GEO for B2B Industries: How Expert Content Gets Cited in ChatGPT and Perplexity

By Edip Kilinc, founder of Webwavers. Published on October 31, 2026, last updated on October 31, 2026. Reading time: about 18 minutes.

The Key Points in Brief

AI search cites expert content that answers a specific question directly, is backed by numbers and sources, and can be clearly attributed to a company or a person. The basic rules are the same in every industry. What probably differs are the sources models draw recommendations from (this is only partly documented), plus the mandatory details, the schema types and the kind of proof that builds trust.

• IT and software: According to a G2 survey, review portals and AI chats are buyers' most important shortlist sources. Your own website is only part of the picture.
• Manufacturing: Datasheets, standards and certificates must exist as readable text.
• Real estate: Property data and energy figures are mandatory. Portals probably play a big role for property queries.
• Energy: Verifiable figures with a reporting date and system boundary beat sustainability prose.
• Aviation: Technical terms, approvals and course data must be exact. English is part of that.
• Trades: Local signals and reviews are probably central.
• Consulting and law firms: People, expertise and approved cases are the citable unit.

Table of Contents

1. What Is the Same in Every Industry
2. Where the Answer Differs: Overview
3. IT, Software and SaaS
4. Manufacturing and Mechanical Engineering
5. Real Estate
6. Energy and Sustainability
7. Mobility and Aviation
8. Trades and Skilled Crafts
9. Consulting and Law Firms
10. The 6-Step Checklist for Every Industry
11. Common Mistakes
12. How We Implement This
13. FAQ

What Is the Same in Every Industry

Before we get to the differences, the baseline. It applies to all seven industries and is the reason GEO is not a special path.

Google states in its documentation on AI features that AI Overviews and AI Mode have no special requirements beyond the familiar ones: indexable, eligible for snippets, content as text, internal linking, no blocked crawlers. Structured data must match the visible text. If you load content only in images, in tabs without markup or via JavaScript, it is understood less well.

On the content side, the research paper on Generative Engine Optimization (Aggarwal et al., KDD 2024) shows that backed statistics, direct quotes and named sources can increase visibility in generated answers by up to 40 percent, depending on the field. The effect differs from field to field. That is exactly what this guide is about.

On the demand side, the direction is clear. Gartner reports for 2026 that 67 percent of B2B buyers prefer to buy without a sales rep. 6sense reports for 2025 that 94 percent of buying groups had already ranked their preferred vendors before first contact. According to Bitkom, half of respondents in Germany use AI chats for internet search at least occasionally. The samples differ, the trend does not: Decision-makers do their research up front, and often in AI answers.

What does that mean in practice? Every page that should get cited needs

• a question in the title or heading that someone really asks,
• the answer in the first two to three sentences,
• numbers with a source and a year,
• a recognizable sender (company, author, date),
• structure: lists, tables, clear subheadings.

Everything else is industry work.

Where the Answer Differs: Overview

The table summarizes what actually changes from industry to industry. Details follow in the sections below.

Industry	Possible sources for recommendations (evidence or hypothesis)	Mandatory or core content on the page	Typical schema types	Typical mistake
IT and software	Review portals and AI chats as buyers' shortlist sources (G2, documented); which pages AI answers cite: hypothesis	Integrations, security, pricing logic, comparisons	SoftwareApplication, Organization, FAQPage	Marketing copy only, no verifiable claims
Manufacturing and mechanical engineering	Hypothesis: manufacturer and trade association sites, standards, product databases	Datasheets, certificates with standard numbers, parameters	Product with additionalProperty, Organization	Key data only in PDF or image
Real estate	Hypothesis: real estate portals, project and location pages	Property data, energy figures, location facts	RealEstateListing, Offer, Place	Mandatory details missing, properties only on the portal
Energy and sustainability	Hypothesis: authorities, funding bodies, studies, trade press	Figures with reporting date and system boundary, funding logic	Article, Dataset, Organization	Unsubstantiated sustainability claims
Mobility and aviation	Hypothesis: authorities, associations, trade media, certifiers	Approvals, course and date data, technical terms, English	Organization, Course, DefinedTerm	Imprecise technical details, German only
Trades and skilled crafts	Hypothesis: map and review data, local directories	Service area, services, reviews, careers	LocalBusiness subtypes, Service	No local clarity
Consulting and law firms	Hypothesis: personal profiles, professional publications, directories	Experts, practice areas, approved cases	ProfessionalService, Person, Article	Anonymous texts, marketing promises

Unless noted otherwise, the sources column rests on industry logic, not on a published statistic. Only your own test with your own prompts shows what the situation looks like in your market, as described in the post "Measure AI Visibility" on our website.

