Updated on 7 November 2026, by João Leite, Founder of murmur.

Explain GEO to your boss or the board through the result: making your company appear in ChatGPT when the buyer asks for a recommendation. In the conversation, bring a number: in how many answers to your buyer’s questions the company appears today, and who appears in its place. Ask for a trial period, with the monthly cost, what counts as success and the date to review the number. Do not promise a position in the answer, a fixed date or a financial return. The AI answer changes with every question, and most of the effect happens without a click.

  • Explain through the result: GEO is making the company appear in ChatGPT, Claude and Google AIO, Google’s AI summary in search.
  • The buyer already asks AI models: ChatGPT has more than 1 billion users per week, according to OpenAI.
  • Bring today’s number: in how many answers the company appears, out of how many, in which AI, on what date and who appears in its place.
  • Ask for a test, not a promise: cost per month, success criterion, stop criterion and review date, written down in advance.
  • Most of the effect happens without a click: fewer than 7 in every 100 questions to ChatGPT bring any source citation, according to Similarweb.

murmur, a GEO agency in Brazil, places brands in the answers of ChatGPT, Claude, Google AIO and other AI agents. For the boss, this work comes down to one number, the brand’s presence. Presence is the number of answers in which the brand appears, divided by the total number of answers. From the diagnostic onward, a dedicated specialist studies the brand’s sector and shows who the AI models recommend in its place. They also take care of the content that seeks to change that number.

Explain to your boss or your partner, in two sentences, the work of appearing in ChatGPT

Two sentences explain GEO to the boss: the buyer already asks ChatGPT what to buy, and GEO is making your company appear in that answer. The first shows that the change has already happened on the client side, and the second says what is being bought, without technical jargon. Neither of them asks the boss to understand how AI works.

A demonstration convinces more than a definition. Ask ChatGPT, in front of your boss, the question your client would ask before buying, and read the names in the answer together. Then show that the same question almost always brings a different list in the next conversation. That is why the decision rests on a number taken from many questions, and not on one screen.

If the boss asks about the acronym, GEO comes from Generative Engine Optimization, a name born in an academic paper from November 2023. Google AIO is what we call the AI Overview: the AI text that Google puts at the top of many searches, before the list of links. The full definition, with the path each AI takes to the answer, is in the guide on what GEO is.

What the boss needs to know about AI answers before deciding

Before deciding, three points matter: the buyer already asks AI models, many companies still do not take care of this, and the answer changes every time. In August 2026, OpenAI counted more than 1 billion weekly active users on ChatGPT. In the same month, OpenAI reported that Brazil is among the three largest markets for ChatGPT by weekly users. The number of users in the country nearly doubled in a year.

Many Brazilian companies still do not take care of their presence in AI answers. RD Station’s Panorama de Marketing e Vendas 2026 (Marketing and Sales Panorama 2026) had 1,351 responses. In it, 36% say they do not optimize content for ChatGPT, Claude and Gemini, and 23% cannot say. But the topic is rising on the agenda: in a survey by 8D Hubify with PipeLovers, with 214 professionals, GEO went from 8.9% to 30.4% of marketing priorities for 2026.

The AI answer changes every time the question is asked, and there is no fixed position to conquer. The 600 volunteers in a study by SparkToro, an American audience research company, ran 12 questions in ChatGPT, Claude and Google’s AI, 2,961 times in total. Two answers from ChatGPT or Google’s AI brought the same list of brands in fewer than 1 case in 100. That is why you track the brand’s presence across many questions, repeated in rounds, and not the position in one answer.

Position in the answer, date and financial return are not promised: the list of brands changes from one answer to another, the timeline depends on the starting point and the competition, and most of the effect happens without a click. What you bring to the boss instead of the promise is the effect that research has already confirmed. Olivier Martinez’s review of 45 studies came out in July 2026 and shows that changing a page the search engine has already found changes the citation or the use the AI makes of it.

When the boss wants to stop investing in content because of AI

Stopping investment in content because of AI interrupts what feeds the answers of ChatGPT and Google AIO. In Google search, the boss is right about the click. Pew Research Center compared searches with and without the AI summary in the history of 900 adults in the United States, in March 2025. With Google’s AI summary on the screen, only 8% of visits had a click on a regular link; without the summary, it was 15%.

Google’s AI summary is written with the pages that search finds. For Google, a page only enters the search AI features if it is in the index and can appear in regular search with a snippet. The index is the list of pages Google has already read and stored. According to OpenAI’s help center, ChatGPT can search the web on its own when the question asks for current information. A company that stops publishing leaves the next chances to appear in ChatGPT to its competitors.

The buyer also checks what the AI says. In January 2026, 1,862 technology buyers answered a TrustRadius survey. Of the total, 63% used AI at some point in the purchase, and 94% of those check the answers at least sometimes. Those who check look for another source, and that is why the company needs its own pages that answer the buyer with facts. More 2026 surveys with buyers are in the guide on how the B2B buyer finds a supplier through AI.

Among murmur’s clients, new content, published on the client’s own website, changed what the AI models answer. murmur produced the FotoRegistro guide, a photo printing e-commerce, and published it at a URL on the store’s website. ChatGPT cited that guide as a source.

Visio, operations software for retail, is a murmur client. On 12 July 2026, part of Visio’s new pages were already in the Google index. In July, across the same 135 questions, Google AIO already showed Visio in 45% of answers, against none in June. Appearances in Claude rose from 1% to 7% of answers, and those in ChatGPT, from 2% to 22%. Why content remains the basis of the work with AI models is in the guide on GEO and content marketing.

