Service · Building AI visibility
When your customers ask AI, the answer should name your company.
AI visibility is built in four work packages: a clear positioning, a website that AI crawlers can read, pages that answer customer questions concretely, and mentions in sources that AI systems cite. We measure from day one in Google AI Search, ChatGPT, Copilot and Gemini. That is how “not mentioned” becomes “recommended”, and this page shows how.
Four problems that decision-makers describe to us
Does one of these sound familiar? Then this page is written for you.
“The AI names competitors who can do less than we can.”
Because the AI does not judge who is better, but who it can clearly assign. An agency with a clear specialisation and target group gets named. A provider that does “everything for everyone” falls through the grid, even with a better product.
Go to positioning“Our website is good. But AI systems do not seem to know it.”
Because a website built for people can be invisible to machines. Content loads only via script, crawlers are locked out, company data exists only as an image. Then the AI reads empty pages and names the competitor whose page it could read.
Go to tech setup“We write a lot. Yet it is the competitor who gets quoted.”
Because AI systems do not cite texts, they cite answers. A page that answers a concrete customer question with figures, differentiation and an example gets adopted. A blog about trends does not. It is not about more content, but about the content your customers query the AI with.
Go to content“We get named, but never recommended.”
Because the AI only partly trusts your website. What gets recommended is what directories, reviews, trade media and partners confirm independently. When Google AI Search last named us, none of the cited sources came from lilox.io. We were named through what others write about us.
Go to external perceptionWhat this looks like in the measurement
Our own visibility and the measures that moved it
We do not sell anything we do not measure ourselves. The chart shows how often lilox.io is mentioned in AI answers to B2B buying prompts, before and after each measure, compared with five competitors. Click a measure to see the course up to that point.
Every AI system ranks differently
The same company, the same questions, five different rankings
Google AI Search, ChatGPT, Copilot and Gemini weight sources differently. That is why it is not enough to be visible “in AI”. You have to be visible where your customers ask. The ranking shows where lilox.io stood at the same five points in time as in the chart above, after each measure M0 to M4, with the movements per system. Competitors anonymised.
Ranking by share of answers with a mention, per system, each in the week of the measure, own measurement with Peec AI. Google AI Search appears twice because Google serves two answer formats. If lilox.io is outside the first six, the row shows the position below that; Google AI Mode was only measured from M3.
The implementation
Four themes we work on. No toolbox, but what you commission.
Each theme addresses one of the four problems above. You read what happens, what is needed from you and what you have in hand at the end.
Positioning
- Buying prompt analysis
- Competitor landscape
- Positioning statement
AI systems assign companies to a category: specialty, target group, region, proof.
“Innovative full-service agency” cannot be assigned, “B2B lead generation for IT service providers and software houses in the DACH region” can. So the starting point is not technology but a decision by management: what should your company be named for, and what deliberately not?
How we prepare the decision
- Identify the 20 to 40 questions your customers put to AI systems, from real buying conversations
- Name the competitors who are named for them today
- Formulate a positioning statement with fixed building blocks that then reads identically everywhere: on the website, in directories, on LinkedIn, in trade contributions
This consistency is no stylistic device. It is the reason an AI assigns you to a question with confidence instead of guessing. For us, this exact sentence led ChatGPT to describe lilox.io as “predictable lead generation for IT service providers, software houses, systems integrators or SaaS”, word for word from our industry page.
Tech setup
- AI visibility check
- Crawler access
- Structured data
- llms.txt
Before an AI can name you, it must be allowed to enter and read your website.
This sounds self-evident and in practice fails at ten points, for example:
- Crawlers from OpenAI, Google or Microsoft are blocked by an old rule
- Content only appears after a script or a click
- Company data exists only as an image
- Structured data that makes the company, services and contacts machine-readable is missing
Our AI visibility check measures exactly these ten criteria and provides the baseline. The fix is seldom elaborate, but it has to be complete. We work on it with your web agency or implement it directly: crawler rules, loading behaviour, structured data according to Schema.org, an llms.txt that tells AI systems what is on the website, and consistent company data on every page.
For us, the relaunch in August lifted visibility from 6.6% to 8.9% within two weeks, without much change to the text. The AI could simply read everything for the first time.
Content generation
- Page structure
- Core pages
- Proof
- Citation report
AI systems do not cite texts, they cite answers.
When a buyer asks “Which provider for X suits a company of our size in Austria?”, the AI looks for a page that answers exactly that: with target group, differentiation, figures and an example.
