Visible in Google — and in AI answers
Technical optimization, local findability and visibility inside AI answers. Measured as four separate signals, not as one invented score.
Free AI access checkThe blind spot
The major AI crawlers do not execute JavaScript — only Gemini renders, because it runs through Googlebot. A modern, client-rendered website can rank perfectly well in Google and still be an empty page to ChatGPT. That is not a ranking problem but an access problem — and it takes minutes to check.
What is SEO?
SEO makes sure a search engine can find your pages, read them correctly and place them properly.
The acronym stands for search engine optimization. It is not a trick that buys a ranking. It is craft: building and maintaining a website so that its structure is unambiguous, its content answers the questions people actually ask, and other sources treat it as reliable.
Most of the work is invisible. It concerns redirects, canonical addresses, language markup, status codes and which pages end up in the index at all. Nobody notices these things while they are correct — and nobody notices when they break.
Three layers
Technical
Can the page be fetched, rendered and indexed? Does it exist once, or simultaneously under three addresses?
Content
Does the page answer the question someone actually types — completely, without requiring another click?
Authority
Do other credible sources point to you? Is your organization mentioned outside your own website?
Why it matters
- For most organizations, search is the largest channel through which new people encounter them at all. If you are not there, you do not exist for those people.
- Technical faults are silent. A site can lose visibility for months while nothing about it looks broken.
- Relaunches are where it gets expensive. In one mandate we found 53 dead addresses carrying 28 percent of all impressions — on a domain whose crawl reported four issues. A crawler only follows links; it cannot see orphaned legacy addresses.
What is GEO?
GEO covers two questions: can people find you where you operate — and will an AI system name you when it answers?
We keep the two apart deliberately, because they work differently and are measured differently. GEO-1 means geographic, local findability. GEO-2 means generative engine optimization: visibility inside the answers produced by ChatGPT, Gemini, Perplexity, Copilot or Google's AI summaries.
GEO-1 — local findability
Someone searching nearby is searching with intent. Whether you appear on the map and in the local result block depends on unglamorous things: identical business details everywhere, clean LocalBusiness schema, a maintained business profile, and keywords tracked per city and per language version. For an organization with a location or a regional catchment, this is the channel with the highest conversion rate.
GEO-2 — visibility in AI answers
A growing share of questions is answered without anyone opening a website. If your brand does not appear in that answer, there is no click left to compete for. These systems access pages differently from a human: they fetch the raw HTML, generally do not execute JavaScript, and favour passages that stand on their own and support their claims.
Why it matters
- The answer is replacing the result list. Visibility then no longer means position three; it means being in the answer, or not.
- Mentions outside your own website carry at least as much weight here as classic links — on current evidence, rather more. Exactly how much varies by study.
- The most common reason a good page is not cited is mundane: the system could not read it. That is testable, and it is fixable.
Why the two belong together
SEO and GEO draw on the same source. An unambiguous structure, clearly named entities and sourced claims help a search engine classify and an AI system select. Do one properly and most of the other is already done.
The order still is not arbitrary. Technical first, then local, then generative. As long as a page is reachable under several addresses or sits behind redirect chains, no system can decide which version is authoritative — and every signal built on top of that goes nowhere.
What is actually measurable about AI visibility
There is no credible single number for AI visibility. We measure four things separately and report them separately.
Access
Can the crawler read the content at all?
Served-HTML check using a crawler user agent
Referral
Does traffic from AI assistants actually arrive?
GA4 AI assistant channel, plus a custom Perplexity group
Citation
Does the page appear in AI answers?
Bing AI Performance; the Google report once available for Switzerland
Representation
How is the brand described when it is mentioned?
A documented prompt sample
We do not add these four up into one number. Doing that sells precision that does not exist: the same brand swings considerably on the same day depending on account and surface.
What we do — SEO
Ordered by effect, not by effort. Critical first, cosmetics last.
Technical analysis
A full crawl of your domain, findings sorted by severity and stripped of noise — rather than a list of thousands of lines about build files.
Canonicalization and redirects
One host, one canonical address per page, redirect chains collapsed to a single hop. Usually the single biggest lever.
Multilingual correctness
hreflang with mutual references and x-default, canonicals self-referencing per language. An EN page that canonicalizes to the DE version will not be indexed.
Indexation
Sitemap, robots.txt, and a sweep against the Search Console pages dimension — that is where orphaned addresses no crawler can find become visible.
Titles, descriptions, structure
Length, uniqueness, heading hierarchy, thin content. Unglamorous, but it decides your click-through rate.
Structured data
Organization, Article, BreadcrumbList — validated by parsing, not by text search. A text match once reported a footer link to us as valid schema.
Loading behaviour
Core Web Vitals in the lab and in the field: LCP under 2.5 seconds, INP under 200 milliseconds, CLS under 0.1.
Keyword and content architecture
Cross-checked against real queries from Search Console rather than estimated lists. For new builds it is settled before the first line of code.
Relaunch protection
Redirect map before go-live, a checklist on staging, a follow-up sweep for dead legacy addresses. This is where most value is lost.
Ongoing checks
Regression check per deployment, rank tracking to position 20, before-and-after against the baseline.
What we do — GEO
GEO-1 and GEO-2 run separately, because they have different levers and different measures.
