Your Next Client Won't Google You — They'll Ask ChatGPT for a Shortlist of 3

A finance director in Derby needs a new IT partner. Her last one dropped the ball on a ransomware scare, and she's had enough. Two years ago she'd have opened Google, typed "managed IT services Nottingham", scanned the map pack, and clicked three websites. This year she opens ChatGPT and types: "I run a 40-person accountancy firm near Nottingham. Recommend three managed IT providers who handle Cyber Essentials and Microsoft 365 for regulated businesses."

Within seconds she has three names, a sentence on each, and a reason to trust them. She never visited a single website. She never saw your carefully worded homepage. If your firm wasn't one of those three, you didn't lose on price or service — you lost before the conversation started.

This is the uncomfortable shift happening right now, and most IT firms are still optimising for a game that's quietly being replaced.

Why the shortlist matters more than the search

Search gave you a fighting chance. Ten blue links, a map pack, some paid slots — plenty of room to appear somewhere on the page and earn a click. A generative answer gives you nothing like that. When someone asks an AI tool for recommendations, they don't get a page of options. They get a handful. Often three. Sometimes one.

That compression is brutal. Being the fourth-best-known IT provider in your area used to mean steady enquiries. Now it can mean invisibility. The tool doesn't scroll, and neither does the buyer.

And here's the part that trips up most marketing budgets: the AI isn't reading your website to build that shortlist. Not primarily. Study after study of AI citations points the same way — the large majority of sources these tools draw on are third-party. Directories, review platforms, industry publications, forum threads, comparison articles, press coverage, and the general chatter of the web. One widely cited analysis put third-party sources at around 85% of what these models reference when they answer questions like our Derby finance director's.

Read that again. Roughly eight and a half times out of ten, the recommendation comes from what other people say about you, not what you say about yourself.

The homepage tweak that won't save you

So the instinct to rewrite your hero section, add another testimonial slider, or squeeze one more keyword into your service pages is understandable — and mostly wasted. Not because your website is unimportant, but because it's no longer where the decision is formed.

Think of it like a reference check. If you're up for a job and the employer rings three people who know your work, it doesn't much matter how good your CV looks. What those three people say decides it. Generative search works the same way. Your website is the CV. The rest of the web is the reference.

This reframes the whole problem. You're not trying to win a website game. You're trying to win a reputation game, played out across sources you don't own — and to make sure the machines can understand who you are clearly enough to connect the dots.

There are three levers that actually move the needle.

Earned mentions. These are the references — being written about, listed, reviewed, and discussed on sites other than your own. A guest article in a trade publication. A profile on a reputable MSP directory. A case study a client publishes about working with you. A mention in a "best IT providers in the East Midlands" roundup. Every one of these is a vote the AI can count.

Entity clarity. The AI needs to understand that "Cloudworks", the Nottingham IT firm, is a single, consistent thing — not confused with a dozen other businesses using similar words. Your name, address, services, and specialisms should read the same everywhere they appear. Inconsistency breeds doubt, and doubtful entities don't make shortlists.

Machine-readable signals. Structured data on your site, clean and consistent business listings, and content written in a way that plainly states what you do, for whom, and where. This is where your own website still earns its keep — not by persuading a human, but by feeding a machine unambiguous facts.

A 30-day starter plan for a UK IT firm

You can't fix all of this at once, and you shouldn't try. Here's a practical month, built for a small IT services business with a real workload and no marketing department.

Week 1: Find out what the machines already think

Open ChatGPT, Google's AI overviews, Perplexity, and Microsoft Copilot. Ask each the questions your buyers would ask. "Best managed IT providers in Nottingham for small businesses." "Who offers Cyber Essentials certification support near me?" "Recommend an IT support company for a 30-person legal firm in the East Midlands."

Write down who gets named, what's said about them, and where you appear — if at all. This is your baseline. Also note which third-party sources the tools cite. Those sources are your target list.

Week 2: Fix your entity foundations

Audit every place your business appears online. Google Business Profile, Bing Places, LinkedIn, Companies House, trade directories, review sites. Make the core facts identical everywhere: exact business name, address, phone number, website, and a one-line description of what you actually do. Add or correct structured data on your website so your organisation, services, and location are spelled out for machines. This is unglamorous work, and it's the load-bearing wall of everything else.

Week 3: Plant earned mentions

Pick three realistic targets. A regional business publication that takes contributed articles. A reputable MSP or IT directory. A local business network or chamber that publishes member profiles. Pitch a genuinely useful piece — say, what SMEs get wrong about Cyber Essentials — rather than an advert. At the same time, ask two happy clients if they'll agree to a named case study or a review on a platform that matters. Three earned mentions beat thirty homepage edits.

Week 4: Sharpen your own content and measure again

Rewrite your key service pages so they answer real questions in plain language: who you help, where, which certifications you hold, what problems you solve. Add a clear, factual FAQ. Then repeat the Week 1 tests. You won't have transformed your standing in a month, but you'll see movement — and you'll have a repeatable process to keep the pressure on.

The firms that adapt early will own the shortlist

None of this is theoretical. Buyers are already asking AI tools for recommendations, and the number doing so climbs every quarter. The firms that treat their reputation as something earned across the web — not something declared on a homepage — are the ones being named when it counts.

The finance director in Derby will never know how the shortlist was built. She'll just trust it, ring three firms, and hire one. Make sure yours is a name the machine already knows how to say.

If you'd like a hand running the Week 1 audit on your own firm, that's exactly the sort of thing we do at Cloudworks — get in touch and we'll show you what the machines currently think of you.