Knowing how to show up in ChatGPT, Gemini, and Google’s generative AI answers is now part of marketing for any business that depends on being found. AI assistants cite pages they can access, understand, and confirm in other sources, and that is the job of GEO.
GEO stands for generative engine optimization. For Google, however, optimizing for AI search is still SEO, as its official guide to optimizing for generative AI features in Search, published in May 2026, states.
This guide shows how to show up in ChatGPT and Google in practice: GEO, SEO, and AEO, how sources are chosen, robots.txt and llms.txt, answer-first content, brand consistency, mentions, tracking, and a step-by-step plan for businesses that use AI.
Key Takeaways
- To show up in ChatGPT, Gemini, and Google, the starting point is a page that is open to each AI’s search crawlers and indexed, when these answers rely on web search.
- Google says GEO and AEO are still SEO, and that llms.txt, “chunking,” and special markup do not help in Search; original, useful content matters more.
- AI visibility is measured with official reports and with a fixed list of questions put to the assistants at regular intervals, with no promise of being cited.
Does your business show up when customers ask an AI?
ConverZAP’s free AI visibility X-ray asks an AI with web search 5 typical customer questions and shows whether your brand appears in the answers, with a score and what to fix on your site. No sign-up required.
Run the free X-rayTest the AI with my websiteWhat GEO Is and How It Differs From SEO
GEO, SEO, and AEO show up together in almost everything written about AI search. Understanding what each term describes keeps a business from paying twice for the same service and helps it prioritize what actually changes in how the site is run. For the full comparison of the three terms, see GEO vs SEO and where AEO fits.
GEO, AEO, and SEO: Three Terms, One Goal
SEO is optimization for traditional search engines, focused on rank and clicks. GEO targets AI-generated answers, such as those from ChatGPT and Gemini. AEO, short for answer engine optimization, is the name used for optimization aimed at direct answers to questions.
Google’s guide acknowledges both terms but says that, from Search’s perspective, optimizing for generative AI is optimizing for the search experience, which means SEO. The same guide advises caution with services sold as AEO or GEO.
What Changes in Practice for a Business
What changes is the format of the result. Instead of a list of links, people get a ready-made answer with links to the sources used. The competition becomes being one of those sources and being described accurately.
That is why GEO fits into a digital marketing strategy as a layer on top of SEO, not a replacement. Technical access, useful content, and reputation remain the foundation; what is new is tracking mentions and cited sources, not just rankings and clicks.
How AI Assistants Choose the Sources for an Answer
Assistants do not pick sources at random. When they search the web, they rely on an index, a ranking system, and a reading of the retrieved pages, which explains why SEO still matters for anyone who wants to be cited.
Search, Retrieval, and Query Fan-Out
A language model on its own answers with what it learned in training. To talk about recent facts, it uses information retrieval: it searches for pages, reads the relevant passages, and writes a response with links. Google calls this technique retrieval-augmented generation (RAG), also known as grounding.
According to Google, AI Overviews and AI Mode can also use query fan-out, where the model runs several related searches at the same time. A question about a lawn full of weeds, for example, triggers searches about herbicides, removing weeds without chemicals, and prevention.
Training vs Real-Time Search
A brand can enter an answer in two ways. The first is training, when the model learned something about it from older data. The second is real-time search, when the assistant queries the web and cites the pages it used.
For anyone figuring out how to show up in ChatGPT, real-time search is the most controllable path: it depends on the page being accessible to the right crawler, indexed, and clear. Training changes slowly, and site owners have no way to track its effect.
How to Show Up in ChatGPT, Gemini, and Google: Each AI’s Crawlers
Each AI company uses different crawlers with different jobs. Blocking the wrong crawler can remove a site from answers without anyone noticing, and allowing all of them without criteria can go against the company’s data policy.
Search Crawlers and Training Crawlers
Among the major AI assistants, OpenAI’s documentation describes three agents, among others. OAI-SearchBot powers ChatGPT search; GPTBot collects content that may be used to train its models; and ChatGPT-User visits pages at a user’s request.