How we proceed
- From the question list of the first theme, a page structure emerges in which each core page answers a concrete customer question
- Existing pages are rewritten rather than replaced, and receive what makes content citable: quantities, durations, industries, references
- We write the pages, you release the proof
- Each week we check which page is cited in which system, so the next text starts where it has an effect
The hard part is not the writing but the release. What is self-evident in a sales conversation is often nowhere written down, because nobody has decided it may be public. Every released proof point is a sentence the AI can say about you.
Backlinks and external perception
- Source register
- Directories
- Reviews
- Trade articles
The difference between “mentioned” and “recommended” is made outside your website.
An AI recommends those whom several independent sources confirm with the same specialisation: industry directories, review platforms, trade media, associations, partner sites, map profiles. When Google AI Search last named lilox.io as a B2B agency for Austria, all the cited sources came from directories and third-party sites, none from our website. And when ChatGPT built a shortlist for IT companies, the first five places came from map profiles.
How we proceed
- First a source register: which sources do AI systems cite for your questions, where do the competitors stand, where are you missing?
- Bring your company consistently into these sources, with the positioning statement from theme 1
- Verified customer reviews
- Trade articles published under the name of your management
This is the slowest of the four themes, and the one that secures the biggest lead, because it cannot be caught up in weeks.
Two real AI answers from our measurement
What a mention reveals about the work behind it
Both answers are real and come from our ongoing monitoring. Competitors are anonymised. They show which theme worked, and what is still open for us too.
- LILOX: digital B2B lead generation for IT, tech and industrial companies
Google names four specialised B2B and tech agencies for Austria, with lilox.io in third place. The route there ran through theme 4: Google did not find us via our website, but via the directory profiles and what other sites write about us. That is exactly how a recommendation arises.
- Additionally recommended: LILOX, “predictable lead generation for IT service providers, software houses, systems integrators or SaaS”
Positioning and content, themes 1 and 3. ChatGPT read our industry page and adopted the target group word for word. That is exactly how a core page should work.
What is still openThe five places in the main list came from map profiles with a location. Whoever is listed there with a complete profile is named first. For us as for you, that belongs to theme 4.
How the collaboration works
Four work packages and one report per month
Your effort: one kickoff, three approval rounds, 45 minutes per month. The packages partly run in parallel, the measurement from the start.
Positioning and starting point
- Kickoff with management, marketing and sales
- Question list from real buying conversations
- Competitor landscape per question and system
- Positioning statement for approval
Tech setup
- AI visibility check of all main pages
- Crawler access, loading behaviour, structured data
- Implementation with your agency or directly
Content generation
- Page structure per customer question
- Write or rewrite five to eight core pages
- Collect and integrate proof
External perception
- Source register: who cites whom
- Directory, map and review profiles
- Trade articles and partner mentions
Monthly report
Six key figures per AI system, defined below. Comparison with the competitors observed, a priority recommendation for the next month, 45 minutes of discussion. One figure for your management meeting and one decision that is due.
The six key figures in the monthly report, defined
- Visibility
- Share of the AI answers to your customer questions in which your company is mentioned. 10% means: in every tenth answer.
- Share of voice
- Your share of all mentions of all providers observed. Shows how much of the conversation you have, not just whether you appear.
- Position
- The place where you are named, when you are named. 1 is the first mention. Decisive, because buyers rarely read beyond third place.
- Citation rate
- How often your website is linked as a source in answers. Measures whether your pages are citable, regardless of whether your name appears in the text.
- Retrieval rate
- How often AI systems read your pages at all before they answer. If it is zero, it is a case for the tech setup, not for the text.
- Tone (sentiment)
- How positively AI systems describe you, 0 to 100. Below 50 means: named, but not recommended. Proven results raise the value, performance promises do not.
What lies with you
Three decisions that no agency can make for you
What do you want to be known for?
AI systems assign companies to a category: specialty, target group, region. “Full service” cannot be assigned. That is a management decision, not a marketing question.
Which proof may become public?
Unit numbers, project durations, customer names, reviews. What is self-evident in a sales conversation is often nowhere written down. Every released proof point is a sentence the AI can say about you.
Who owns the number?
AI visibility works when it is a key figure in reporting, not a project of the marketing department. Decide who sees it monthly and who draws decisions from it.
Clear limits
What we do not do, so you know what you are commissioning
No position guarantee
Nobody can guarantee a mention in Google AI Search or ChatGPT. We guarantee the work in the four themes and transparent measurement. Anyone promising first place has not understood the mechanics.
