AI Engines · 9 min read · July 15, 2026
How your IT systems house gets named in ChatGPT and Perplexity recommendations
When a managing director asks ChatGPT "Which systems house can migrate our 40 workstations to Microsoft 365?", a language model decides in seconds which providers it names. For IT service providers, this AI visibility is the new referral source: whoever doesn't show up there simply no longer exists for a growing share of decision-makers – despite good Google rankings and full order books.
Why IT decision-makers today ask the AI, not Google
Your customers are tech-savvy – that's the decisive difference from many other industries. An IT manager, a CTO or a mid-sized managing director has long used ChatGPT, Perplexity and Microsoft Copilot in their working day. These very people ask the AI not just for PowerShell scripts, but also: "Which IT service provider in East Westphalia specialises in Sophos firewalls?" or "Who can support us with an ISO 27001 audit?" The AI's answer is a preselection for them that boils the market down.
The tricky part: the AI names three to five names, rarely more. There's no "page 2" like on Google. Whoever is missing from this short list doesn't even make it into the request for a quote. For a systems house that has so far lived strongly off referrals and local recognition, the stage thus shifts. The first contact happens before a human has ever seen your website – and before your sales team even gets a chance.
And something else is new: the AI answers with reasoning. It doesn't just say "Company X", but "Company X, because it specialises in mid-sized manufacturing businesses and employs certified Microsoft partners". The model draws this reasoning from what it finds about you on the web. This is exactly where Generative Engine Optimization comes in: you deliver the building blocks the AI builds its recommendation from.
How a language model arrives at a systems-house name at all
A language model doesn't invent recommendations out of nothing. It relies on training data and, with Perplexity and ChatGPT with web search, on live-retrieved sources. For IT service providers that means concretely: your website, your Google Business Profile, entries in directories like IT-systems-house lists, manufacturer partner directories from Microsoft, Sophos or DATEV, trade articles and reviews on ProvenExpert or Kununu flow together into one picture.
Structured, unambiguous signals weigh especially heavily. When Microsoft officially lists you as a "Solutions Partner for Modern Work" and that stands publicly in the partner directory, that's a robust fact for the AI. When your website clearly says "We look after tax firms with DATEV environments in the Nuremberg area", the model can assign this niche exactly. Vague phrasings like "innovative IT solutions for every need", by contrast, deliver nothing a model can hold on to.
Important to understand: the AI loves consistency. If you call yourself "Müller IT-Systemhaus GmbH" on the website, "Müller EDV" on Google and "Müller Netzwerktechnik" in the Sophos directory, your profile splinters. The model can't safely merge the building blocks into one provider and leaves you out when in doubt. A uniform name, one address, one core message across all sources is the basic prerequisite.
The niche is your strongest lever
Generalists have a hard time in AI recommendation. When someone asks "Who does IT for small businesses?", the field is huge and the AI reaches for the best-known brands. But when someone asks "Which IT service provider looks after dental practices with CGM software and handles KRITIS-adjacent data security?", the circle shrinks to a handful of providers. In these very specific questions you can win, even as a small systems house.
So consider honestly: which three to five questions should the AI answer with your name? Such as "Who migrates trades businesses from on-premise servers to the Azure cloud?" or "Which managed service provider offers 24/7 monitoring for production networks in Baden-Württemberg?". These questions are your target corridor. Every page, every article, every reference project should pay into at least one of them.
The mistake many systems houses make: they want to be found for everything and are thereby recommended for nothing. Dare to be pointed. A provider that clearly says "We're number one for DATEV and Microsoft 365 environments in tax firms between Cologne and Bonn" is named more reliably by the AI than one that advertises ten industries and all technologies at once.
Writing content an AI can quote
Language models prefer content that answers a concrete question directly and completely. Instead of a glossy landing page with "Your partner for digitalisation", you need pages that pick up real questions from your target customers. Example: a page titled "Microsoft 365 migration for tax firms: process, duration, DATEV specifics". In it you answer precisely how long a migration takes, what happens to the DATEV data-centre data and which downtimes are realistic.
Write in clear declarative sentences and put the most important information first. A model extracts "A migration of 40 workstations typically takes us four to six weeks" more easily than a flowery paraphrase. Use subheadings as real questions, work with bullet points for scopes of service and name technologies explicitly: Microsoft 365, Entra ID, Veeam, Sophos XDR, Datto, N-able. These technical terms are anchors for the AI on which it pins your competence.
Add concrete figures and facts wherever you can name them honestly: SLA response times, number of managed endpoints, years on the market, certifications with dates. Fact-dense, verifiable statements are preferentially quoted by AI systems, because they lower the hallucination risk. Vague superlatives like "industry-leading" the model ignores – or worse, it trusts your page less overall.
Technical foundation: structured data and clean pages
As an IT service provider you have a home advantage here that you should use. Add structured data to your website via Schema.org: the type "ProfessionalService" or "Organization" with exact name, address, service area, offered services and certifications. That makes it easy for crawlers and AI systems to absorb your profile in machine-readable form instead of having to guess it from running text.
