B2B companies and consultancies
Your buyers ask ChatGPT or Perplexity “which providers for…” and compare before they call. The assistant needs to know your offer, your area and your specialities, and must not confuse you with a namesake.
Found and cited by AI,
on solid SEO foundations.
Visibility in AI answers, or GEO (generative engine optimisation), means making sure ChatGPT, Perplexity, Gemini, Copilot and Google’s AI Overviews understand your business, describe it correctly and can cite your pages. Brutux Studio, a product studio in Thonon-les-Bains near Geneva, treats it as an extension of solid SEO: an audit of what the assistants say, verifiable fixes, follow-up with repeated test questions. No guaranteed citation: nobody can guarantee one.
What we actually know.
AI answer engines do not replace the web: they rely on it. Google’s AI Overviews and AI Mode draw on the Google Search index, Copilot on Bing’s, ChatGPT searches the web with its own crawler, OAI-SearchBot, and Perplexity with PerplexityBot. Google says so in its documentation for site owners: SEO best practices remain the foundation for its AI features, and no special optimisation is required to appear in them. Microsoft says much the same for Bing and Copilot: clear, well-structured content that answers the question directly. Google also warns against multiplying pages for every variant of a question and against artificial mentions.
In other words, there is no shortcut. A page that is not crawled, not indexed or not clear will not be picked up by an AI, whatever the trick. Conversely, a business that is well described on its own site, consistent wherever it is mentioned and cited by reputable sources stands a good chance of appearing correctly in answers. The work therefore looks like SEO, with three differences: we check what the assistants say, we take care of the company’s identity, and we measure differently.
What remains uncertain deserves saying too. Generated answers vary from one run to the next, from one tool to another and with the wording of the question. llms.txt files are not used by Google, according to its own representatives, and no major engine has confirmed using them to choose sources. We build the offer on what is documented, and flag what is not.
Who is this for?
GEO makes sense when your customers put their questions to an assistant before contacting you.
Local SMEs and shops
“Which tradesperson near me” questions also go through AI Overviews and the Google map. A complete Google profile, reviews and precise pages by service and by town remain the first lever here.
Businesses misdescribed by AI
An old name, a discontinued activity, an outdated address: when an assistant repeats false information, you have to fix the sources it reads, not the model itself.
Six AI visibility workstreams.
Each action rests on a recommendation published by Google, Microsoft, OpenAI or Perplexity, or on established SEO practice.
AI answer audit
Your customers’ questions, put to ChatGPT, Perplexity, Gemini, Copilot and Google: what is said about you, what is wrong, who is cited in your place.
Crawler access
robots.txt, firewall and indexing: OAI-SearchBot (ChatGPT search), PerplexityBot, Bingbot and Googlebot must be able to read your pages.
Structured data
Organization, Service, FAQPage: accurate schema.org markup on who you are, what you do and where. Useful to every engine, with no special pass for AI.
Entity consistency
The same name, description and contact details on your site, Google Business Profile, LinkedIn, directories and professional profiles, linked together with sameAs.
Content that answers
Service pages and Q&As written to answer buyers’ real questions in the very first sentence, with verifiable facts and your business clearly named.
Mentions and monitoring
The third-party sources assistants cite in your sector, honest steps to appear in them, and monthly monitoring.
What you get.
Dated findings, fixes actually made and a measurement method you can reproduce.
- A dated record of how ChatGPT, Perplexity, Gemini, Copilot and AI Overviews answer your test questions
- A list of false or outdated information and the sources spreading it
- A check of robots.txt, firewall and indexing, with recommended settings
- Structured data corrected or added on key pages
- A company identity sheet, to be copied word for word across all your profiles
- A content plan: questions to cover, pages to create or rewrite
- Monthly monitoring with the same test questions, to compare over time
Our GEO approach.
We measure before acting, fix what is documented, then monitor without over-reading.
- 01
Test questions
Together we list your buyers’ questions: “best provider for… in Geneva”, “alternative to…”. They become the measurement protocol.
- 02
Baseline
Each question is put to the main assistants, several times, on a fixed date: mention, wording, sources, errors.
- 03
Technical foundations
Crawler access, indexing, speed, markup, structured data: fixes come first, because a page an engine cannot read is one an AI cannot read either.
- 04
Entity and content
A harmonised identity sheet, updated profiles, pages rewritten or created to answer directly, outreach to third-party sources.
- 05
Monthly monitoring
Same questions, same tools, every month: we look at trends across several readings rather than reacting to a single answer.
What to know about GEO.
What is documented, what is not, and what we apply to our own site.
Crawlers: telling the roles apart
Each company runs several crawlers. At OpenAI, OAI-SearchBot powers ChatGPT search and GPTBot model training: you can allow the first and block the second. At Perplexity, PerplexityBot feeds the search index. At Google, the Google-Extended token covers Gemini and Vertex AI, not Search: blocking it removes nothing from AI Overviews. An overly strict firewall can also block these crawlers without anyone noticing.