No bought or fake proof
No invented reviews, no comparison articles in which we rate ourselves or competitors. What AI systems are to say about you must be true and verifiable.
No replacement for sales
AI visibility puts you on the list of providers that a buyer takes from the AI. The call afterwards is warm instead of cold, but you have to lead it yourself. Anyone who wants to turn the list into appointments combines this theme with our LinkedIn and email lead generation.
What it costs
AI optimisation alone from €1,900 per month. Integrated into ongoing marketing support from €2,900 per month.
You pay for work packages and measurement, not for promises. The entry is free, the exit is possible monthly. How high the amount ends up depends mainly on one question: On which technical setup do we implement?
AI visibility check
Online check of your home page or advanced check of the main navigation, plus a 30-minute conversation about your result and your customer questions. No obligation.
€0
Four work packages plus measurement
Positioning, tech setup, content, external perception, daily measurement in four AI systems, monthly report. AI visibility only, nothing else.
from €1,900 / month, cancellable monthly
AI visibility as part of marketing support
The four work packages run within the ongoing support of website, content and campaigns. Marketing as a service: AI visibility is one component of it.
from €2,900 / month, cancellable monthly
The implementation plan sets the price
Every optimisation has to reach the website: new answer sections, structured data, crawler rules, corrections after the monthly measurement. How quickly and how cheaply that works depends on who builds the change and on which system. That is why the tech setup is so decisive: it determines whether we stay agile for optimisations or every change becomes a project.
Astro, we implement
We build changes in directly, preview for approval, live on the same day. Shortest cycle between measurement and correction, lowest ongoing effort.
WordPress, with your agency or editorial team
Changes run via editorial staff or plugins, technical corrections via your agency. Works, but costs more time and coordination per optimisation.
Framer, only with Framer expertise in the team
If you have a Framer expert or work with AI and MCP in the system yourselves. Without that expertise we no longer recommend Framer.
You implement it yourselves
We deliver analysis, specifications and texts, your team builds them in. The cheapest plan if someone internal has time and access, otherwise the slowest.
For a change of setup without a new design there is the website migration: same content, same design, new technical foundation.
Questions from management meetings
What decision-makers ask us before they commission
How does building AI visibility work in practice?
In four work packages: positioning, tech setup, content generation, external perception. Measurement in Google AI Search, ChatGPT, Copilot and Gemini runs from the first day, and each month you receive a report with five key figures and a recommendation. The packages partly run in parallel, and the first new mentions typically come after four to eight weeks.
How long does it take until we are named?
The tech setup takes effect within days to a few weeks, once AI systems have re-read your pages. Mentions from third-party sources take two to four months, because directories, reviews and trade articles need time. For us, the path from 5% to 10% visibility took around three months.
What does building AI visibility cost?
AI optimisation alone from €1,900 per month, integrated into ongoing marketing support from €2,900 per month, each cancellable monthly. What comes on top depends on the implementation plan: if we implement on Astro ourselves, it runs via WordPress or Framer with your agency, or your team builds the changes in. The AI visibility check beforehand is free.
How do you recognise a good GEO agency?
By whether it measures its own AI visibility and discloses it, whether it shows mentions in real buying prompts instead of rankings, and whether it says which sources it intends to influence. We have put our selection criteria, our approach and our proof together on a separate page: GEO agency for B2B.
GEO agency for B2BWe already do SEO. Is that not enough?
SEO makes sure Google finds and assesses your pages. AI systems additionally need directly citable answers and independent confirmation from third parties. Good SEO is a prerequisite, but it does not create a mention in an AI answer. Our example from Google AI Search shows it: named via directories, not via our own website.
Our competitors are already being named. Has the train left?
No, but it is speeding up. The source register in the first work package shows exactly which sources the competitors draw their mentions from. In most cases these sources are just as reachable for you. The lead can be caught up, as long as it is not years.
Can we do this in-house?
Parts of it, yes, especially the release of proof and the maintenance of profiles. The measurement across several systems, the source register and the rewriting of core pages are specialist work that rarely pays off for a single company to build up. We work in a way that afterwards your team knows what it can continue itself.
How does this differ from the AI visibility check and the report?
The check shows the technical current state of your website in two minutes. The report provides the data basis to read up on. This page describes the service: how the current state turns into mentions.
AI visibility checkThe first step takes 30 minutes and costs nothing
In the AI visibility check we go through together how Google AI Search, ChatGPT and Copilot read your website today and who is named for your customer questions. Afterwards you know whether and where to start.
