AI Visibility · AEO/GEO
So the agent mentions your name.
More and more often, a customer asks an AI assistant first and only then calls. We make sure your business shows up in the answers: in ChatGPT, Gemini, Claude, Copilot, Perplexity and the other major systems. Proven, not claimed.
In short
What is AI visibility?
AI visibility means that AI assistants mention your company when someone asks for what you do. For that, your business has to be unambiguous, readable and verifiable for machines: on your own website and on the pages the systems trust.
- AEO
- Answer Engine Optimisation. Aligning content and technology so that an answer engine names a company in response to a specific question.
- GEO
- Generative Engine Optimisation. Umbrella term for optimising for generative AI systems that write a new answer from several sources, choosing and linking those sources as they go.
- Goal
- Being part of the answer, not just an entry in a list.
- Systems
- ChatGPT, Gemini, Claude, Microsoft Copilot, Perplexity, Grok, Meta AI, Le Chat by Mistral and DeepSeek.
- Status
- October 2026, checked against the providers’ documentation.
The situation
The research is moving elsewhere.
Someone looking for a supplier describes the problem to an AI assistant and gets two or three recommendations. That is not a forecast. It is already happening.
Answers instead of links
The assistant summarises and names two or three companies. There is no second page.
Facts instead of ads
AI assistants prefer verifiable information: numbers, locations, services, results, sources.
The lead is up for grabs
In many technical niches the answer is still open. Whoever starts early has the better cards.
Foundation
Three crawlers, three jobs.
Most providers send not one crawler but three. Mixing them up locks you out of the answers without you noticing.
- Search
- Builds the index the assistant draws its answers from and links its sources to. Block it and you will not be mentioned.
- Fetch
- Retrieves a page the moment a user asks about it or pastes a link. Block it and you stay unreadable in that conversation.
- Training
- Collects text for future models. Block it and you stay in search but are missing from future models’ background knowledge.
The nine systems
Every assistant reads differently.
The assistants do not write from memory. They search live and summarise what they find. If you do not get into their search index, you do not get into the answer.
| System | Searches via | Crawlers | What matters |
|---|---|---|---|
| ChatGPT (OpenAI) | its own search index, plus signals from Bing | OAI-SearchBot (search), ChatGPT-User (fetch), GPTBot (training) | Access for OAI-SearchBot, indexing at Bing, content in the delivered HTML |
| Gemini (Google) | the Google search index, queried live when needed | Googlebot (index), Google-Extended (switch for training and Gemini answers) | Indexing at Google, do not block Google-Extended |
| Claude (Anthropic) | its own search index | Claude-SearchBot (search), Claude-User (fetch), ClaudeBot (training) | Access for Claude-SearchBot, cleanly structured HTML |
| Microsoft Copilot | the Bing index | Bingbot | Indexing at Bing, report new content instantly via IndexNow |
| Perplexity | its own index, cites every source | PerplexityBot (index), Perplexity-User (fetch) | Access for PerplexityBot, current pages, clearly supported statements |
| Grok (xAI) | the live web and posts on X | no reliably documented crawlers | Presence on X with links to your own content, current pages |
| Meta AI (Meta) | its own search index for Meta AI | Meta-WebIndexer (search), meta-externalfetcher (fetch), meta-externalagent (training) | Access for Meta-WebIndexer, which Meta names as the condition for being cited |
| Le Chat (Mistral AI) | its own search index | MistralAI-Index (search), MistralAI-User (fetch), MistralAI-Training (training) | Access for MistralAI-Index and MistralAI-User |
| DeepSeek | web search when needed | no official crawler documentation | Presence on well-indexed, trusted pages |
Status October 2026. Providers change their crawlers and rules all the time; we check them again before every baseline. Where a provider documents nothing, we say so rather than claim something.
Approach
Five steps.
01
Baseline
We ask the assistants your industry uses the questions your customers ask and document who gets mentioned and which sources the systems link to.