Answers that vary
Ask an assistant the same question twice: the answer, and above all the sources cited, often change, as several recent studies observe. A single screenshot therefore proves nothing. We measure with a fixed set of questions, asked several times, in several tools, and read trends.
Third-party sources carry weight
Assistants cross-check: what your site says matters, but so does what local press, trade directories, associations, comparisons and reviews say about you. Microsoft and Google both recommend reliable, verifiable content, and Google advises against inauthentic mentions. The job is to earn mentions, not to manufacture them.
What is not proven
The llms.txt file is a proposal with no adoption confirmed by the major engines; Google has said it does not use it. “Magic prompts”, hidden text aimed at AI and pages written only for bots are a gamble at best and manipulation at worst, which Bing now addresses in its guidelines. No provider controls what a model will answer tomorrow.
What we apply at home
On brutux.studio, we apply these principles to our own site: Organization, Service and FAQPage structured data, a Q&A centre that answers buyers’ questions directly, pages in French and English, a robots.txt open to search crawlers, and llms.txt and llms-full.txt files published as a precaution, without presenting them as a proven lever. It is our testing ground, not a promise of results.
GEO and classic SEO.
The two share the essentials; what changes is mainly how you measure and how much care goes into the company’s identity.
| Criterion | Classic SEO | GEO (AI visibility) | Promises to avoid |
|---|---|---|---|
| Goal | Appear in results and earn clicks | Be understood, correctly described and cited in an answer | Be “recommended first” by ChatGPT |
| Foundations | Crawling, indexing, speed, markup | The same, plus access for AI search crawlers | A single file or tag that would be enough |
| Content | Pages that answer queries | Direct answers, verifiable facts, business named | Copy rewritten “for AI” or pages per variant |
| Reputation | Links and reviews | Consistent mentions in the sources AI cross-checks | Bought or fabricated mentions |
| Measurement | Rankings, clicks, Search Console | Repeated test questions across several assistants | A screenshot presented as proof |
| Guarantee | No guaranteed ranking | No guaranteed citation | Citation guaranteed within X days |
Frequently asked questions.
Visibility in AI answers (GEO): answers to the questions we are asked.
Can you guarantee we will be cited by ChatGPT?
No. Nobody controls a model’s answers, and they vary from one question to the next and from one day to the next. We work on what improves your chances, crawler access, clear content, a consistent entity, third-party sources, and we measure honestly what changes.
Does GEO replace SEO?
No, it depends on it: Google’s and Microsoft’s AI answers rely on their search index. “GEO” will not make up for a site that is poorly crawled or poorly ranked. Our offer therefore starts with an SEO check and works hand in hand with our SEO service.
How do you measure results?
With a fixed set of test questions, asked several times in ChatGPT, Perplexity, Gemini, Copilot and Google, on a regular schedule. We record whether you are mentioned, how and with which sources. Because answers vary, we read trends over several months, not a single reading.
Should we block AI crawlers?
It depends on what you want. Blocking GPTBot or Google-Extended limits the use of your content for training or for Gemini, with no effect on Google Search. Blocking OAI-SearchBot or PerplexityBot, on the other hand, removes you from ChatGPT or Perplexity search answers. We explain every setting before changing it.
Is an llms.txt file useful?
It costs almost nothing, but no major engine has confirmed using it to choose sources, and Google has said it does not use it. We publish one on our own site as a precaution, without counting it as a lever. It is never the priority.
Does structured data get a page cited?
Not on its own. It helps engines understand who you are and what you offer, and powers Google’s rich results. Google states that no special markup is needed for its AI features: markup simply has to be accurate and match the visible content.
ChatGPT says false things about our company: what can we do?
Find where the error comes from (an old profile, an outdated directory, an obsolete page, a namesake), fix those sources and publish clear information. The model does not correct itself on request, but answers grounded in web search gradually reflect up-to-date sources.
How long before we see an effect?
According to OpenAI and Perplexity, a robots.txt change is picked up within about a day. For the rest, it depends on how pages are crawled, the competition and third-party sources: think months, and we do not promise a timeframe.
Have you specialised in GEO for long?
GEO is new for everyone. Our foundation is technical SEO and the design of websites and content, and we apply these principles to our own site. We would rather say what is documented than sell a recipe.
Do you work in Switzerland?
Yes, throughout Switzerland, especially French-speaking Switzerland. We then add local.ch, search.ch and Swiss directories to the sources to harmonise, and test answers in your customers’ languages.
(Let’s work together)
Let’s talk about your project.
30 minutes to talk about your project, in Thonon or by video call. I’ll tell you honestly what will make the difference.
Book 30 minutes ↗