FAQ: AI visibility, aeo and ainomics

ainomics is a Vienna-based consultancy for AI visibility. ainomics GmbH makes sure companies are visible, correctly described and citable in AI-powered search, answer and recommendation systems such as ChatGPT, Perplexity, Google AI Overviews and Microsoft Copilot. Under the name aeo (ainomics engine optimization), ainomics combines digital consulting, Generative Engine Optimization, and communications and PR in one measurable steering loop.

ainomics is led by an interdisciplinary team of two managing partners and two partners. Mag. (FH) Tina Bernold is a digital strategist with more than 15 years of experience building and scaling digital business models. DI (FH) Philipp Schneider, MSc specializes in how companies are positioned in AI-driven information ecosystems. Mag. Gregor Schütze leads economic and crisis communications and is managing director of the PR agency Schütze Public Results. DI Christina Wilfinger was CEO of SAP Austria and held leadership roles at SAP and Microsoft.

ainomics GmbH is based at Berggasse 11/3, 1090 Vienna, Austria. From Vienna, ainomics serves companies across the German-speaking region and beyond. AI visibility is not tied to a location: it is measured and optimized per market, per language and per AI system.

You can reach ainomics by email at office@ainomics.at, by phone at +43 1 208 07 56-0, or via LinkedIn. The usual entry point is a free initial conversation covering the current state of your AI visibility and the scope of a possible audit.

aeo is particularly relevant for advisory-intensive and YMYL industries: healthcare, finance, insurance, tourism, retail and politics. YMYL stands for “Your Money or Your Life” — topics where wrong information can have direct financial or health consequences. Google states that it gives particular weight to expertise and trustworthiness on such topics. It is precisely there that a wrong representation in an AI answer weighs most heavily.

Source: Google – Creating helpful, reliable, people-first content

aeo is aimed at organizations whose customers prepare decisions with the help of AI systems — from mid-sized specializts to corporations and institutions. What matters is not headcount but how much explanation the offering requires and how long the decision process is. The more research and comparison happens before a purchase, the greater the impact of AI visibility.

aeo stands for ainomics engine optimization and is the name of the integrated approach of ainomics. It does not invent a new discipline but combines three existing ones in one loop: Generative Engine Optimization for technical and content findability, digital consulting for goals and KPIs, and communications and PR for independent third-party sources. Every measure feeds into a defined KPI.

No. AEO (Answer Engine Optimization) is a general industry term for optimizing content for answer engines such as ChatGPT, Perplexity or Google AI Overviews. aeo – in lowercase – stands for ainomics engine optimization and is the name under which ainomics runs GEO, digital consulting, and communications and PR together. GEO and AEO techniques are one building block, complemented by consulting, communications and continuous measurement.

GEO (generative engine optimization) optimizes for being found; aeo also optimizes for being understood. GEO focuses on content, structured data and technology. aeo adds strategic consulting and communications expertise and treats AI visibility as a management task with goals, KPIs, ownership and a fixed rhythm. The aim is not only to be mentioned, but to be correctly described and regularly cited.

Classic SEO targets placements in result lists through keywords, backlinks and meta tags. aeo targets what AI systems have learned, stored and rated as trustworthy about a company. The difference is where the decision happens: people used to scan a list of results, today they trust a generated answer directly. What is measured is therefore not ranking position but mention, description quality, recommendation, linking and citation.

SEO optimizes for placements in classic search engines. GEO (generative engine optimization) is the term coined in research for visibility in generative AI systems; it was published at the KDD conference in 2024. AEO (answer engine optimization) is used as an umbrella term for optimizing towards answer systems. LLMO (large language model optimization) emphasizes how language models learn and represent brands. None of these terms has a single agreed definition. aeo is ainomics’ approach that ties these layers together with consulting and communications into one steering system.

Source: Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024)

An aeo project runs in four phases. Phase 1, Audit & Diagnosis, delivers persona mapping, a prompt set, measured AI visibility and an action plan. Phase 2, Strategy & Roadmap, defines the project plan, targets, KPIs and a roadmap aligned with corporate strategy. Phase 3, Content & Third-party sources, builds citable content, structured data, studies, media relations and alliances. Phase 4 is the aeo loop with monitoring, reporting, Wikipedia stewardship and continuous optimization.

Yes. Without a measured baseline there is no way to prioritize or to prove success. The audit is deliberately kept compact — around six weeks — and delivers an immediately actionable plan. The baseline is what makes visible which of the five stages of AI visibility a company currently sits on, and which measures carry the greatest leverage.