02
Sharpen the entity
Name, address, services and numbers identical everywhere, on your own site as in directories, profiles and trade media. Ambiguity is one of the most common reasons for not being mentioned.
03
Open up the technology
A robots.txt that allows the search and fetch crawlers of all assistants, content in the delivered HTML instead of loaded later, structured data following Schema.org, indexing at Google and Bing, IndexNow. We also check whether a firewall or the hosting quietly turns the crawlers away.
04
Write to be quoted
Clear answers to real questions, concrete numbers, named sources and a visible date. What an AI cannot quote is of no use to you.
05
Measure and adjust
The same questions every month, the same evaluation. What works, we expand; what does not, we drop.
What the evidence says
What works, and what is overrated.
In a study by Princeton and IIT Delhi (KDD 2024), the visibility of sources in generative answers rose by up to 40 per cent when texts contained source references, quotations and statistics. Marketing speak and keyword lists did not help.
Works
- Concrete numbers, sources and quotations in the text
- Identical company data everywhere
- Mentions on pages the systems trust
- Access for the search crawlers of all assistants
- Current pages with a visible date
- Clear structure: headings, lists, tables
Overrated
- llms.txt as a cure-all: not a standard, no provider commitment
- Chopping text mechanically into answer blocks
- Stacking keywords: did little in the study
- Guaranteed mentions: they do not exist
Scope
What you get.
- Baseline
- A record of your customers’ questions, the assistants’ answers and the sources they link to.
- Entity
- A check of your company data across website, Google Business Profile, directories and trade portals, with a list of discrepancies.
- Technology
- robots.txt for all search and fetch crawlers, Schema.org graph, indexing at Google and Bing, IndexNow and a check whether the hosting turns crawlers away.
- Fact page
- A page of its own with the verified facts about your company, written for AI systems. This is ours.
- Content
- Rework of your most important pages: answers first, numbers and evidence, questions and answers in your language.
- Measurement
- A monthly spot check with a fixed set of questions, plus Bing’s Copilot report and the visits arriving from AI assistants.
- Timing
- Technology in days to a few weeks. Noticeable changes in the answers usually after three to six months.
- Price
- Fixed price after the free brand check. Ongoing also as part of the Digital Brand Partner.
Our own experiment
We prove it on our own website.
This site is the evidence for what we sell. Every page carries a validated schema graph, every page is fully readable in the delivered HTML. The robots.txt explicitly allows the crawlers of OpenAI, Google, Anthropic, Microsoft, Perplexity, Meta, Mistral, Apple and DuckDuckGo. And there is a separate fact page for AI systems.
Use the switch at the right edge of the screen to change to the agent view: you will see this page the way an assistant reads it.
What every page ships with
- One validated schema graph per page
- Content fully in the HTML, readable without JavaScript
- 21 AI and search crawlers explicitly allowed
- Changes reported to Bing and Copilot instantly via IndexNow
- Visible review date and maintained dateModified
- Questions and answers, directly quotable
Frequently asked questions
AI visibility means that AI assistants such as ChatGPT, Gemini, Claude, Copilot or Perplexity mention your company when someone asks for what you do. The discipline behind it is called AEO or GEO, short for answer engine optimisation and generative engine optimisation. The goal is not a position in a list but a place in the answer.
The nine major ones: ChatGPT, Gemini, Claude, Microsoft Copilot, Perplexity, Grok, Meta AI, Le Chat by Mistral and DeepSeek. For each system we check how it searches, which crawlers it sends and what it looks for when choosing sources. We measure mainly where your customers actually ask.
SEO aims for positions in a search engine’s results list. AEO aims for an answer engine to name your company in response to a specific question. GEO is the umbrella term for all generative systems that write a new text from several sources. The foundation is shared: almost all assistants search the Google or Bing index or an index of their own.
Buyers, engineers and applicants increasingly research with AI assistants. An AI answer names two or three companies instead of ten links and has no second page. If you are not among them, you are not on the shortlist. In many technical niches the answer is still open.
