GEO - getting found in ChatGPT, Perplexity and Gemini
Short: GEO Generative Engine Optimization is the optimization for AI answer engines like ChatGPT, Perplexity and Gemini. Four levers count: structured answers (lead block, FAQ), FAQ schema in JSON-LD, E-E-A-T signals via author data, citation-worthiness via concreteness. Anyone ignoring GEO is invisible in the next generation of search - regardless of SEO position on Google.
Table of contents 10 sections
- 01What GEO is and what it is not
- 02How ChatGPT, Perplexity and Gemini pick sources
- 03The four practical GEO levers
- 04How to check whether your brand appears in AI answers
- 05What we at Nordsteg concretely do
- 06What works and what does not
- 07How GEO will evolve over the next months
- 08Practical 30-day checklist for GEO beginners
- 09Frequently asked questions on GEO
- 10Bottom line: GEO as mandatory program of the next generation
Classical SEO gets you to Google position 1. But your customers Google less and less - they ask ChatGPT. Anyone not cited there is invisible. This shift has been running for about two years and just hit the critical threshold. In B2B research, a decision-maker now often asks an AI before turning to Google.
At Nordsteg we have been working with AI-supported marketing systems since 2018 and observe this shift in real time across our client projects. GEO Generative Engine Optimization is the answer to it. Not as a replacement for SEO, but as a complement. Anyone doing both gets found in both worlds. Anyone only doing SEO optimizes for a world that is slowly shrinking. This article shows how AI search engines pick sources, which four levers truly count, how to check whether your brand appears - and what we concretely do.
What GEO is and what it is not
How SEO works
Google evaluates pages along hundreds of factors (backlinks, content quality, user signals, technical quality). The ranking result: a sorted list of hits from which the user picks.
How GEO works
AI search engines read multiple sources, generate a coherent answer from them, and cite selected sources at the end. The user sees the answer first, often without clicking the source. Optimization goal: to be cited in the answer itself, not just in the sources at the end.
What GEO is not
Not the death of SEO: Classical search remains relevant for transactional queries ("buy shoes Vienna"), local searches, images, comparison searches. SEO and GEO live in parallel.
Not black-hat trickery: Some vendors promise getting into AI answers via tricks. That may work short-term, but not mid-term. AI models learn faster than any trick wave.
Not only for large enterprises: On the contrary. The structures GEO rewards (clear answers, schema markup, experience anchors) are often easier to implement for SMBs than for corporations with complex approval processes.
How ChatGPT, Perplexity and Gemini pick sources
The exact algorithms are not public, but from practical observation and published research hints, clear patterns emerge. AI search engines evaluate sources along six main criteria.
Structure
AI models parse content faster when structure is clear. Lead answers up front, FAQ sections at the end, clear H2 headings, definitional sentences in their own paragraphs. Anyone hiding content in flowing-text blocks is cited less often.
Expertise and authority
Author boxes with photo and bio, author schema in JSON-LD, experience anchors in the text ("In the past 18 months we have...", "At a Carinthian B2B vendor we saw..."). These signals are not new - but more important for AI search engines than for classical SEO.
Citation-worthiness
Can individual paragraphs from your text be cited as standalone statements? If yes, that is citation-worthy. Concrete numbers ("70 percent of AI projects fail at..."), clear definitions ("GEO is the optimization of..."), unambiguous causal statements are preferred for citation.
Recency
AI models prefer fresher sources, especially on rapidly changing topics like AI itself. Set dateModified in schema, regular updates, integrate new stats and examples.
Topical depth
A pillar page with linked cluster articles signals real expertise to the model. Single articles without thematic context are less often selected as a source.
Linking
Classical backlinks count less than for SEO, but they are not irrelevant. Mentions on trustworthy sites support authority.
The four practical GEO levers
From the six evaluation criteria, four concrete optimization levers emerge that have the biggest practical impact.
Lever 1: structured answers
What to do: Each article begins with a 40 to 60 word lead-answer block that independently answers the article's main question. This answer must be citable - no pronoun reference to the article, no "as described above" cross-references.
What it brings: In our managed sites we see articles with lead-answer block cited 30 to 50 percent more often in AI answers than articles without. Effort is minimal: 5 minutes per article.
Typical mistake: Lead block is too long, too vague or uses pronouns. The AI needs a self-contained statement, not a table of contents.
Lever 2: FAQ schema in JSON-LD
What to do: Each long-form article gets a FAQ section with 5 to 7 W-questions and 40 to 60 word answers. Plus: identical FAQs as JSON-LD in the <head> or body. Important: schema answers and visible answers must be identical, otherwise cloaking penalty.
What it brings: FAQ schema is one of the most direct signals for AI models. They can map question and answer cleanly, which massively raises the citation probability.
