Your name pops up in an AI answer and still the phone stays silent. This article shows which part of your GEO budget actually brings in customers.
Getting mentioned is the first step, getting chosen is the goal
Have you noticed that an AI answer almost always gives several names?
As a business owner you hear it everywhere. Make sure AI mentions you. That advice is sound, but a mention is only the beginning. An AI answer often names five providers in the same breath. The reader then picks one.
That is where the difference lies. Getting mentioned brings visibility. Getting chosen brings an appointment, an order or a quote request.
2,000 mentions are worth little if they never ring the till.
I notice it in almost every first conversation. A business owner proudly tells me that AI recommends their company. My first question is always how many enquiries that has brought in.
So the question for you is sharper than how often you appear. The question is what convinces an AI assistant to recommend you over your competitor.
Take Fysiotherapie Smits, a physiotherapy practice in Den Bosch. Use Google's AI Mode to look for a good physiotherapist for a knee injury in the city. Smits comes third, with Knienetwerk Brabant (a regional knee network) and ultrasound as the reasons. Both are clearly stated on the practice's own knee page. Visible, then. You only get chosen once the answer puts you at the top.
Line up your spending with what you are judged on
Many businesses pay for one thing and are judged on another. The budget goes towards visibility, while management steers on enquiries, revenue and customer value. That gap costs money.
So start with your goal. If your business is judged on new customers, every GEO project deserves a clear line to new customers. Without that line, you spend time on something your own scoreboard ignores.
Pay for what you measure and measure what you pay for.
A machine has a say
Imagine a buyer asking an AI assistant for three suppliers in your region. Are you on that list?
Optimising for a machine is less new than it sounds. You already did it for Google's algorithm, for the social media timeline and even for the spam filter of a mailbox.
What is new is where it happens. AI assistants now sit right in the middle of the buying process. They compare, summarise and give advice, to consumers and business buyers alike.
This is moving fast in the Netherlands too. In 2025, 33% of Dutch businesses with 10 or more employees used AI, reports Statistics Netherlands (CBS). In 2023 it was still 14%. So your customer, your supplier and your competitor are all working with the same assistants.
My own experience with this is down to earth. I increasingly ask an AI assistant for advice before I buy something, from a laptop to an accountant.
So you are writing for two readers at once. The machine has to understand your offer, the person has to trust you. To win a new customer, you need both.
Foundation first, then more content
Ever heard of content that costs you more than it brings in?
When a team sees AI search growing, the same reflex often follows. Get dozens of extra articles made, fast. If an AI system struggles to find or read your existing pages, all that content just piles up as clutter.
The foundation rests on four pillars.
- Crawlability. AI crawlers visit and read your site, free of technical blocks.
- Structured data. Schema states exactly who you are, what you offer and which experts work with you.
- A clear entity. Search engines and AI recognise your brand and show the right details.
- Up-to-date information. Prices, opening hours and services are correct today.
Every page you publish needs maintenance afterwards. Prices change, services shift and figures go out of date. An AI assistant reads your old texts just as readily as your newest ones.
Ahrefs analysed 17 million citations. What AI assistants cite is on average a year newer than what appears in the regular search results, as that study shows. If you have a hundred quick articles made, you also have a hundred to keep up to date.
A marketing colleague of mine calls a website full of outdated blogs a cluttered attic. Everyone puts something there, nobody tidies up. That is why we clear out that attic first for every client. Only then do we add anything new.
A well-stocked site needs maintenance too. Fysiotherapie Smits has thirty condition pages under Specialisations, from osteoarthritis to fall prevention. When the reimbursement from health insurers changes, you check every text that covers it. Ten strong pieces are quicker to update than thirty loose ones.
How to lay that foundation step by step is explained in the article on the new search funnel.
Improving strong existing pages often pays off faster than writing a hundred new ones.
Why AI picks your answer
What do you know that your competitor has never written down?
Every provider in your industry publishes and many of those texts are interchangeable. So the question that counts is why an AI system would prefer your answer over a hundred pages that say the same thing.
An AI assistant likes to cite what is easy to pick up. Put the answer at the top. Let every key sentence stand on its own. Put comparisons in a table. Add experience that a model could never make up.
The strongest reason to cite you is something only you have. Your own figures, hands-on experience or a fixed way of working. Read how to build that up in the article on authority and AI mentions.