Through four levers: identical company data everywhere, a website that the search crawlers of all assistants may read in full, with structured data following Schema.org, content with concrete numbers and sources instead of marketing speak, and mentions on pages the systems trust.
It depends on the goal. The major providers separate crawlers for training from crawlers for search and for fetching on a user’s request. Blocking GPTBot, ClaudeBot or MistralAI-Training only prevents training. Blocking OAI-SearchBot, Claude-SearchBot, PerplexityBot, Meta-WebIndexer or MistralAI-Index removes you from those systems’ answers.
According to Sistrix they appear for around 20 percent of searches in Germany, and the first result falls there from just over 27 to 11 percent of clicks. Mainly knowledge questions are affected, less so searches for a supplier. Being named in the overview as a source keeps you visible.
The technical work on your own website takes days to a few weeks. Mentions on other sites take longer. Changes usually show up in the answers after three to six months.
No, and nobody can honestly promise that. AI assistants write every answer anew and choose their sources themselves. We improve the conditions and measure every month whether and how often your company is mentioned.
A text file in the root directory that gives AI systems the key facts about a company in compact form. It is not a binding standard, and no major platform has committed to using it. We still create one because it takes little effort and some agents read it. We do not sell it as a lever.
Yes. Every month we ask the assistants the same questions your customers ask and record who gets mentioned and which sources are linked. On top of that come Bing’s report on how often a site is cited in Copilot and the analysis of visits arriving from ChatGPT, Perplexity and others. Individual answers fluctuate, the trend over months is meaningful.
Not necessarily. Often adjustments to the technology, structured data and content of the existing site are enough. A new website makes sense if the content is only loaded via JavaScript, because most AI crawlers do not run JavaScript and then see an empty page.
Further reading
More on the topic.
How companies get mentioned in AI answers
Facts about Studio BUBORI for AI systems
Sources for this page
- Aggarwal et al.: GEO: Generative Engine Optimization. KDD 2024 (Princeton, IIT Delhi)
- OpenAI: Overview of crawlers (GPTBot, OAI-SearchBot, ChatGPT-User)
- Google: Crawlers and fetchers, including Google-Extended
- Anthropic: Anthropic’s crawlers (ClaudeBot, Claude-SearchBot, Claude-User)
- Perplexity: PerplexityBot and Perplexity-User
- Meta: Meta web crawlers (Meta-WebIndexer, meta-externalagent, meta-externalfetcher)
- Mistral AI: Crawlers (MistralAI-Index, MistralAI-User, MistralAI-Training)
- Bing Webmaster Tools: AI Performance (citations in Copilot)
- IndexNow: protocol for instant indexing
Does the assistant mention your company?
In the free AI check we ask ChatGPT, Gemini, Perplexity and Copilot the questions your customers ask. Afterwards you know whether your business comes up.
agent-view
What an answer engine reads from this page
A search engine returns ten links and the human picks. A language model returns one paragraph and names two or three companies in it. That is the difference between being found and being named. This page is built for the second case.
Retrieval
read: /llms.txt · /robots.txt · JSON-LD @graph · this page
answered: AI visibility means that AI assistants such as ChatGPT, Gemini, Claude, Copilot or Perplexity mention your company when someone asks for what you do. The discipline behind it is called AEO or GEO, short for answer engine optimisation and generative engine optimisation. The goal is not a position in a list but a place in the answer.
Allowed agents (21)
Anthropic: ClaudeBot · Claude-User · Claude-SearchBot
Perplexity: PerplexityBot · Perplexity-User
Google: Google-Extended · Googlebot
Microsoft: Bingbot
Apple: Applebot-Extended · Applebot
Meta: Meta-WebIndexer · meta-externalfetcher · meta-externalagent
Mistral AI: MistralAI-Index · MistralAI-User · MistralAI-Training
DuckDuckGo: DuckAssistBot
rule: Allow: / · source: https://www.bubori.studio/robots.txt
Entity
url: https://www.bubori.studio/en/ai-visibility
inLanguage: en-GB
alternate: de-DE · https://www.bubori.studio/ki-sichtbarkeit
@type: WebPage
updated: 2026-10-30
Questions this page answers
- What is AI visibility?