According to OpenAI, sites that block OAI-SearchBot do not appear in ChatGPT search answers, although they may still show up as navigational links. At Perplexity, PerplexityBot surfaces and links sites in results and, according to the company, is not used to train models.
| Crawler | Company | What it does |
|---|---|---|
| OAI-SearchBot | OpenAI | Shows sites in ChatGPT search results |
| GPTBot | OpenAI | Collects content that may be used to train models |
| ChatGPT-User | OpenAI | Visits pages at a user’s request; robots.txt may not apply |
| PerplexityBot | Perplexity | Surfaces and links sites in Perplexity results |
| Googlebot | Crawls and indexes pages for Search, the basis of AI Overviews and AI Mode | |
| Google-Extended | Control token for use in Gemini training and grounding in Gemini Apps |
Gemini, AI Overviews, and Google’s Index
According to Google’s page on AI features, to appear in AI Overviews and AI Mode a page must be indexed and eligible to be shown with a snippet; the 2026 guide adds that the site must be included in Search generative AI features in Search Console.
The Google-Extended token controls whether content can be used to train Gemini models and for grounding in Gemini Apps. It does not affect inclusion or ranking in Search; blocking it only removes the site from those Gemini uses.
Technical Access: robots.txt, Indexing, and llms.txt
With the crawler map in hand, the next step is to check what the site allows. An old robots.txt file or a misconfigured security plugin can block AI crawlers without warning. The guide to llms.txt explains the format and whether the file is worth creating.
robots.txt, noindex, and Snippet Controls
The robots.txt file sits at the root of the domain and defines what each crawler can crawl. The minimum is to allow Googlebot, OAI-SearchBot, and PerplexityBot on public pages. OpenAI says changes to the file take about 24 hours to take effect.
Beyond robots.txt, the noindex, nosnippet, data-nosnippet, and max-snippet directives limit what Google shows, including in AI features. A page with noindex or nosnippet applied by mistake stays out of AI Overviews, even if the content is excellent.
A basic access audit checks three things:
- robots.txt: AI search crawlers allowed on public pages.
- Indexing: key pages free of noindex and listed in the sitemap.
- Text in the HTML: main content available as text, not only in images.
What llms.txt Is and What It Does Not Do
llms.txt is a proposal by Jeremy Howard, published in September 2024: a Markdown file at the root (or in a folder) of the site, with the project name, a summary, and lists of links, to help language models use the site.
Google, however, says Search ignores llms.txt and similar files: creating one neither helps nor hurts visibility on Google. The file may serve other services that read it, but it does not replace robots.txt, indexing, and useful content.
Answer-First Content: The Answer Comes First
Answer-first is a format in which the direct answer appears at the start of the page or section, followed by the details. It serves readers in a hurry and makes it clear which passage answers each question, without requiring text written only for bots.
How to Structure an Answer-First Page
The structure starts with one sentence that answers the question in the heading, in plain language. Then come the context, the steps, the numbers with sources, and the exceptions, as a good help center or a WhatsApp API glossary does.
Microsoft recommends clear headings, tables, and FAQ sections to make it easier for Bing’s AI to use content accurately. Google, on the other hand, says there is no need to rewrite text just for AI or to break content into chunks.
A quick test is to see whether an AI can answer using only the text on the page. ConverZAP’s free tool Test the AI with your website shows how an agent would answer customers using only the text of the page you enter.
To see whether the brand already shows up in answers, the free AI visibility X-ray asks an assistant with web search 5 typical customer questions and shows in how many answers the brand was cited.
Original Content, First-Party Data, and Cited Sources
Google gives weight to non-commodity content: a unique point of view, first-hand experience, and analysis beyond common knowledge. Summaries of what already exists online, or of what an AI could generate on its own, get lost among similar sources.
The GEO study, presented at KDD 2024, tested text changes on a benchmark and found that citing sources and adding quotations and statistics raised visibility by 30% to 40%, in relative terms, on one of the metrics. Stuffing the text with keywords brought little or no improvement.