Make sure central content is in the initial HTML and not loaded only via JavaScript. Many modern systems-house websites are based on heavy page builders where the actual text only appears client-side. Some AI crawlers render JavaScript only to a limited extent. A simple test: disable JavaScript in the browser and check whether your service descriptions are still visible. If they're missing, they're missing for the AI too.
Don't accidentally block AI crawlers in robots.txt. Check whether GPTBot, PerplexityBot, OAI-SearchBot and ClaudeBot have access. Whoever locks these bots out disappears from the web-search-based answers. That's a deliberate decision: visibility versus control. For most systems houses that want to win new customers, the visibility gain clearly outweighs it.
Third-party sources: where else the AI reads about you
Language models often trust independent sources more than your own website. That's why your presence outside your own domain is decisive. Maintain your Google Business Profile fully, with the correct category "computer service" or "IT service provider", current opening hours and real reviews. Ensure entries in relevant directories: the Microsoft partner directory, Sophos and DATEV partner lists, regional IT networks and industry portals.
Reviews are doubly valuable. They're a trust signal for the AI and at the same time deliver authentic phrasings that confirm your niche. When customers write "Switched our entire firm IT to Microsoft 365, smoothly and with genuine DATEV understanding", that's worth gold. Actively ask satisfied customers for reviews and to describe concretely what was done – not just "great company".
Professional visibility helps additionally: a guest article in an IT trade magazine, a talk at a regional digitalisation event, a mentioned case study at the manufacturer. Every independent mention that links your name to your niche reinforces the picture the AI assembles of you. It's not about mass, but about consistent, credible evidence.
Measuring whether it works: asking the AI itself
The big advantage over classic SEO: you can test the result directly. Regularly ask ChatGPT, Perplexity, Copilot and Google Gemini exactly the questions you want to be found for. "Recommend me three IT service providers for a Microsoft 365 migration in the Stuttgart area." Note whether you get named, in which position, and with what reasoning. Repeat this monthly and document the development in a simple table.
Watch for differences between the systems. Perplexity and ChatGPT with web search react quickly to new content, because they crawl live. Pure training-data answers lag months behind. So if you've freshly published a niche page, first check Perplexity to see whether it picks up your name. Also phrase your test questions sometimes without a local reference or with a different technology, to see how broadly your profile carries.
Also observe your competitors' mentions. Who gets named instead of you and why? Often the AI's reasoning reveals exactly the gap: "because they publicly show references in healthcare". That's your assignment for the next content round. This turns the diffuse goal of AI visibility into a concrete, iterative improvement process you can steer.
What you can tackle concretely this week
Don't start with a website reprogramming, but with the three target questions. Sit down with your sales team and write down which questions customers actually come with. Choose the three most lucrative niches and phrase them as concrete AI questions. That's the map for everything else and doable in half a day.
Then take on your existing website and build a dedicated, fact-dense page for each target question – or sharpen an existing one. In parallel, check robots.txt, add Schema.org data and unify your company name across Google, manufacturer directories and review portals. This tidying-up costs little and often works faster than new content.
And be patient but consistent. AI visibility isn't a one-off project but a discipline like patch management: regular, documented, measurable. Whoever starts now as an IT service provider has a real head start, because most competitors don't yet take the topic seriously. In two years the question won't be whether your systems house shows up in AI recommendations, but how prominently.
Common questions
Isn't it enough if my systems house ranks well on Google?
No, those are two different worlds. Google shows ten blue links, ChatGPT and Perplexity name three to five names with reasoning. A model draws its recommendation from training data and live-crawled sources, not from your Google ranking. Good SEO helps indirectly, because crawlable, clear content benefits both systems, but it guarantees no AI mention. You additionally need fact-dense niche pages, consistent third-party entries and machine-readable data so the AI can assign and recommend you reliably.
We're a small systems house with ten staff. Do we even stand a chance against the big MSPs?
Yes, especially on specific questions. For "Who does IT for mid-sized businesses?" the big brands win. But for "Who looks after dental practices with CGM software in the Kassel area?", specialisation counts, not size. If you clearly occupy your niche, with concrete references, matching certifications and reviews that confirm exactly this niche, the AI names you more reliably than a generalist. Small providers with a sharp profile are often at an advantage in AI recommendation, because they cover a precise question completely.
Should we allow or block AI crawlers like GPTBot and PerplexityBot in robots.txt?
For most systems houses that want to win new customers, clearly allow them. If you lock out GPTBot, OAI-SearchBot, PerplexityBot and ClaudeBot, you disappear from these systems' web-search-based answers. That's a pure loss of visibility without any return, as long as you don't publish content worth protecting. Check your robots.txt specifically for these user agents. A deliberate block only makes sense if, for legal or strategic reasons, you don't want to appear in AI answers, which is rarely the case for a sales-oriented IT service provider.
Read on