IT, Software and SaaS

The answer differs here in the sources. According to G2 (survey of more than 1,000 software buyers, August 2025), half of software buyers start their research in an AI chatbot instead of Google, and AI chats are considered the most important source for shortlists there. That makes review portals and comparison sites documented as important for shortlists. That they are also cited especially often in AI answers to vendor questions is plausible but not documented. A software company probably has a harder time getting named if nobody talks about it on G2, Capterra or in expert comparisons.

That changes the priorities:

1. Build the product as an entity. Name, category, target group, main function in two sentences, consistent across website, profiles and directories. This helps models assign you to a category.
2. Comparison and alternative pages. Questions like "alternative to X" or "X vs. Y" get asked. If you answer them fairly yourself, with a table and clear criteria, you are more likely to get cited than someone who ignores them.
3. Integrations, security, data protection. Buying committees are large: Forrester cites 13 internal and 9 external participants as typical for 2026. Each role asks its own questions. Engineering wants APIs and integrations, data protection wants server location and data processing agreements, procurement wants pricing logic. Each of these questions deserves its own page with a direct answer.
4. Maintain review profiles. This is outreach work, not writing work, and often the biggest lever.
5. Structured data. SoftwareApplication with category, operating system or platform and offer information, provided it is visible on the page. Google offers rich results for this only in certain cases, but as a machine-readable clarification it is useful anyway.

Example from our work: For ScaleITS, an IT service provider, we evolved the brand in 2026, from the logo symbol and color system to the website and materials. The point for GEO: A clear, consistent brand description helps models place a company unambiguously. You can find more about our work for IT companies on the page IT, Software and SaaS.

Manufacturing and Mechanical Engineering

The answer differs here in the form of the data. Industrial buyers search for parameters, standards and proof: material, tolerance, manufacturing process, certification, lead time. Contentful surveyed 900 decision-makers from manufacturing, industry and technology in EMEA (2025 B2B Buyer Benchmark Report). 84 percent say self-service tools are critical for vendor selection. The biggest friction points respondents name are missing price transparency (39 percent), cumbersome reordering (35 percent) and slow vendor response (24 percent).

For GEO, that means:

• Datasheets not only as PDF. Models can read PDFs, but they are harder to attribute and to keep current. Put the core parameters on the product page as an HTML table and link the PDF in addition.
• Certificates with name and number. "Certified" says nothing. "ISO 9001, ISO 3834" with validity and certification body is a verifiable statement. Every certificate belongs on its own page or in its own section, with an explanation of what it means for customers.
• Answer technical questions. "Which welding process suits stainless steel frames with high dimensional accuracy?" is a prompt a buyer really asks. A guide from the manufacturer that answers this question seriously is more citable than any corporate brochure.
• Structure product data. Schema.org Product with additionalProperty for technical attributes. Caution: Only mark up what is visible on the page.
• Multilingual. Export-oriented companies need English with its own URL structure and hreflang. Models answer in the language of the question and search accordingly.

The data on research in mechanical engineering is thin. There is no current German study on online supplier search for 2024 to 2026. We therefore rely on Contentful for EMEA and on prompt tests.

Example from our work: For Heigl GmbH & Co. KG in Blaubeuren, an industrial welding shop and frame builder, we redesigned the web presence and built it in Webflow. The site covers welding, CNC milling, bending and laser technology, plus certifications to ISO 9001 and ISO 3834, manufacturing depth, references and careers. Clearly named processes and standards like these are the building blocks that technical questions can be answered from. Everything about the industry is at Industry and Engineering.

Real Estate

The answer differs here in the mandatory details and probably in portal power. For property and vendor questions, portals probably play a big role. Your own website wins where portals cannot reach: project developments, location knowledge, portfolio strategy, developer expertise and questions no listing answers.

