This case study on B2B SEO for IT covers an ongoing organic search (SEO & GEO), and content programme run for a European IT company with annual revenue above 35 million euros, a marketing department of roughly a dozen people, and a website in the hundreds of thousands-of-URLs range.
The mandate was not “rank for more things”. It was to turn a large, ageing content estate into a channel that produces pipeline, without hiring three more writers to do it.
Since the start of 2025 we have produced around 230 pieces between articles and pages for them, with, among others, deep analysis on AI performance, competitive landscape and optimization opportunities, and CMO-ready SEO strategies.
Year to date, SERP clicks from Google are up close to 173% year on year (approaching six figures) and impressions are up close to 263% (tens of millions). Below is what we built, and which parts of it mattered.
The starting point: volume without architecture
The company was not invisible. It had a blog, a reasonable domain profile, and a brand that people in its niche already searched for by name across several European markets.
What it did not have was a structure and a thorough optimization plan for production and revision.
Articles existed because somebody thought they were a good idea, not because they sat at a defined point in a buying journey, while honoring the algorithm rules.
Topic clusters overlapped, product pages were orphaned from the editorial content that should have been feeding them, and nobody could say which URLs were carrying the site and which were dead weight.
That is the normal condition of a mid-market B2B content estate. It is also why “publish more with AI” is usually the wrong first instruction.
What we built – semantic architecture and recurrent analysis
An intent architecture instead of a keyword list
Together with the existing SEO team, we rebuilt the editorial plan around search intent and buying stage rather than volume. Every planned piece had to answer 3 questions before it was commissioned: to which ICP is this related, which decision does this help someone make (or which intent does it match), and which commercial page does it hand the reader to.
That removed a surprising amount of planned work and focused our efforts on what counted, including optimization tasks, sales ancillary material, and ToFU content.
It also meant the surviving briefs carried internal linking instructions from the start, so editorial traffic had somewhere to go. Our B2B tech SEO consulting work almost always begins here, because no amount of production quality rescues a badly mapped content estate.
An AI-assisted production line with human checkpoints
Today, AI handles roughly 85% of the creation and production flow: research consolidation, outline generation, first drafts, metadata, schema, internal link suggestions, and translation into the additional market languages.
Humans still own the parts where being wrong is expensive. Subject-matter accuracy, first party data, strategic match, SEO cross-revision, topcial positioning, product claims, and final editorial judgement stay with people.
The AI layer is there to remove the eight hours of assembly work that sit around every good article, not to replace the person who knows what the product actually does.
The practical result is throughput. Around 230 articles in the period, across multiple languages, with a team that did not grow.
Technical and on-page groundwork
Production at that pace only pays off if the pages get crawled, rendered, and indexed. In parallel with the editorial work we ran continuous technical SEO cleanup: indexation hygiene, internal link distribution, template-level metadata, structured data, and hreflang consistency across markets.
Also, we started briefing content for passage-level retrieval as well as for classic ranking. Clear definitional openers, self-contained sections, explicit entity naming, comparison tables, and answers that survive being lifted out of the page.
This is the practical core of AI visibility work, and how it supports AI search tracking.
None of this is exciting. All of it is the difference between 230 published articles and 230 indexed, ranking articles.
Recurrent analysis and in-depth experiments
From the beginning of 2025 until today, we’ve been running recurrent analysis, among which:
- competitor gap (mainly on SERP positioning, content angles, SERP clicks, keyword volumes, topical & semantuc, backlinks, etc.)
- ICP fit (real commercial opportunities derived from organic a specific market share, etc.)
- AI visibility presence (AI bot hits from server log drain from all core LLM, fan out coverage, url retrievability, technical stability, topical and factual gaps, etc.)
The results
Numbers below are year to date and deliberately expressed as magnitudes rather than exact figures, since this is a live client engagement.
| Metric | Movement |
|---|---|
| Clicks from search | up around 173% YoY, approaching six figures |
| Impressions | up around 263% YoY, in the tens of millions |
| Brand queries | average position at or near 1 across eight core markets |
| Blog pages generating traffic | several hundred distinct URLs, including top 3 in SERP |
| Users arriving from AI assistants | up roughly 50%, thousands of users |
| Share of production handled by AI | about 85% |
Search performance
Impressions grew faster than clicks, which is what you expect when a site expands into a much wider query surface. The site is now being shown for categories of query it simply did not appear for before.
Clicks followed at a slower but still triple-digit rate, and brand demand held its position at the top of the page across every core market.
Non-brand discovery grew on top of a brand base that was already healthy, rather than cannibalising it.
Visibility inside AI assistants
Users arriving from LLM referrals grew by roughly half over the period, into the low thousands (ChatGPT is the dominant source by a wide margin).
The more useful detail is which pages get cited (not the homepage, but mid-funnel comparison and explainer articles, the ones written to answer a question completely inside a single section).
Retrieval rewards self-contained passages, and the referral data says so quite plainly.
Production economics
The programme runs at roughly ten times the article output the team could previously sustain, at a cost per piece that made the volume defensible in a budget conversation. That is the part most B2B marketing leads care about, and the part that rarely appears in case studies.
What makes the difference in B2B SEO for IT
Three things were game-changing for us, in order of impact.
Architecture before volume. The single highest-leverage decision was refusing to publish for six weeks while the content map was rebuilt. Everything produced afterwards had a job.
AI as an assembly layer, not an author. The 85% figure is only defensible because the remaining 15% is the expensive, judgement-heavy part. Teams that invert that ratio produce filler and then wonder why nothing ranks.
Treating AI search as a content problem. No separate GEO workstream was needed. Content briefed for clean retrieval performs in both classic search and AI answers, because both systems are looking for the same thing: an unambiguous, well-structured answer attached to a credible entity.
Can this approach for B2B SEO be replicated?
The scale is specific to this client, the method, we can replicate it.
The same architecture-first sequence, the same AI-assisted production line, and the same retrieval-aware briefing work at 500 URLs as well as they do at 5,000. What changes is how long and to which specificity the first phases take.
If you want to see the workflow layer that runs this kind of programme, you can contact us for a discovery: we will tell you honestly what’s stopping your B2B IT business from growing from an organic standpoint.