My favourite in this list is figures from your own practice. A small overview stands out more than a long article full of general tips.
The knee page of Fysiotherapie Smits explains meniscus injuries and knee osteoarthritis clearly. That explanation appears on plenty of other sites too. What AI finds nowhere else is a figure of its own, such as the average number of weeks of recovery for the practice's patients.
A mention and a citation are two different things
In reports, two terms often get mixed up. A citation is a link in the AI answer pointing to your website. You earn it with clear answers, your own data and up-to-date insights.
A mention means the answer recommends your brand by name. You earn it with trust and authority in your field.
So look beyond the AI referral traffic in your analytics. A good share of citations never leads to a click. Visitors who do arrive via AI also behave differently. According to Adobe they are 23% less likely to bounce and view more pages per visit.
A recommendation can bring you revenue, even when nobody clicks.
Ask Google's AI Mode which practice in Den Bosch has experience with running injuries. Fysiotherapie Smits comes first. The source label under that recommendation belongs to competitor Van Lith Healthness Center. Smits gets the mention, Van Lith the citation. Each says something different about your position.
Busywork and return often go separate ways
Some tasks feel productive because they keep the team busy. Rewriting pages, ticking off technical checklists, scaling up content. Being busy is a different thing from getting a return.
The tasks that actually set you apart often get the smallest budget. Think of your own research, a clear vision from your experts and mentions elsewhere. That is exactly where a brand wins in the long run.
Ask yourself one question for every task. Would a customer notice if this task stopped tomorrow? If the answer is no, chances are the work mainly provides reassurance internally.
Also take a critical look at tricks that seem to work fast. Mass-producing pages to win rankings falls under Google's policy against scaled content abuse.
What scores quickly today can eat your budget tomorrow.
The budget exercise. Imagine your budget shrinks tomorrow
A simple exercise shows what you really value. Imagine your marketing budget shrinks by 24% tomorrow. What do you cut first and what do you hold on to?
What remains are your real priorities. Chances are that mass-produced content goes first. Rightly so, because it delivers the least. The technical maintenance of your site often stays.
Do this exercise with your team or your adviser. The conversation often gives more insight than a whole dashboard.
Quick wins, building work and experiments
A good GEO plan divides the work into three buckets.
- Do now. Quick improvements, such as fixing crawl errors, allowing AI crawlers and updating your most important service pages.
- Build next. Work that takes longer and delivers more, such as your own research, a customer panel or mentions in trade media.
- Test carefully. Small trials on a new platform or with a new format.
A handy split is the 80/20 rule. Put about 80% of your resources into approaches that are proven to work. Keep 20% free to experiment. That way you grow steadily and stay curious.
At Fysiotherapie Smits, the specialisation of every therapist is already on the team page. So the quick step is done. An overview of the practice's own recovery times is the building work for later. A trial with short exercise videos can run alongside it.
Seven questions for every plan
Torn between two projects? Run them both past the same seven questions.
- Revenue. Does this bring in customers directly or mainly awareness?
- Opportunity. Do your customers use AI to compare providers like you?
- Room. Are your competitors already in the answers or is there an open spot?
- Existing authority. Are you already known in this field, so you can build on that?
- Effort. How much time does it take to carry out?
- Time to impact. How quickly do you see results?
- Measurability. Can you track success reliably?
Always weigh impact and effort together. A quick improvement builds confidence in the approach. A long project builds a lead that is hard to copy. Together they make you less dependent on a single source of visitors.
Measure GEO as a growth channel
Have you noticed that you yourself read the answer more often than you click through?
The classic click figure tells an ever smaller part of the story. In the United States, 68% of Google searches now end on the search page itself, according to research by SparkToro covering early 2026.
The same country comparison puts that share in Germany and France between 62% and 65%. Our own measurement for the Netherlands comes out at 63%.
Think of this path. Someone asks an AI assistant for advice and sees your brand in the answer. The tab closes. Three weeks later the same person sees your ad, recognises your name and books. Who gets the credit? Anyone who only measures the last click gives all the credit to the ad.
I notice that most dashboards are blind to this. A client gets more enquiries, while the traffic in the analytics stays flat.
If you only measure the last click, the influence of AI stays invisible.