Two peculiarities shape the industry.

Mandatory details in listings. If a property with an energy performance certificate is advertised in commercial media, the law prescribes details such as the type of certificate, the energy figure, the energy source and, for residential buildings, the year of construction and efficiency class (§ 87 para. 1 in the version on gesetze-im-internet.de, retrieved on October 31, 2026; the provision is listed there under the Gebäudemodernisierungsgesetz, the German Building Modernization Act). Warning letters and lawsuits over missing details are documented for 2024, for example by Deutsche Umwelthilfe. For property pages this means: The details appear in the visible text, not only on the portal. As a side effect this also helps citability, because the page supplies complete facts.

Personal trust. A survey of 500 people in Germany (Immoscout24, March 2026) found that 74 percent consider personal contacts decisive when looking for an apartment. The figure concerns people searching, not investors or municipalities, but it shows how strongly real estate decisions depend on people. Agent profile, team and referral options are therefore part of GEO work.

What you can do concretely:

• Property pages as complete data sets. Location, area, rooms, year of construction, energy figures, features, price or pricing logic as readable text. Add RealEstateListing and Offer, as far as the data is visible.
• Location pages with real knowledge. Infrastructure, price trends with a source, zoning situation. Not a text kit for 200 towns; Google counts that as scaled content abuse.
• Project pages with construction progress and facts. Sales status, floor plans, schedule, parties involved. This is exactly what investors and buyers ask.
• Agent profiles as people. Name, license, area, specialization. This lets a model assign a competence to a person.
• Maintain portals and directories. Recommendations can originate there.

Example from our work: For Atoria Immobilien, we developed a corporate identity in 2026 that relies on structure, precision and authority, with applications from the website to letterhead and business cards. For GEO this is the foundation: A distinct brand with clear positioning is easier to cite than an interchangeable presence. You can find more information at Real Estate.

Energy and Sustainability

The answer differs here in the burden of proof. Energy and climate topics are fields where claims quickly become open to attack. Models and readers check whether a statement is backed by numbers, time periods and methods. A page that writes "climate neutral" and proves nothing is less citable than one that states emissions with a system boundary, base year and source.

The need for information is high. The EY/BDEW municipal utilities study 2024 (100 utilities) names digitalization, heat planning and recruiting skilled workers as the most important topics for 87 percent of respondents. 99 percent are involved by their municipality in heat planning, and 79 percent consider the federal funding for efficient heating networks insufficient. These are topics customers, municipalities and specialists ask questions about, and on which a utility or energy company can take a professional position.

What you can do:

• Figures with reporting date, method and system boundary. Every number gets a year and a source. Don't publish reports only as PDF, but with an HTML summary and a table.
• Answer funding and planning questions. Anyone who explains how heat planning, funding and connection fit together becomes a source. Link primary sources such as funding bodies and statutes.
• Make technology understandable. Explain technical terms, present processes in steps.
• Have claims reviewed. Sustainability claims are subject to competition law. Have wording reviewed legally before it goes live.
• Mark up Dataset and Article. If you publish data series, markup with Dataset can clarify the structure. Prerequisite: The data is visible on the page.
• Separate the customer portal from required pages. Advisory content belongs on public, indexable pages, not behind a login.

For energy startups there are no reliable market figures on how decision-makers research. The same logic applies as for IT: Build the product entity, case studies, integrations and reviews.

Example from our work: For BRIC Energy, a European EPC partner for the construction, procurement and operation of large-scale photovoltaic plants, we developed the identity in 2026. The goal was technical credibility with utilities, landowners and institutional partners, deliberately not as a green brand but as an infrastructure company with industrial precision. That fits the GEO logic of the industry: sober facts instead of sustainability rhetoric. More at Energy.

Mobility and Aviation

The answer differs here in precision. Technical terms, approvals, and course and time details must be exact. A model that quotes a wrong prerequisite for a course or a wrong certification from your page hurts you twice: through misinformation and through lost trust. You should therefore always check details on approvals, license levels and authorities against the responsible body and keep them current. We did not research the aviation regulations (EASA, national authority) for this guide and make no statements on them.