GEO-1 — local
Unify business details
Name, address and phone identical on the website, in the schema and in directories. Divergence is the most common reason an entry is missing from the local block.
Local schema
LocalBusiness and Organization with address, area served and geo coordinates — complete rather than half filled in.
Business profile
Reconciling what the profile says with what the website claims. Contradictions cost trust with humans and machines alike.
City-level tracking
Keywords per city and per language version, so that effect becomes visible instead of disappearing into a national average.
Local landing pages
Where the content justifies it: separate pages per service and location — with real content, not a template with the place name swapped out.
GEO-2 — generative
Access check
We fetch your pages with a crawler user agent and without JavaScript, then compare what remains. If the content is missing from the served HTML, everything else is pointless.
Crawler rules
robots.txt with a deliberate split: bots that serve live answers, and bots that collect training data. Blocking both wholesale costs you visibility, not training data.
Citable structure
Question-and-answer blocks that stand on their own. Answer first, reasoning after, evidence included. Short sentences, because answer systems quote from them.
Unambiguous entities
One consistent name, a supported organization markup, verified profile references. Invented references do more harm than missing ones.
Structured data for machines
Schema as a comprehension and trust signal, not as decoration for search results. We no longer sell FAQ markup as SERP real estate — that display was withdrawn in May 2026.
Setting up measurement
Bing AI Performance and the GA4 channel for AI assistants including a custom Perplexity group — in your accounts, not ours. Google's report joins once it reaches Switzerland.
Prompt sampling
A fixed set of questions, asked on a schedule, with language, account state and timing on the record. It is a sample and is reported as one — never as coverage.
Mentions beyond your website
Where else your brand appears counts at least as much for AI answers as links do. We capture the current state and name which sources are realistically reachable.
Two entry points
Track A — new build or relaunch
Keyword architecture, content structure and redirect map are settled before the first line of code. That is the point where this work is cheapest.
Track B — existing website
Capture the baseline, analyse, prioritise, implement, measure the effect. We change nothing before the starting point is documented.
How it runs
Seven phases. Each with a deliverable and a sign-off.
- 0
Goals and baseline
Shared objectives, access, a measured starting point.
- 1
Analysis
Technical health, local findability and AI access — de-noised into one backlog.
- 2
Roadmap
Prioritised by impact ÷ effort, in understandable blocks rather than an error list.
- 3
Sign-off
You decide priorities and scope. Decisions are put on the record.
- 4
Implementation
Fixes in your website's code — by us or your team. Verified live, not merely committed.
- 5
Effect
Before and after against the baseline. Regressions are named, not buried.
- 6
Iteration
Schedule what is left, plan the next round.
What you get
- A measured baseline before anything is changed
- An access report: which search and AI crawlers can read your content — and which cannot
- AI reports set up inside your own accounts, so the data belongs to you
- A readable summary per round: goals, baseline, measures, progress, next steps
- A roadmap prioritised by impact ÷ effort, quick wins first
- Task and decision cards with a traceable history
- A before-and-after delta at the end of every round
- Implementation in code — by us or your development team
Data sovereignty
Crawling and rank tracking run on our own infrastructure. The only things that leave it are services you already use.
External services
SERP lookup · Google Search Console · GA4 · PageSpeed and CrUX · Bing Webmaster Tools · IndexNow
No US-hosted AI visibility tool without a transfer impact assessment. If a dashboard is wanted, it will be one operated in the EU.
Honest limits
- Citation rate inside AI answers is not fully measurable. The providers do not release their answer logs. What we deliver is a documented sample — and we call it that.
- No guaranteed rankings and no guaranteed AI citations. Anyone promising those is guessing.
- We report rankings down to position 20. Below that, since long result pages were switched off, it is expensive noise. Time series crossing September 2025 are flagged as a break rather than smoothed over.
- Deployments and external approvals stay with you. We provide analysis, prioritisation and — on request — the implementation.
Frequently asked questions
What is the difference between SEO and GEO?
SEO optimises for the classic search index. GEO-1 is local findability, GEO-2 is whether an AI system can read, understand and cite your page as a source.
Can you measure how often ChatGPT cites us?
Not fully — the providers do not publish their answer logs. Four things are measurable: access, referral traffic, the first reports from Google and Bing, and a documented prompt sample. Anyone selling you a single AI visibility number is selling an extrapolation.
Does an llms.txt improve rankings in AI answers?
No. Google states it ignores the file for ranking and visibility, and the vast majority of published files are never requested. We create one where it serves documentation — we do not sell it as a ranking measure.
We rank well in Google. Is that not enough?
Not necessarily. Most AI crawlers do not execute JavaScript. A page that assembles its content in the browser may rank well and still be empty to those systems.
Where is our data held?
Crawl and ranking data on our own infrastructure. The only external calls are the SERP lookup and the Google and Bing interfaces you already use. We do not deploy a US-hosted tool without a transfer impact assessment.
Who implements the fixes?
Either us, in your website's code, or your own development team. We deliver the prioritised roadmap; deployments and approvals stay with you.
Is this worth it for a site that does not exist yet?
That is when it pays off most. Keyword architecture, content structure and a redirect map before the build cost a fraction of repairing them afterwards.
Free AI access check
We check whether the major AI crawlers can read your content at all, and which AI reports are already available for your domain today. The result is one page, with no obligation.
Request the check