across four funnel stages and took a baseline: 44 citations, 12% coverage, and the site trailing Cleverence among vendors. In September we approved 53 new topics and loaded 265 more queries for them, which brought the full set to 425.
• Content and publications. The formula stayed the same: one topic produces two assets, a narrow article on the site for a specific query and a rewrite on an external platform (VC.ru, RBC Companies, Dzen). The tracker grew to 65 links. AI engines already pick up 53 of them, 210 mentions in total.
• Measurement and scaling. We moved to AI Visibility as the main measurement tool. It collects exactly the engines the client cares about.

Top pages by citations in October:

The leaders shifted from overview pages to commercial ones. Pages about choosing and buying moved up, and the funnel moved the same way.

Among external publications, three VC.ru pieces collect the most mentions. "Developing corporate web services" has 18 mentions, "GitLab in corporate development" has 15, and "LLMs in corporate processes" has 14.

The headline numbers:

• 71 citations of the site in AI answers on the full set of 425 queries;
• growth from 44 to 59 citations (+34%) in two months on a comparable 160-query funnel, plus 12 citations on new topics;
• Decision-stage coverage up from 5% to 22.5%;
• ChatGPT and Gemini cited the site for the first time;
• traffic up from about 165 to about 300 visits a month between May and October.

Why 71 and not 112?

In September we reported 112 citations, but a different tool counted them: Semantica, across nine engines. Now the measurement runs in AI Visibility on five engines with a different query set. The absolute numbers from the two tools cannot be compared. We measure dynamics inside one tool, which is why the growth rests on the pair 44 and 59.

Engine	Queries with a citation, 26 Oct	18 Aug	What changed
Alice (Yandex)	26	16	Main channel
Yandex AI Mode	24	not measured	New channel
ChatGPT	8 (9 links)	2	First working result
Gemini	7	0	Out of the blind spot
Google AI Overviews	4	0	Weak channel, but present
Perplexity	not measured	26	Dropped out because of a provider outage

The Yandex ecosystem is the backbone. Alice and Yandex AI Mode together account for 50 of the 69 queries with a citation, or 72%. ChatGPT linked to the site for the first time and cited it twice in one answer. Visibility no longer depends on one or two engines.

Here is how it looks inside the answers. Alice answers "How do I choose a mobile solution for corporate infrastructure?" and cites intellms.ru twice.

In a Google AI Overview, intellms.ru is cited four times and appears as the first card in the sources panel.

Perplexity is missing from this measurement because of an outage at the provider. In August it gave 26 queries with a citation. We will put it back into collection once the outage is resolved.

The most important change is at the Decision stage. This is where a buyer chooses a contractor and asks things like "what are the risks of ordering a corporate web service" or "how to choose SaaS for project management". Earlier, the site appeared in answers to only 2 of 40 such queries. Now it appears in answers to 9.

The drop in Awareness from 35% to 17.5% looks structural. Perplexity fed the top of the funnel, and it is missing from this measurement. We will test that assumption at the next measurement.

Site	Citations	Rank
Bitrix24	72	1
Intellms	71	2, level with the leader
Cleverence	60	3
Elma365	53	4
workspace.ru	38	5
KT Team	38	6
SAP	35	7
IT Integration	20	8

Intellms passed Cleverence, Elma365 and SAP. One citation separates it from Bitrix24. At the previous measurement Intellms was behind Cleverence.

Where do AI engines take their answers from in this niche? Legal-reference and government services have pushed user-generated platforms out of the top of the list. The leaders are consultant.ru (315), habr.com (247) and base.garant.ru (177). Habr and VC.ru, where we publish, hold their positions. intellms.ru itself is already among the eleven main sources. So for topics like electronic document flow and ERP migrations, publications next to such platforms are needed.

According to Yandex Metrica, visits to intellms.ru grew from about 165 in May to about 300 by 27 October. October was incomplete at that point. Two sources drive the growth: search (544 visits over six months) and referrals from other sites (222). The second source comes from our publications on VC.ru and other platforms. The same content that AI engines cite also brings human traffic.

• Comparing the two measurements. Semantica and AI Visibility numbers cannot be compared directly. We compare growth only inside AI Visibility.
• Small samples. Each funnel stage has 40 queries. One query moves a figure by 2.5 percentage points, so the Decision move from 2 to 9 queries is a strong signal, not a precise value.
• The missing Perplexity. Because of it, the Awareness drop cannot be considered final.
• Traffic. Metrica records a visit, not a citation. A person can read an answer with our link and not click, so traffic figures understate the real effect. Metrica does not prove a causal link between publications and visit growth.
• Leads. In this measurement we counted citations, funnel coverage and visits. The number of leads is not part of the case.

November plans include 9 topics and 18 materials on the same formula: one topic produces an article and an external publication. Goals for the November measurement:

• raise coverage on the new set of 265 queries from 2.6% to 8-10%;
• keep Decision at 20% or higher;
• return Perplexity to collection.

• Look at the funnel, not the total number of citations. A lead forms at the choosing stage, so that is where coverage should grow first.
• Write a narrow page for one buyer question and pair it with a rewrite on an external platform. That pairing worked for Intellms.
• Track dynamics with one tool. Otherwise a change of measurement system turns growth or decline into an artifact of method.
• Find out in advance which platforms AI engines already cite in your topic, and plan publications next to them.

Want the same for your niche? Get in touch and we will show you how AI engines answer queries in your market right now. For the method behind this work, read the first part of the case and the cases section on the Humanswith.AI website.

1. Humanswith.AI. Intellms: from 53 to 112 citations in a month (in Russian). https://humanswith.ai/ru/cases/intellms-ot-53-do-112-citirovaniy-za-mesyac/
2. Intellms. Company website. https://intellms.ru/