Two groups with different needs:

Suppliers and manufacturers. The market is under pressure: According to VDA (October 2025), the German automotive industry has lost almost 120,000 jobs since 2019, and suppliers lost 24 percent of their workforce from August 2019 to August 2025. Many suppliers are therefore looking for new markets, for example aviation. Buyers vet suppliers heavily through certificates and proof. For GEO this means: a certificate page with standard designations, datasheets as HTML parameters, manufacturing processes explained and English as a full-fledged language.

Flight schools and service providers. Here it is about courses, prerequisites, dates, locations and prices. Prompts read like "How do I become a private pilot" or "Which flight school in Bavaria offers trial flights". Good pages answer these questions completely: prerequisites (age, medical fitness), process, duration, costs as a range with conditions, instructors, location. Add Course and Organization as structured data, as far as they are visible.

More levers:

• Glossary with technical terms. Definitions in one to two sentences, with a source, marked up as DefinedTerm. Technical terms are prompts, and a clean entry gets cited.
• Check accessibility. If you offer online booking for consumers, you probably fall under the Accessibility Strengthening Act (BFSG, Barrierefreiheitsstärkungsgesetz). The exemption is in § 3 para. 3 BFSG: The obligation does not apply to micro-enterprises that offer or provide services. According to § 2 no. 17 BFSG, a micro-enterprise employs fewer than ten people and has either annual revenue of at most €2 million or an annual balance sheet total of at most €2 million. The exemption concerns services, not products. If in doubt, get legal clarification.
• Careers section. Skilled workers for engineering and training are becoming scarce. Concrete job and apprenticeship pages are citation targets of their own.

There are no current German studies on demand and enrollment numbers at flight schools. The only concrete figure on the skills gap in air transport comes from IW in 2022 and lies outside the period we use here.

Example from our work: For Lenti Aviation, a private aviation company, we developed the brand identity in 2025, deliberately calm, exclusive and trustworthy. According to the project page, it forms the basis for the current web design in Webflow. More on the industry at Aviation.

Trades and Skilled Crafts

The answer differs here in local focus. Anyone looking for a tradesperson searches locally. There is good reason to think that answers to local questions are fed heavily by map, review and directory data. This is not a published statistic and may change; check it for your area with your own prompts. If it holds, the work shifts from the website to its surroundings.

What counts:

• Clear local identity. Name, address, phone number consistent across website, Google Business Profile and directories. Service area named in towns, not as "nationwide".
• Service pages with concrete answers. "What does a heating modernization cost?" with a range, influencing factors and process. Prices only as a range with conditions.
• References with place and photo. Project, place, service, time period, with the customer's consent.
• Actively collect reviews. They are considered a strong local signal, but for AI answers this is not documented.
• Structured data. LocalBusiness or a suitable subtype, opening hours, service area. It must match the visible text.
• Careers and apprentice page. The skills gap in the trades is documented: According to KOFA-Kompakt 07/2026 (IW, 2025 data), 209,444 vacancies faced 139,239 unemployed skilled workers, and the gap was 97,482. Anyone who wants applicants needs a page that answers questions like "electronics technician apprenticeship in the region".
• Accessibility. The BFSG exempts micro-enterprises for services (§ 3 para. 3 BFSG). Micro-enterprises are businesses with fewer than ten employees and annual revenue or an annual balance sheet total of at most €2 million (§ 2 no. 17 BFSG). The exemption applies to services, not to products. Larger businesses with an online shop or booking tool for consumers should check whether they are affected.

There is no current source on how private and commercial customers research trades, so statements on that remain assumptions. More on the industry at Trades and Skilled Crafts.

Consulting and Law Firms

The answer differs here in the citable unit. In consulting, a product is not recommended, a person or a team is. A law firm or management consultancy sells judgment. If you want to be named in AI answers, you therefore have to make people, practice areas and cases visible. At the same time, stricter standards apply to consulting: Legal, tax and financial topics count at Google among the areas where trust is judged especially strictly (YMYL). The abbreviation stands for Your Money or Your Life.

User behavior is also special. A Bitkom survey (1,004 people aged 16 and over, late 2024) shows that only 12 percent would definitely ask an AI first with legal problems, and 14 percent at most for simple questions. 42 percent would go to a lawyer in person. 54 percent believe AI cannot empathize with those affected. This concerns consumers; for management consultancies in the B2B world, the buyer findings from the software world apply more. But it shows how strongly consulting rests on trust and personal relationships. The consequence: AI is often not the decision but the entry point and the check. Gartner reports that 69 percent of B2B buyers additionally validate AI insights with a sales rep.