Tools and first-party data follow the same logic. A WhatsApp API cost calculator or a WhatsApp Business API guide with rates and rules checked against the official source are examples of pages that answer real questions with verifiable information.
Entity Consistency: AI Needs to Know Who the Brand Is
To cite a business, the system needs to recognize it as a single entity, without confusing it with another one that has a similar name. The same name, description, address, and contact details on every channel make that identification easier and prevent answers with outdated data.
The Same Name, Description, and Details Everywhere
Consistency starts on the site itself: the About page should state, in text, what the business does, for whom, where it operates, and how it works. The same short description should appear on social media, in directories, and on the Google Business Profile.
Small deviations cause confusion. Different spellings of the brand, old phone numbers, and outdated addresses scattered across the web create competing versions of the same business. On WhatsApp, the WhatsApp Business display name should also follow the official spelling used on the site.
For local businesses, Google says businesses with complete and accurate information in their Business Profile are more likely to show up in local search. The generative AI guide also mentions the Business Profile as a way to give products and services visibility.
Organization Structured Data and sameAs
Organization structured data describes the business in code: name, logo, website, contact details, and profiles. According to Google’s documentation, this markup helps the search engine understand the organization’s administrative details and tell it apart from others in results.
The sameAs property lists URLs of the company’s profiles on other sites, such as social networks and review sites. Google notes that structured data is not required for AI search, but it must match the visible text.
Off-Site Mentions and Brand Reputation
AI answers are not limited to the company’s own site. Google says its AI features can show what people say about products and services in blogs, videos, and forums, which brings off-site reputation into GEO.
Where Mentions Count: Reviews, Press, and Communities
For the Business Profile, Google explains that local prominence considers how many sites link to the business and how many reviews it gets. Real reviews, answered with care, also help customers decide and shape what is said about the brand.
A practical way to generate reviews is to ask for feedback right after the conversation, with a WhatsApp customer satisfaction survey. Interviews with industry press, events, and helpful answers in communities round out the reputation.
Authentic Mentions vs Risky Shortcuts
Google considers chasing inauthentic mentions unhelpful: its ranking systems prioritize quality content, and other systems block spam. Buying reviews, creating fake testimonials, or spreading undisclosed paid posts exposes the brand and does not bring lasting visibility.
Another risky shortcut is hiding instructions for bots on the page to try to push the AI into recommending the brand. The technique resembles prompt hacking and goes against Google’s guidance on helpful content, which is meant to help people, not to manipulate results.
How to Measure Brand Visibility in AI Answers
Without tracking, knowing how to show up in ChatGPT becomes a bet. Combining official reports with your own question tracking shows where the brand appears, which pages are cited, and what is missing. The guide on how to measure AI visibility covers metrics, questions by funnel stage, and official reports in detail.
Official Reports: Search Console and Bing Webmaster Tools
In June 2026, Google launched generative AI performance reports in Search Console, rolled out to all sites on August 31. They show impressions in AI Overviews, AI Mode, and Discover, with pages, countries, devices, and dates.
In February 2026, Bing Webmaster Tools launched the AI Performance report in public preview. It shows citations in Copilot and in Bing’s AI summaries, the cited pages, and the grounding queries that led to the content.
Question Tracking and Share of Voice
Official reports do not break out ChatGPT or measure mentions without a link. For that, the business builds a fixed list of questions a customer would ask, such as “best med spa in Denver,” and asks the assistants the same questions at regular intervals.
Each round records whether the brand was cited, in what position, and alongside which competitors and sources. The total becomes share of voice, the share of answers that include the brand. Questions written the way a real customer would phrase a prompt keep the sample faithful.
At ConverZAP, the AI Visibility module asks up to 50 questions per week, with web search, to ChatGPT, Gemini, and Claude (whichever have an API key registered), compares the brand with up to 15 competitors, and audits the site for AI crawlers.
The module is part of the Completo plan, described on the pricing page, and uses the company’s own API keys. Because API answers can differ from those in the apps, the trend over several weeks is worth more than a single round.
