GEO optimization

Get found in ChatGPT, Perplexity and Google AI

Get mentioned in ChatGPT, Perplexity and Google AI Overviews: VirtuoLab handles technology, content and monthly measurement.

SEO and GEO in five questions

What is GEO?

GEO, short for generative engine optimisation, is the practice of making a business visible in the answers of AI assistants such as ChatGPT, Perplexity, Google AI Overviews, Gemini and Claude. The term comes from the research paper GEO: Generative Engine Optimization by researchers at Princeton University and IIT Delhi. The paper appeared on arXiv in November 2023 and was accepted at KDD 2024. Classic SEO aims for a position in a list of links. GEO aims for a mention or a citation inside the generated answer. An AI assistant names a business when three conditions are met. First, the assistant can read the website. Second, the assistant understands what the business offers. Third, other sources confirm the same picture. In practice, GEO therefore works on access for AI crawlers, on content with direct answers, and on one consistent business description, for example through LinkedIn and Google Business Profile. VirtuoLab offers GEO together with SEO, because GEO builds on a website Google can already find.

What is the difference between SEO and GEO?

The difference between SEO and GEO is the goal: SEO aims for a high ranking in search results, while GEO aims for a mention inside an AI answer. SEO targets Google and Bing, and success means a click to the website. GEO targets assistants such as ChatGPT, Perplexity and Google AI Overviews, and success means a recommendation, often with a link as the source. The shift matters: in a study by Pew Research Center, users clicked a traditional result in 8% of visits with an AI summary, against 15% without one. Additionally, GEO does not replace SEO. There are no additional requirements to appear in AI Overviews or AI Mode, according to Google Search Central, and a page must be indexed and eligible for a snippet. Google last updated the guidance in December 2025. No special schema.org markup is needed. Measuring differs as well. For example, SEO tracks rankings with Google Search Console, while GEO tracks mentions and citations on a fixed set of questions.

How do AI assistants choose their sources?

An AI assistant chooses sources in two steps: first the assistant must be allowed to crawl a page, then the page must answer the question well. Step 1 is access. OpenAI surfaces websites in ChatGPT search with the OAI-SearchBot crawler, and sites opted out of OAI-SearchBot are not shown in ChatGPT search answers. Robots.txt changes take about 24 hours to apply, according to OpenAI. Perplexity surfaces and links websites in Perplexity search results via PerplexityBot. AI Overviews and AI Mode may use query fan-out, according to Google Search Central: the system runs multiple related searches across subtopics to build one answer. For example, one question about software prices can trigger separate searches for web apps, websites and mobile apps. Step 2 is content. The GEO study from Princeton University and IIT Delhi, published in 2023, showed citing sources, adding quotations and adding statistics raised visibility by up to 40% in generative engine responses. Keyword stuffing offered little to no improvement.

How does VirtuoLab approach a GEO project?

A GEO project at VirtuoLab is a cycle of four steps, carried out by Mustafa Kose. Step 1 is the baseline. Together with the client, VirtuoLab drafts a set of questions customers would realistically ask an AI assistant, records where the business is mentioned today, and scans the website for AI access and structured data. Step 2 is the foundation. Crawlers such as OAI-SearchBot and PerplexityBot get access, the site gets an llms.txt file and schema.org data, and Bing receives updates through IndexNow. Business details become consistent everywhere, for example on LinkedIn and in Google Business Profile. Step 3 is answer content: new and rewritten pages address the baseline questions with the answer first, plus sources and statistics, which raised visibility by up to 40% in the GEO study from 2023. Step 4 is measuring and adjusting. Every month, the same questions go to ChatGPT, Perplexity, Google AI Overviews, Gemini and Claude, and the results are compared with the baseline. The website virtuolab.dev uses the same setup.

What does GEO deliver?

The result of GEO is a website AI assistants such as ChatGPT, Perplexity and Google AI Overviews can read, understand and cite, so a business gets mentioned faster and more often. VirtuoLab builds the result in three layers. First, the technology: crawler access, an llms.txt file and structured data. Second, the content: clear answers to customer questions, with the answer in the first sentence. Third, the entity: one consistent business profile across the web, for example through LinkedIn and Google Business Profile. Research supports the approach. The GEO paper on arXiv reports visibility gains of up to 40% on GEO-bench, a benchmark of 10,000 queries, and was published at KDD 2024. Citing sources, adding quotations and adding statistics worked best, while keyword stuffing even performed 10% worse on Perplexity. In practice, VirtuoLab therefore writes for readers, not for keywords. AI answers change per question and per model, so VirtuoLab measures progress every month on a fixed set of questions and reports the results in plain figures.