The aeo loop is the fourth phase of the method and the ongoing operation of AI visibility. It covers monthly monitoring across all relevant AI systems, reporting against the defined KPIs, Wikipedia stewardship and continuous refinement of content and third-party sources. The loop is necessary because generative systems keep changing: models are updated, retrieval logic is adjusted, source preferences shift.

A prompt set is a structured collection of real user questions against which every measurement is taken. It covers three groups. Branded prompts ask about the company directly and test whether AI systems describe it correctly. Non-branded prompts are generic subject questions without any brand mention and measure the hardest form of visibility. Customer-journey prompts map the questions along the decision path and show at which stage visibility is missing.

Persona mapping describes who is asking: role, prior knowledge, decision situation and the language they use. AI users do not search, they ask — in natural language and with context. A managing director asks differently than a buyer, a patient differently than a referring physician. Only by knowing who asks can you identify the relevant prompts before they are asked. Persona mapping is therefore the foundation of the prompt set.

Funnel mapping assigns prompts to decision stages, because AI systems choose different sources depending on the type of question. Early on, definitions and orientation knowledge count; in the middle, comparisons and criteria; at the end, concrete provider information and trust signals. Each funnel stage gets its own prompts, its own content and its own metric. This shows whether visibility is missing early in the research or only at the selection stage.

AI visibility is built stage by stage. Stage 1, Mention: the company appears in AI answers at all. Stage 2, Correct description: services, positioning and facts are reproduced accurately. Stage 3, Recommendation: the company is actively recommended for fitting questions, even without brand mention in the question. Stage 4, Linking: AI systems point to your own pages. Stage 5, Cited source: the company is regularly the source AI systems draw their answers from.

The baseline is the documented first measurement of AI visibility before any measure is taken. It records, per AI system, per topic area and against competitors, how a company currently appears in answers. Without a baseline there is no reliable reference point for progress: changes in AI answers could not be distinguished from model updates or random variation. The baseline is created in Phase 1 and carried forward in the aeo loop.

Generative engine optimization (GEO) covers all measures that make content visible in AI systems such as ChatGPT, Perplexity or Google AI Overviews. The guiding question is not “how do I rank higher?” but “why does an AI system select precisely this source?”. GEO therefore works with clearly structured, self-contained content, explicit definitions, structured data and solid third-party sources rather than keyword density.

ainomics optimizes visibility in ChatGPT and ChatGPT Search, Claude, Google Gemini and Google AI Overviews, Microsoft Copilot and Perplexity. Each system has its own selection logic: Perplexity exposes its source choices particularly transparently, Google is strongly influenced by classic crawling and structure signals, and Claude and ChatGPT each run their own crawlers. That is why measurement happens per system rather than averaged across systems.

Because the place where decisions are made has shifted. More and more people ask an AI system directly and trust the generated answer instead of scanning a list of results. Only a handful of sources appear in that answer. Anyone not present there does not exist for that user — regardless of how well the page ranks on Google. SEO remains a prerequisite, but it is no longer sufficient.

From two places. First, from training: what the model learned up to its knowledge cutoff is stored in the model and is hard to influence in the short term. Second, from retrieval: for current questions the system fetches content live and generates the answer from it — and this path can be influenced. GEO measures therefore take effect through retrieval first, while training knowledge only catches up with later model generations.

That depends on the objective. OpenAI separates its crawlers: OAI-SearchBot is used for search in ChatGPT, GPTBot crawls content that may be used to train models, and ChatGPT-User fetches pages when users request it. These settings are independent: a site that blocks GPTBot can still appear in ChatGPT search via OAI-SearchBot. Other providers such as Anthropic (ClaudeBot), Perplexity (PerplexityBot) and Google (Google-Extended) have their own rules. We document bot rules in robots.txt and verify them against server logs and actual requests.

Structured data following schema.org, delivered as JSON-LD, makes content machine-readable and relationships between entities explicit. The most relevant types for AI visibility are Organization, WebSite, Person, Article, FAQPage, BreadcrumbList and DefinedTerm. They do not replace good content, but they reduce the risk of misinterpretation: what is marked up explicitly does not have to be guessed from running text.

An llms.txt file is not required for Google Search or Google’s AI features; according to Google it neither helps nor harms visibility or rankings there. Some other services or tools may read it. If we publish one, we treat it as an optional guide to existing content and check in the server logs whether it is actually fetched. Indexable pages, clear navigation, current facts and verifiable sources remain the priority.

Both are among the sources generative systems draw on most often. Wikipedia supplies structure, definitions and entity relationships and strongly shapes how a company is described; Reddit supplies lived experience and influences recommendation questions. Neither can be bought or directly controlled. What does work is Wikipedia stewardship: continuous, rules-compliant monitoring and maintenance of the factual record based on verifiable evidence.