What you can do:

• Expert pages per person. Name, photo, role, practice areas, education, publications, mandates or projects in permissible form, contact. As a Person with sameAs links to LinkedIn, association profile or chamber directory. This lets a model connect name and competence.
• Expert articles with an author. Every insight article has an author with a profile, a date and sources. Anonymous texts under "Editorial team" are less citable. This is exactly the pattern consultancies can change: Many publish insights that are hardly found because author, structure and question are missing.
• Case studies with approval. Attorney-client confidentiality and discretion limit what you may show. Anonymized case examples with starting point, approach, result and lessons learned are possible. You should clarify professional rules and advertising restrictions for lawyers, tax advisors and auditors beforehand.
• Answer professional questions. Answer the questions clients ask directly and clearly, with a note on the limits of general information.
• Structured data. ProfessionalService or a suitable subtype such as LegalService for the organization, Person for the advisors, Article for expert posts. Check everything against the visible text.
• Directories and chamber profiles. Professional directories and professional profiles can be sources for recommendations of people. There are no reliable studies on this.

Example from our work: For the tax firm Steuerkanzlei Spengler, we designed a new web presence in 2025 meant to convey tax topics in an understandable and accessible way. A photo shoot showing the people behind the firm was added. The project page describes the impact qualitatively. For GEO this is one building block: make people visible, explain clearly. Expert pages and case studies for consultancies are the topic of a guide of their own. More on the industry at Consulting and Agencies.

The 6-Step Checklist for Every Industry

This order works regardless of the industry.

1. Collect prompts. 20 to 40 real questions from your customers, split by phase: understand the problem, find vendors, compare vendors, vet vendors.
2. Measure the baseline. The same prompts in ChatGPT, Perplexity, Gemini and Google AI Mode, repeated several times. Who gets named, which sources appear below?
3. Map the source landscape. Which portals, directories, associations or media show up? Where are you missing?
4. Revise core pages. One page per main question with a short answer, numbers with a source, a table and an FAQ. Author and date visible.
5. Check technology and structure. Indexing, rendering, performance, structured data matching the text, multilingual pages where needed.
6. Measure, adjust. Repeat after four and after eight weeks, close gaps.

The llms.txt file is a small building block that is quick to implement among several, but not a switch for visibility and no substitute for good content and clean structure. Google said in June 2025 that no AI system currently uses it. Include it, but don't expect miracles.

Common Mistakes

• Generating industry pages from a template. One page per industry with swapped words is exactly what Google treats as scaled content abuse. Every industry needs its own facts, examples and sources.
• Certificates and data only as image or PDF. Models read text. Images stay invisible.
• Schema that does not mirror the text. Anyone who marks up reviews that are not visible risks violating the guidelines.
• Forgetting outreach. In IT, real estate, trades and consulting, recommendations probably often originate outside your own website.
• Statements that are too general. "We are your partner for quality" cannot be cited by anyone. "We manufacture welded frame structures to ISO 3834" can.
• Expecting guarantees. Answers fluctuate. A serious plan works with probabilities and measurement series. How to judge agencies on this is described in the guide "Choose a GEO Agency".

How We Implement This

Webwavers is a premium agency for branding, web design and web development, founded in 2021 in Günzburg, with a second location in Munich. We are one of seven Webflow Premium Partners in Germany and also work with WordPress, TYPO3 and headless systems. At our company, GEO belongs to the Growth area. It typically works like this: First there are queries on your customers' questions and a technical check of whether AI crawlers can read the site. Then technical blockers are fixed and the most important pages are made citable, after which the presence in external sources is built up and measured on an ongoing basis. What this looks like at our company is described on the page GEO and AI Visibility.

FAQ

Does GEO really differ by industry?

The basics are the same: indexable, directly answering pages with evidence and a recognizable sender. Differences arise in the sources models draw recommendations from, in mandatory details and in the kind of proof. A law firm needs personal profiles, a manufacturer needs datasheets, a tradesperson needs local signals.

Which schema types do I need per industry?