Three layers, from visibility to revenue
So build your reporting in three layers and read them from the bottom up.
- Visibility. Are you mentioned and cited more often for the questions that matter to you? Split them by target group and by stage of the customer journey.
- Demand. Are more people searching for your brand name and are there more direct visitors?
- Business results. Are your qualified enquiries and your revenue growing along with it?
For the first layer, choose a fixed set of questions your customers really ask. Think ten to twenty, spread across orientation, comparison and choice. Ask them again every month and note who gets mentioned. That way you see movement, even though every AI answer is slightly different.
Classic figures such as rankings, traffic and click-through rate simply belong in there too. For most websites, Google remains the biggest source of visitors. The order of measuring is worked out further in the GEO playbook.
Ask your customer where they found you
The simplest measurement costs next to nothing. Add one question to your contact form. Where did you first hear about us?
Put those answers next to the figures in your analytics. The difference shows how much influence stays out of sight.
That gap can be large. One agency compared 620 enquiries side by side. The software gave search engines 78% of the credit, while customers themselves named a search engine in 12% of cases, according to the analysis by Refine Labs.
This is the measurement I recommend to every client. That one line on the contact form often tells you more than a month of analytics.
On the online appointment form of Fysiotherapie Smits, you choose a complaint, a therapist and a time. With that one question added, the practice sees straight away whether patients come in via AI, their GP or someone they know.
The ladder from found to chosen
How high are you on that ladder right now?
See GEO as a ladder with six rungs. Being found, being understood, being cited, being mentioned, being chosen and turning it into revenue.
Each step leans on the one before. A brand that AI understands poorly is rarely chosen. So focus on your weakest rung first. Once that is solid, take the next one.
Your weakest rung decides how high you climb.
Where does your next euro go
The answer depends on where you stand now.
- Hard to find? Fix the basics of your site first.
- Found but rarely cited? Improve the information on your pages.
- Does AI mostly name your competitors? Build authority beyond your own site.
Ask Google's AI Mode for the best physiotherapist in Den Bosch. The answer knows Smits well, right down to the ultrasound. Yet Fysio Moov comes first, with 333 reviews against 67. So the weakest rung for Smits is getting chosen. The next euro goes towards patient reviews and a listing in a regional health guide.
Comparison articles help in all three cases, because they show AI how you compare with others. If something works, roll it out across all your services and regions.
If you can only back one horse, a place in trade media and independent overviews often pays off most. It is exactly those sources that build the trust AI systems base their recommendations on. Want to know where your weakest rung is? A free opportunity analysis maps it out.
How to tackle it in four steps
Take your time for each part. Here, a steady pace pays off more than speed.
Step 1. Measure
Record where you stand today. Ask an AI assistant a few questions about your business, such as who you are and what you are known for. Note your mentions, your citations, those of your competitors and your most cited pages.
Step 2. Repair
Fix technical errors, make clear what you want to be known for and update outdated information. Improve important pages that perform poorly instead of deleting them.
Step 3. Build
Create something AI likes to cite. An overview of your own, a benchmark, a case study, a handy calculator or a fixed way of working from your experts.
Step 4. Test
See what works and do more of it. If an approach falls behind, find out why, adjust and try again. You can find help with AI visibility on our AI SEO page.
Frequently asked questions about GEO and budget (FAQ)
What is GEO?
GEO stands for generative engine optimisation. You make sure AI assistants and the AI answers in Google understand, cite and recommend your business.
How do I measure whether GEO pays off?
Measure in three layers. First your visibility in AI answers. Then demand, such as searches for your brand name. Finally your enquiries and your revenue.
What is the difference between a mention and a citation?
With a citation, the AI answer links to your website. With a mention, the answer names your brand as a recommendation. Both count, but for your revenue a recommendation often carries more weight.
Should I create more content for AI?
Improve what you already have first. Making strong existing pages more accessible usually gets results faster than publishing new articles en masse.
Which GEO tactic pays off most?
That depends on your weakest point. If you are hard to find, fix the basics. If you are found but rarely cited, improve your information. If AI mostly names your competitors, build authority beyond your site.
How much budget should I set aside for experiments?
A workable split is about 80% for proven approaches and 20% for trials. That way you keep growing and learn what new things work at the same time.