E-E-A-T stands for experience, expertise, authoritativeness and trustworthiness — a quality framework for evaluating sources. The principle is central to AI visibility, but the weighting differs: identifiable authors with a profile, verifiable evidence, consistent terminology and independent third-party sources count for more than signals a site can simply assert about itself. Authority is topic-specific, not general.

First document it: which system, which prompt, which answer, which date — answers are not deterministic and they change. Then find the cause: outdated content of your own, contradictory statements across channels, incorrect third-party sources, or a hallucination with no source at all. The correction happens at the source: unambiguous facts on your own pages, consistent terminology, corrected third-party sources and structured data. Then measure again to see whether the correction lands in the answers.

YMYL stands for “Your Money or Your Life”: topics where wrong information has financial or health consequences. In healthcare, finance, insurance and comparable fields, AI systems handle sources more cautiously and prefer evidence-based, institutional and editorially reviewed content. For companies that means a higher hurdle to being included in answers — and greater damage when an outdated or wrong representation is what stands there.

Measurement runs against the prompt set: each question is put to the relevant AI systems regularly and the answer is evaluated in a structured way — does the company appear, is it described correctly, is it recommended, is it linked, is it cited. Because generative answers are not deterministic, the work relies on repetition and time series rather than single queries. The measurement is complemented by technical visibility, brand perspective and competitive benchmarking.

The KPIs follow the five stages of AI visibility. They include the mention rate per AI system and topic area, the accuracy of the description, the recommendation rate on non-branded prompts, links coming out of AI answers, and the citation rate — the share of relevant AI answers in which your source is named or cited. Share of model against competitors is tracked alongside these.

No, generally accepted benchmarks do not currently exist. Values depend heavily on industry, topic area, language, prompt type and AI system, and the systems themselves keep changing. What carries meaning is therefore not an absolute number but the development against your own baseline and the distance to direct competitors within the same prompt set.

First effects usually show where AI systems retrieve live: structure and content measures can appear in retrieval-based answers within a few weeks. Impact through third-party sources — press, studies, Wikipedia, expert articles — takes longer but is more stable. Anything anchored in model training only shifts with new model generations. A realistic horizon is two to four quarters, with monthly measurement.

No — and nobody credibly can. Generative systems do not deliver fixed placements but answers that vary with phrasing, context, user history and model version. What can be steered are the prerequisites: citable content, clean structure, consistent facts, open crawler access and solid third-party sources. What aeo guarantees is measurability and a systematic loop — not a rank in a list that does not exist in this form.

Partly. From stage 4 — linking — visibility turns into measurable traffic that can be traced in analytics and Google Search Console. Part of the effect, however, stays traffic-free: people read the answer, remember the name and later search for the brand directly. That is why AI visibility is not judged by sessions alone, but also by mention, description quality and recommendation rate.

Source: Pew Research Center (2025) – Fewer clicks when an AI summary appears

aeo is billed by consulting days. The scope per phase is defined during the audit and depends on the number of topic areas, markets and languages as well as the state of existing content and data structures. We discuss the concrete frame in an initial conversation: office@ainomics.at or +43 1 208 07 56-0.

The Phase 1 audit takes around six weeks. It is deliberately kept compact and delivers persona mapping, a prompt set, measured AI visibility as a baseline and an immediately actionable plan. After six weeks you can therefore decide whether and to what extent the following phases make sense.

Above all, access to knowledge and decisions. Concretely: contacts from marketing, communications and the relevant business unit, insight into existing content and data structures, access to the website, analytics and Google Search Console, and someone at management level who supports the goals and priorities. Ongoing effort on the client side concentrates on subject-matter input and approvals.

No. aeo does not replace SEO work, it builds on it. Technical cleanliness, indexability, site architecture and load times remain the precondition for AI systems to pick up content at all. aeo adds the question of source selection in generative systems — plus consulting, KPIs and communications. In practice ainomics works alongside existing agencies and in-house teams.

Parts of it, yes. Structure, definition-first content, structured data and opening up AI crawlers can be handled by an experienced content and tech team. Harder to run in-house are systematic measurement across several AI systems, building solid third-party sources and Wikipedia stewardship. A mixed model is common: ainomics provides method, measurement and steering, the internal team handles implementation.

With an initial conversation. In it we clarify which questions your customers are asking AI systems today, how your company is currently represented there, and what scope makes sense for Phase 1. Contact: office@ainomics.at, +43 1 208 07 56-0, ainomics GmbH, Berggasse 11/3, 1090 Vienna, Austria.