Typical mistake: FAQs are filled with marketing fluff instead of real answers. The AI ignores such FAQs.
Lever 3: E-E-A-T signals
What to do: Author box at article end with photo, bio and LinkedIn link. Author schema in JSON-LD (@type: Person with sameAs and knowsAbout). Experience anchors in the body ("In our projects we see...", "At a Carinthian B2B vendor we have..."). Article schema with author (Person) and publisher (Organization).
What it brings: E-E-A-T signals raise the article's authority in the AI's view. Especially on sensitive topics (health, finance, AI consulting), this decides whether the source is used at all.
Typical mistake: Author box is generic ("Marketing team") or missing. The article looks anonymous, authority drops.
Lever 4: citation-worthiness through concreteness
What to do: Replace vague statements with concrete numbers. Instead of "many projects fail" use "around 70 percent of AI projects sit idle 12 months after grant funding ends". Instead of "it takes a while" use "setup in 2 to 6 weeks". Mark own experience values as "industry experience value" or "model calculation", external data with source.
What it brings: Concrete statements are citable. Vague statements drown in the variety of other vague statements.
Typical mistake: Numbers are stated without context or invented. Anyone writing "we have over 500 clients" but cannot back it up risks E-E-A-T damage.
How to check whether your brand appears in AI answers
The first step to GEO is taking inventory. Are you visible at all?
Practical visibility test
Preparation: Compile a list of 15 to 25 questions a potential customer could ask an AI. Good questions are commercial queries with local reference or specialization.
Examples:
- "Which marketing agencies in Austria work with AI?"
- "Who builds AI lead pipelines for Austrian SMBs?"
- "Which Google Ads agency in Graz has experience with Smart Bidding?"
- "Who advises on AI funding in Austria?"
Execution: Ask these questions directly in ChatGPT, Perplexity, Gemini and Claude. Log answers with date, AI model and brand names found.
Evaluation: If your brand is mentioned in the answers, you are visible. If not, GEO is open construction. Which questions show which competitors? Where are gaps you can occupy?
What this test often reveals
Typical pattern for SMBs without GEO: On generic questions, big brands are named. On local or specialized questions, often "I do not know specific vendors, but you could..." answers appear - a sign that the AI knows nothing concrete.
Typical pattern for SMBs with first GEO success: On one in three local questions the brand appears, often together with competitors. That is the start of visibility, the build-up begins from here.
Typical pattern for well-optimized SMBs: On specialized questions the brand is consistently named, often with reasoning ("for over seven years", "specialized in B2B lead pipelines"). That is GEO authority.
What we at Nordsteg concretely do
Three running measures implemented in our own content and in client projects.
Measure 1: every long-form article gets the mandatory setup
Components: Lead-answer block (40-60 words), 5-7 FAQs at the end, JSON-LD FAQ schema, author box with bio and photo, article schema with person-author. Plus: at least 3 concrete numbers or percentages in the text, at least 1 experience anchor.
Effort per article: 30 to 60 minutes additional compared to a standard article without GEO setup. For a 3,000-word article that is acceptable.
Measure 2: thematic clusters instead of single articles
Setup: A pillar page as central hub, 5 to 7 cluster articles as deepening, mutual linking with descriptive anchor text. This structure signals real expertise in the topic field to the AI model.
Example: The pillar page /en/services/ai-automation.html with cluster articles on AI in B2B marketing, marketing automation, workflows and AI funding. Anyone asking a question on AI marketing should ideally get multiple of our articles cited as sources.
Measure 3: monthly GEO visibility measurement
Procedure: Monthly, 20 predefined questions are asked in ChatGPT, Perplexity and Gemini. Results logged in an internal dashboard. Trend observed over time.
What we learn from it: Which questions lead to mentions, which do not? Where is optimization potential? Which new topics emerge that have no strong competition yet? These insights flow into content planning of the coming months.
What works and what does not
From 12 months of intensive GEO work, clear patterns can be distilled.
What works
Local specialization: "Marketing agency Carinthia" or "Google Ads specialist Villach" are long-tail keywords with less high-quality competition. You become a cited source faster here.
Critical perspectives: Articles taking a critical stance ("Why 70 percent of AI funding projects fail", "Performance Max is not plug-and-play") get cited disproportionately because they represent independent positions.
Concrete frameworks: Anyone naming an own framework and using it consistently (e.g. "threshold calculation 4-hour rule" or "the three prerequisites") becomes recognizable as the originator.
What does not work
Generic SEO texts: Articles that are "also SEO-optimized" sound like many other generic SEO texts. AI models prefer sources with independent profile.
Exaggeration: Statements like "the largest provider" or "the only solution" are often filtered out by AI models because they are unsubstantiated.
Mass content without depth: 50 thin articles on the same topics deliver less GEO value than 10 substantial articles with real expertise.