GEO vs SEO: what is the difference?

SEO targets the search results; GEO targets the answers of AI assistants. GEO builds on SEO.

Goal
SEO: A high position in search results
GEO: Being mentioned and cited in the AI answer
Where
SEO: Google and Bing
GEO: ChatGPT, Perplexity, Google AI Overviews, Gemini, Claude, Copilot
What counts
SEO: Keywords, links, technical health
GEO: Clear answers, structured data, confirmation by other sources
Outcome
SEO: A click to your site
GEO: A recommendation — often with a link as the source
Measuring
SEO: Rankings and traffic
GEO: Mentions and citations on a fixed question set

What we actually do

Four parts, each with a tangible result you can check yourself.

01

Technology for AI crawlers
Making sure every AI assistant is allowed to — and able to — read your site.
robots.txt with explicit access for OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot and Google-Extended
llms.txt and llms-full.txt: your business and offering in plain text for language models
Structured data (schema.org): Organization, Service, FAQPage, Article and Breadcrumb
Bing Webmaster Tools and IndexNow — the index that Copilot and ChatGPT search, among others, rely on
Sitemap with language variants, fast load times and clean HTML

02

One recognisable entity
The same story everywhere, so an AI recognises your business with confidence.
One consistent business description: name, address, phone, services and service area
Linking your official profiles (sameAs) — LinkedIn, Instagram, Google Business Profile
Founders or experts as recognisable authors (Person schema)
Mentions on relevant industry and comparison sites

03

Content that is an answer
Pages built around the questions your customers ask AI.
A question set per audience: what do your customers ask ChatGPT?
The question as the heading, the answer in the first sentence
Figures with source and date, so an AI is willing to cite them
FAQs, comparisons and topic clusters instead of isolated blog posts

04

Measuring and reporting
Every month, in black and white, where you stand.
A fixed question set, asked again every month in ChatGPT, Perplexity, Google AI Overviews, Gemini and Claude
Per question: are you mentioned, which competitors are, which sources are cited
AI crawler visits in the server logs
A monthly report on what changed and what we do next

How we work

1
Baseline
Together we draw up the question set, record where you are mentioned today and scan your site for AI access and structured data.
2
Foundation
Technology and entity in order: crawlers, llms.txt, schema, Bing and IndexNow, consistent business details.
3
Answer content
New and rewritten pages around the questions from the baseline, with the answers up front.
4
Measure and adjust
The same questions again every month, compared with the baseline, and adjusting based on what works.

Who is GEO worth it for?

GEO is worth it for any business whose customers ask for advice before they choose, because a growing share of such questions now goes to an AI assistant first. According to OpenAI chief executive Sam Altman, ChatGPT had more than 800 million weekly users in October 2025. First, service providers and agencies benefit, for example when a buyer asks which agency in Rotterdam can build an app with React Native. Second, online stores and brands benefit when shoppers ask which product suits a specific need. Third, SaaS and B2B companies benefit when buyers ask for alternatives to a well-known tool. Finally, local businesses in the Drechtsteden and Rotterdam benefit when customers ask for the best supplier in the region. VirtuoLab starts every GEO project with a set of real customer questions, so the work focuses on the moments where a recommendation decides the sale.

Service providers and agencies: "which agency can …", "best … in Amsterdam"
Online stores and brands: "which product is best for …"
SaaS and B2B: "alternatives to …", "which software for …"
Local businesses with competition in their region

How we work

We write for readers, not for keywords: clear answers with facts and sources.
AI answers differ per question and per model. That is why we measure progress on a fixed question set, not on one lucky screenshot.
GEO is ongoing work: models and sources change, so we measure every month and adjust.