Depending on the industry, SoftwareApplication, Product, RealEstateListing, Dataset, Course, LocalBusiness or ProfessionalService fit. More important than the type is that the markup mirrors the visible text. Google requires no special markup for AI features. Schema is a clarification for machines, not a substitute for good content.

How do I find out what AI search cites in my industry?

Put 20 to 40 real customer questions to several models and note the vendors and sources named below the answers. Repeat the questions several times, because answers fluctuate. The most frequent sources show where you need to be visible: your own site, portals, associations, media or directories.

May a law firm show cases on its website?

Only to a limited extent. Attorney-client confidentiality and discretion limit what is allowed. Anonymized case examples with consent are a common route. You should clarify professional advertising restrictions and competition law with a qualified body before publishing. The legal assessment in the individual case is up to your legal counsel or chamber.

Is GEO worthwhile for small trade businesses?

Often first through local basics: a well-maintained Google Business Profile, consistent details, reviews and service pages with clear answers. A comprehensive GEO program usually only pays off once the basics are in place and you want to be found beyond your local area.

Do manufacturing companies need English pages?

If you export or want to reach international buyers, yes. Models answer in the language of the question and search in that language. A separate URL structure with hreflang and a fully fledged English version in terms of content are better than automatically translated pages without review.

Can I start GEO without an agency?

Yes, with a prompt list, a baseline measurement and the revision of three to five core pages. An agency is more worthwhile when technology, content and outreach have to be tackled at the same time. How to compare vendors is covered in the guide "Choose a GEO Agency".

Sources

1. Google: AI features and your website
2. Google: General structured data guidelines
3. Google: Spam Policies for Google Web Search
4. Aggarwal et al.: GEO, Generative Engine Optimization (KDD 2024)
5. Search Engine Roundtable: Google zu llms.txt, Juni 2025 (Google on llms.txt, June 2025)
6. Gartner: 67 Prozent der B2B-Käufer bevorzugen Kauf ohne Vertrieb, 09.03.2026 (67 percent of B2B buyers prefer buying without sales, March 9, 2026)
7. Gartner: 69 Prozent validieren KI-Erkenntnisse beim Vertrieb, 20.05.2026 (69 percent validate AI insights with sales, May 20, 2026)
8. 6sense: 2025 B2B Buyer Experience Report, Pressemitteilung (press release)
9. Bitkom: Hälfte nutzt KI-Chats für die Internetsuche, 20.11.2025 (Half use AI chats for internet search, November 20, 2025)
10. G2: AI search surging for B2B buyers, 2025
11. Forrester: State of Business Buying 2026
12. Contentful: 2025 B2B Buyer Benchmark Report
13. Immoscout24 / Scout24: Umfrage Wohnungssuche, März 2026 (apartment search survey, March 2026)
14. Gesetze im Internet: § 87 (Pflichtangaben in Immobilienanzeigen), abgerufen 31.10.2026 (mandatory details in real estate listings, retrieved October 31, 2026)
15. Deutsche Umwelthilfe: Klage zu Energieangaben in Inseraten, 2024 (lawsuit on energy details in listings, 2024)
16. BDEW / EY: Stadtwerkestudie 2024 (municipal utilities study 2024)
17. Verkehrsrundschau: VDA zum Stellenabbau in der Automobilindustrie, Oktober 2025 (VDA on job cuts in the automotive industry, October 2025)
18. Görg: Was das BFSG für Unternehmen bedeutet, 2025 (What the BFSG means for companies, 2025)
    • Statutory text: BFSG § 2 (Begriffe) (definitions) and BFSG § 3 (Anforderungen, Ausnahme Kleinstunternehmen) (requirements, micro-enterprise exemption)
19. IW Köln: KOFA Kompakt 07/2026, Handwerk (trades)
20. LTO: Bitkom-Umfrage zu digitalem Rechtsrat, 2025 (Bitkom survey on digital legal advice, 2025)
21. Webwavers: Heigl Blaubeuren project
22. Webwavers: ScaleITS project
23. Webwavers: Atoria Immobilien project
24. Webwavers: BRIC project
25. Webwavers: Lenti Aviation project
26. Webwavers: Steuerkanzlei Spengler project

Change log: Published on October 31, 2026.