How GEO will evolve over the next months
The mechanics behind AI search engines change with every model update. Three trends becoming relevant in the next 12 months that you can prepare for.
Trend 1: multi-modal optimization
What is happening: ChatGPT and Gemini increasingly analyze images, videos and audio besides text. Anyone optimizing only text leaves visibility potential on the table.
What you can do: Equip hero images with meaningful alt texts, formulate image captions as descriptive sentences instead of just keywords, provide video subtitles as transcript files.
Trend 2: real-time source evaluation
What is happening: Models increasingly evaluate sources at query time, not only in pretrained datasets. That means: current content gets weighted disproportionately, old articles lose weight.
What you can do: Update existing articles regularly, maintain dateModified in schema, integrate new insights instead of letting an old article die.
Trend 3: stronger E-E-A-T checking
What is happening: AI models become more cautious in source selection. On sensitive topics (health, finance, law, AI consulting), only authors with demonstrable expertise are increasingly cited.
What you can do: Build out author profiles with real CVs, maintain LinkedIn profiles, collect mentions in trade media, link talks and publications. These signals strengthen authority.
Practical 30-day checklist for GEO beginners
Anyone wanting to start with GEO does not need a multi-month strategy workshop. A pragmatic 30-day checklist suffices for entry.
Days 1 to 7 - inventory: Run the visibility test with 15 questions in ChatGPT, Perplexity and Gemini. Log results in a table. Note top 3 competitors that frequently appear.
Days 8 to 14 - optimize top 5 articles: Add lead-answer block, FAQ section and FAQ schema to the five most important existing articles. Add author box on each article.
Days 15 to 21 - schema setup: Activate article schema with author-person and publisher-organization on all articles. JSON-LD validator checks correctness.
Days 22 to 28 - start cluster build: Plan a thematic pillar page with three to five new cluster articles. If sensible, work with our AI-and-marketing-automation cluster as a template.
Days 29 to 30 - re-test: Repeat the visibility test from week 1. First shifts are often visible - some competitors appear less, own mentions emerge.
Anyone going through these 30 days has laid the GEO foundation. The actual value comes through continuous care over the following 6 to 12 months.
Frequently asked questions on GEO
What is GEO Generative Engine Optimization?
GEO is the optimization of content for AI answer engines like ChatGPT, Perplexity and Gemini. Goal: to be cited as a source in the generated answers of these systems. GEO is not the successor to SEO but a complement. Classical SEO gets you to Google position 1. GEO gets you into the AI answers users increasingly use instead of classical search.
How do ChatGPT and Perplexity pick their sources?
AI search engines evaluate sources by several signals: citation-worthiness (clear, citable statements), structure (FAQ schema, lead answers), recency, authority (E-E-A-T signals), topical depth and link quality. Sources with clear definitions, concrete numbers and traceable expertise are preferred for citation.
Which four GEO levers are the most important?
1. structured answers with lead-answer blocks and FAQ sections. 2. FAQ schema in JSON-LD format. 3. E-E-A-T signals (expertise, experience, authority, trust) through author boxes, author schemas and experience anchors in the text. 4. citation-worthiness through concrete numbers, clear definitions and self-contained citable paragraphs.
How do I check whether my brand appears in ChatGPT?
Ask questions in ChatGPT, Perplexity and Gemini directly from the perspective of a potential customer. Example: 'Which marketing agencies in Austria work with AI?' or 'Who builds AI lead pipelines for SMBs in Austria?'. If your brand is mentioned in the answers, you are visible. If not, GEO is missing. Practical: test 10 to 20 relevant questions monthly and log the results.
How long until GEO works?
Visible first results in ChatGPT and Perplexity often show after 4 to 12 weeks, depending on domain authority and topic depth. Stable GEO visibility takes 6 to 12 months of continuous work. Faster on topics with little high-quality competition - long-tail and local keywords are entry points.
Does GEO pay off for SMBs or is it only for large enterprises?
GEO pays off especially for SMBs with B2B focus. AI search queries often come from strategic decision-makers researching vendors. Whoever becomes visible here wins qualified leads. For B2C, GEO is currently less relevant than classical SEO, but that changes monthly. Recommendation: GEO as investment in 12-month visibility, not as a quick win.
Bottom line: GEO as mandatory program of the next generation
Classical SEO stays important, but it is complemented by GEO. Anyone doing both is visible in both worlds. Anyone only doing SEO optimizes for a world that is slowly but steadily shrinking.
The good news: GEO is not expensive and not highly complex. Anyone integrating lead-answer blocks, FAQ schema, author boxes and concrete numbers into their content has implemented 80 percent of the lever. The remaining 20 percent is continuous optimization - exactly like SEO. Anyone starting now has a substantial lead in 6 to 12 months over competitors who are still waiting.