We do it for ourselves first

virtuolab.dev is a working example of this approach, because VirtuoLab applies GEO to its own website first. In October 2026, we measured the citability of nine core pages before and after the changes, on a scale from 0 to 100. The English median rose from 33 to 71, and the number of text blocks with the optimal length of 134 to 167 words rose from 0 to 71. First, the robots.txt gives OAI-SearchBot, PerplexityBot, ClaudeBot and Google-Extended explicit access. Second, an llms.txt and an llms-full.txt summarise the site for language models in English, Dutch and German. Finally, every page carries structured data that points to one Organization, with an address in Papendrecht and a Chamber of Commerce number. For example, the case study on the blog lists every improvement with the figures per page. Feel free to take a look:

See what we improved →Our llms.txt →Our robots.txt →

Frequently Asked Questions

How do you get found in ChatGPT?

Getting found in ChatGPT is a matter of three things at once: access, answers and confirmation. ChatGPT is a large channel: OpenAI CEO Sam Altman reported 800 million weekly users in October 2025. First, ChatGPT must be allowed to read the website. Sites opted out of the OAI-SearchBot crawler are not shown in ChatGPT search answers, according to OpenAI, so robots.txt must allow OAI-SearchBot. Second, the pages must answer customer questions directly, with the answer in the first sentence and concrete facts such as prices and lead times. Third, the business must appear consistently elsewhere, for example with the same name, address and services on LinkedIn and in Google Business Profile. After a robots.txt change, ChatGPT search needs about 24 hours to adjust. VirtuoLab then checks the progress with ChatGPT itself, using a fixed set of questions asked again every month.

What is llms.txt and do I need it?

An llms.txt file is a plain-text summary on a website, written for language models: who a business is, what the business does and which pages matter most. Jeremy Howard proposed the format in September 2024. First, the file is not an official standard: the proposal page describes the format as open for community input. Second, no special optimisations are needed to appear in AI Overviews or AI Mode, according to Google Search Central, and Google names no such file as a requirement. Still, the file costs little to create and gives AI assistants a reliable summary in one place. For example, the VirtuoLab website, built with Next.js, publishes both an llms.txt and an llms-full.txt file. VirtuoLab includes both files as standard in every GEO project. A summary file never replaces good pages, because the website content remains the source an AI assistant cites.

How do you measure AI visibility?

AI visibility is measured with a fixed set of questions customers would realistically ask an AI assistant. Step 1 is asking the same questions every month in ChatGPT, Perplexity, Google AI Overviews, Gemini and Claude. For each question, VirtuoLab records, for example, whether the business is mentioned, which competitors are mentioned and which sources are cited. A fixed set matters, because AI answers differ per question, per model and per moment. Step 2 is reading server logs: visits from crawlers such as OAI-SearchBot, PerplexityBot and ClaudeBot show whether AI systems read the site at all. Step 3 is checking the data through Google Search Console. Traffic from AI Overviews and AI Mode is included in the overall Web search traffic, according to Google guidance last updated in December 2025. Finally, each month ends with a report on what changed and what happens next.

How quickly do you see results from GEO?

GEO is a gradual process with three steps: technical changes can take effect within a day or so, while mentions in AI answers build up over months. Step 1 is technical access. ChatGPT search needs about 24 hours to adjust after a robots.txt update, according to OpenAI, and Perplexity gives a window of up to 24 hours for PerplexityBot. Step 2 is indexing. Microsoft Bing, Naver, Seznam.cz, Yandex and Yep receive a notice through IndexNow when a page is added, updated or deleted. New content must still be crawled, indexed and picked up in answers, and each assistant updates at its own pace. For example, OpenAI, Perplexity and Google each use a separate crawler. Finally, step 3 is mentions, which grow per question and per assistant. VirtuoLab therefore starts with a baseline measurement and repeats the measurement every month, 12 times per year, so progress stays visible.

Does GEO also make me visible in Google AI Overviews?

Google AI Overviews and AI Mode are part of GEO, and the groundwork is the same as for SEO. The feature reaches many searches: in a study of 900 US adults by Pew Research Center, 18% of Google searches produced an AI summary in March 2025. There are no additional requirements to appear in AI Overviews or AI Mode, according to Google Search Central. A page must be indexed and eligible to be shown in Google Search with a snippet. Google advises the familiar basics, for example allowing crawling in robots.txt, linking pages internally, keeping important information in text, and making structured data match the visible content. Additionally, no special schema.org markup is needed. Both features may use query fan-out, running several related searches across subtopics to build one answer. A page with clear answers to several subquestions therefore has more chances to be linked. VirtuoLab tracks AI Overviews in the monthly measurement and through Google Search Console, where the traffic appears within the Web search type.