GEO Step by Step for Small and Midsize Businesses
Small and midsize businesses do not need a long project to get started. A two-phase plan, first access and diagnosis, then content and reputation, covers the essentials with the team already in place.
Week One: Access and Diagnosis
GEO starts by removing blocks and recording the baseline, before touching any copy. The first week follows this order:
- Check robots.txt: allow Googlebot, OAI-SearchBot, and PerplexityBot, and decide on GPTBot and Google-Extended as a matter of policy.
- Verify the site in Search Console and Bing Webmaster Tools: check inclusion in generative AI features and the reports.
- Review indexing: home, services, and contact pages without noindex or nosnippet.
- List 20 to 30 customer questions: use the questions that reach customer service.
- Run the first measurement: note mentions, competitors, and sources for each assistant.
The most useful questions come from customer conversations. Small businesses that already use AI in customer service have a conversation history that shows, in the customer’s own words, what people want to know before they buy.
The First 90 Days: Content, Entity, and Reputation
Once the diagnosis is done, the next three months focus on content, entity, and reputation:
- Rewrite the main pages in answer-first format: services, terms, and FAQ.
- Create one clear page per service or audience: without copying the same text for every search variation.
- Standardize the entity: About page, Organization with sameAs, Business Profile, and social profiles with the same description.
- Publish something original every month: first-party data, a case study, or a tool.
- Repeat the measurement every week: compare the trend and adjust pages that are not cited.
Google warns that creating near-identical pages for every search variation, just to manipulate answers, violates its scaled content abuse policy. Templates such as industry pages help, and the rest fits into digital marketing automation.
The Limits of GEO and What to Expect in the Coming Years
GEO is a new field, with rules that change fast and many unfounded promises. Knowing what you cannot control matters as much as knowing what to do, especially when hiring tools or services.
No One Can Promise a Citation
Google says that meeting every requirement does not ensure crawling, indexing, or display, and warns that no third-party tool has access to its ranking or AI systems. Anyone researching how to show up in ChatGPT will find “first place” promises that deserve suspicion.
The GEO study measured gains on a controlled benchmark, not across every assistant on the market. Answers can also change from one query to the next, so a single question says little about the brand’s real visibility.
AI Agents and the Future of Search
Google’s guide already includes a section on AI agents, systems that carry out tasks on people’s behalf, such as booking an appointment or comparing products. Browser agents can read the page, its code structure, and the accessibility tree.
That reinforces the basics: a fast, accessible site with clear, up-to-date information. Keeping up with artificial intelligence trends helps a business notice when a new assistant starts to matter in its industry.
For businesses that want to start without a long project, ConverZAP combines tracking of AI answers with a free test that shows how an agent reads the company’s website. The AI visibility solution explains how the weekly rounds and the site audit work.
Frequently Asked Questions
The basic steps cost nothing: allow OAI-SearchBot in robots.txt, keep pages indexed, write direct answers, and take care of reviews. Paid tools help track progress, but they are not a requirement for the brand to be found.
There is no official timeline. Once access is open, a citation depends on the page being found, read, and chosen over other sources, which varies by question. That is why it makes sense to compare several weeks of tracking instead of waiting for a specific day.
There is no official confirmation. OpenAI’s crawler documentation does not list llms.txt as a requirement for ChatGPT search, and Google says Search ignores the file. It may serve tools that read it, but the foundation is still access and content.
SEO measures rank and clicks in a list of links; GEO measures whether the brand is cited and described correctly in AI answers. The techniques overlap a lot, and for Search, Google treats both as SEO.
No, according to OpenAI, because ChatGPT search relies on OAI-SearchBot. The GPTBot decision is about data policy, not visibility: the business chooses whether it accepts that its public content may help train models.
For Google, the generative AI report in Search Console shows impressions. For ChatGPT and Gemini, the way is to repeat a list of customer questions at regular intervals and note mentions, competitors, and sources, by hand or with a tool.
