How Intellms got 112 AI citations in one month and entered Google AI Overviews

Intellms builds ERP systems, LLM solutions, and enterprise web services. The client wanted AI engines to name its site as a source when companies look for such a contractor. The second goal was a new lead channel beyond classic search: a person sees a link in an AI answer, goes to the site, and leaves a request.

By August 21, at the end of the first stage, the site had 53 citations, a link in answers to 15% of queries, and third place among the niche's domains. Details are in the previous case.

The second stage had three constraints:

• the site is young and has almost no history in AI results;
• Google AI Overviews, Yandex AEO, and Gemini were not measured in the August run;
• ChatGPT, Claude, DeepSeek, and Grok did not cite the site at all.

We built a pool of 160 LLM queries, 40 for each funnel stage: Awareness, Consideration, Evaluation, and Decision. Each query was run through 9 engines in Semantica AI, which gives 1,440 answers per measurement. We compared two measurements, August 21 and September 24, on the same query pool, broken down by engine and by stage.

The case uses three metrics:

• citation: an AI answer links to a page on intellms.ru;
• coverage: the share of queries whose answers contain at least one such link;
• answer share: the part of the 1,440 answers that contain citations of the site.

We wrote narrow pages on intellms.ru for specific queries and released rewrites on external sites: VC.ru, RBC Companies, and Dzen. In September we put the emphasis on Decision topics: choosing a SaaS platform, GitLab, and corporate messengers. By September 24, AI engines cited 24 pages of the site. The six most cited are in the table.

Each of the six pages answers a specific query, and each is cited by at least two engines.

In one month citations grew from 53 to 112, and 62 of them came from engines that were not measured in August. Google AI Overviews reached Alice's level in its very first correct measurement.

Alice went from 23 citations to 32, while Perplexity dropped from 30 to 18. Gemini produced its first hits. ChatGPT, Claude, DeepSeek, and Grok remain the next front of work.

The emphasis on Decision topics worked: coverage of the bottom stage grew from 7.5% to 15%. The top of the funnel grew too, while Evaluation dropped.

The Evaluation dip from 27.5% to 17.5% should be closed by the October block of topics, because in September we deliberately bet on Decision.

Besides the domain itself, we track all of the client's publications, including rewrites on external sites. By September 24, 53 of 65 such links were found in AI answers.

Fresh publications sometimes enter answers on the same day they are added to the tracker. A VC.ru article about measuring the results of a mobile platform pilot got two mentions right away.

Among the developers we track, Intellms leads. The site is linked in 7.8% of answers, against 3.4% for the closest competitor.

Of ten competitors, only Surf.ru has noticeable visibility. On Decision queries it almost caught up with Intellms: 13 citations against 14.

Over nine months, intellms.ru traffic grew from zero to 240+ visits a month, in parallel with the publication program. From January to September 2026 the site received 1,126 visits. Of these, 503 were direct, 417 came from search engines, 167 from links on other sites, and the rest from other sources (Yandex Metrica data).

A caveat: by default Metrica does not separate AI traffic into its own source. Visits from ChatGPT, Perplexity, and Alice land in "link referrals" and "search engines" together with regular ones. We will calculate the share of the AI channel from the referrer for the next measurement.

• October will bring 7 articles for the site and 2 rewrites. They will strengthen Evaluation (17.5%) while holding Decision (15%) and the top of the funnel.
• The goal for the end of October: the first citations in ChatGPT and more than 5 in Gemini. For that we are increasing external publications with links to intellms.ru.
• The next measurement runs at the end of October. The checkpoint: an answer share with a site link of 10% or more (7.8% now).

• Narrow pages for a specific query work. AI engines cite 24 pages of the site, and each of the six leaders is cited by at least two engines.
• Measurement without blind spots shows the full picture. Google AI Overviews gave 32 citations in its first correct measurement, as many as Alice.
• Publications on external sites widen reach. 53 of 65 tracker links produced 210 appearances in AI answers, while the domain itself has 112 citations.
• Growth is uneven. Perplexity dropped, and ChatGPT, Claude, DeepSeek, and Grok do not cite the site yet.

1. Humanswith.AI. How Intellms outranked every vendor competitor in AI answers. https://humanswith.ai/blog/how-intellms-outranked-every-vendor-competitor-in-ai-answers/