- AI visibility means that AI assistants such as ChatGPT, Gemini, Claude, Copilot or Perplexity mention your company when someone asks for what you do. The discipline behind it is called AEO or GEO, short for answer engine optimisation and generative engine optimisation. The goal is not a position in a list but a place in the answer.
- Which AI assistants do you cover?
- The nine major ones: ChatGPT, Gemini, Claude, Microsoft Copilot, Perplexity, Grok, Meta AI, Le Chat by Mistral and DeepSeek. For each system we check how it searches, which crawlers it sends and what it looks for when choosing sources. We measure mainly where your customers actually ask.
- What is the difference between SEO, AEO and GEO?
- SEO aims for positions in a search engine’s results list. AEO aims for an answer engine to name your company in response to a specific question. GEO is the umbrella term for all generative systems that write a new text from several sources. The foundation is shared: almost all assistants search the Google or Bing index or an index of their own.
- Why does it matter for mid-sized companies?
- Buyers, engineers and applicants increasingly research with AI assistants. An AI answer names two or three companies instead of ten links and has no second page. If you are not among them, you are not on the shortlist. In many technical niches the answer is still open.
- How do you become visible to AI assistants?
- Through four levers: identical company data everywhere, a website that the search crawlers of all assistants may read in full, with structured data following Schema.org, content with concrete numbers and sources instead of marketing speak, and mentions on pages the systems trust.
- Should you block AI crawlers in robots.txt?
- It depends on the goal. The major providers separate crawlers for training from crawlers for search and for fetching on a user’s request. Blocking GPTBot, ClaudeBot or MistralAI-Training only prevents training. Blocking OAI-SearchBot, Claude-SearchBot, PerplexityBot, Meta-WebIndexer or MistralAI-Index removes you from those systems’ answers.
- What do Google's AI Overviews mean for my business?
- According to Sistrix they appear for around 20 percent of searches in Germany, and the first result falls there from just over 27 to 11 percent of clicks. Mainly knowledge questions are affected, less so searches for a supplier. Being named in the overview as a source keeps you visible.
- How long until the answers change?
- The technical work on your own website takes days to a few weeks. Mentions on other sites take longer. Changes usually show up in the answers after three to six months.
- Can you guarantee that ChatGPT will mention us?
- No, and nobody can honestly promise that. AI assistants write every answer anew and choose their sources themselves. We improve the conditions and measure every month whether and how often your company is mentioned.
- What is an llms.txt, and does it help?
- A text file in the root directory that gives AI systems the key facts about a company in compact form. It is not a binding standard, and no major platform has committed to using it. We still create one because it takes little effort and some agents read it. We do not sell it as a lever.
- Can mentions in AI answers be measured?
- Yes. Every month we ask the assistants the same questions your customers ask and record who gets mentioned and which sources are linked. On top of that come Bing’s report on how often a site is cited in Copilot and the analysis of visits arriving from ChatGPT, Perplexity and others. Individual answers fluctuate, the trend over months is meaningful.
- Do I need a new website for this?
- Not necessarily. Often adjustments to the technology, structured data and content of the existing site are enough. A new website makes sense if the content is only loaded via JavaScript, because most AI crawlers do not run JavaScript and then see an empty page.
Contact
email: hello@bubori.com
address: Industriestraße 3, 78052 Villingen-Schwenningen, Baden-Württemberg
geo: 48.0916, 8.5046
Machine-readable entry points
sitemap: https://www.bubori.studio/sitemap.xml
JSON-LD: in the head of this page
Ten links leave the choice to the human. One paragraph has already made the choice. A company that does not appear in that answer is not on the list being chosen from.