What to bring to the boss on one page: today’s number, the cost, the criteria and the review date

Bring the boss a single page, with today’s number, the cost, the success and stop criteria and the review date. The page needs to make sense without you in the room. According to Forrester, an American research company, the typical business-to-business purchase decision involves 13 people from inside and nine from outside. How the buying committee asks, person by person, is in how to do GEO for a B2B company.

  1. Today’s number: in how many answers to the buyer’s questions the company appears, out of how many, in which AI and on what date.
  2. The cost: the amount per month and what it includes.
  3. The success criterion: the KPI, the number that says whether the work is delivering results. Here, in how many of the same questions the brand needs to appear.
  4. The stop criterion: what ends the test if the number does not move, written down before starting.
  5. The review date: when the number is measured again, on the same questions.

To build today’s presence number, follow the guide on como medir a presença da marca no ChatGPT (how to measure the brand’s presence in ChatGPT). The questions where the competitor appears and your company does not carry more weight, because they show who the buyer hears instead of your company. If the boss asks why the company does not appear, the causes are in the guide on por que o ChatGPT não recomenda a sua empresa (why ChatGPT does not recommend your company).

The request to the boss is a trial period, not an endless budget. In Forrester’s 2026 research, more than 60% of business buyers use a trial period to evaluate a solution.

The monetary return is not included as a projection, because most of the influence of AI models happens without a click. What is counted, and what is not counted, when the company does not appear is in the guide on quanto sua empresa perde por não aparecer no ChatGPT (how much your company loses by not appearing in ChatGPT). ROI, the return on investment, compares the money a piece of work brought in with what it cost. Its calculation and market prices are in the guide on quanto custa uma agência de GEO (how much a GEO agency costs).

What finance tears down in a slide about AI, and what to put in its place

Finance tears down five common claims in a slide about AI, and each one is replaced by a fact it can check. The five are a percentage without a base, mentions added together, a revenue projection, a promised timeline and a superlative.

what usually goes on the slidewhy it falls apartwhat to say instead
“we have 30% visibility in AI models”the percentage without the number of questions cannot be checked, and 30% of 10 questions says much less than 30% of 300“the brand appears in X of Y of the buyer’s questions, in ChatGPT, on such date”
“we are cited 200 times”the sum mixes questions that already carry the brand name with questions from someone who does not know it yet, and the first ones inflate the totaltwo separate numbers: the questions with the brand name and the questions without it
a revenue or ROI projectionmost of the effect happens without a click, and the buyer who read the name in the answer arrives later, through search or by typing the site addressthe cost of the test and the success criterion, written down beforehand
a promised timeline (“in 90 days we will be in the answers”)the AI answer changes with every question, and the timeline depends on the starting point and the competitionthe average timeline of those who do the work, stated as an average, and the date of the next measurement
a superlative (“we are leaders”, “we were the first”)a board member disproves it with a question to ChatGPT, on the spotwhat the company does and the number that supports it

The five swaps make the slide about AI harder to tear down, because every number can be checked. The question set changes what the percentage means. Otterly.AI, which sells an AI monitoring tool, measured one brand across groups of questions randomly drawn from a list of 320, over 30 days. In groups of 10 questions, 9 out of 10 presence measurements fell between 39% and 71%; in groups of 100, between 51% and 60%.

Every number on the slide carries the value, out of how many questions, the date, what was counted as appearing and the link to the proof. murmur’s case studies carry the client’s name, the number of questions and the period of each measurement. One of these case studies is Mindo, a motion design studio in São Paulo. In July 2026, Mindo appeared recommended by ChatGPT in 23% of the 134 questions in its market. In June, on the same questions, it had been 12%.

A screenshot of an answer illustrates the slide, but it does not prove presence, because the same question, asked again, almost always returns another list of brands. The guide An AI answer screenshot proves an answer, not brand presence (An AI answer screenshot proves one answer, not the brand’s presence) separates the screenshot that helps from the screenshot that misleads.

If appearing in ChatGPT turns into leads and sales: the answer in three parts

Answer about leads in three parts: most of the effect comes without a click, the click shows up in Google Analytics and the work delivers presence. In May 2026, fewer than 7 out of every 100 questions asked to ChatGPT had any source citation, by Similarweb’s count. The buyer keeps the name read in the answer and reaches the company later, through search or by typing the site address.

The click that comes from AI models is recorded: in Google Analytics 4, the AI Assistant channel gathers visits from ChatGPT, Gemini and other AI models. The channel leaves out the AI Overview and AI Mode, the conversational mode of Google search. Across the 2,425 Brazilian sites in Leadster’s overview, with 2025 data, visits coming from AI converted at a median of 7.80%, and those from organic search at 2.39%.

GEO delivers what the click does not show: in how many AI answers the brand is, next to each competitor, round after round. The sale is tracked in Google Analytics and in the CRM, the system where the sales team records each negotiation. To check any AI traffic number before taking it to the boss, use the guide on the reliable metric for the brand’s presence.

The timeframe for appearing in ChatGPT goes into the request as an average, never as a date

On average across murmur’s clients, significant improvement came from 60 days onward, depending on the starting point and the competition. The first citation can come earlier. At FotoRegistro, a murmur client, the first recorded citation of the store’s guide in ChatGPT came 30 days after launch.

In the request to the boss, no date goes in as a promise: what goes in is the date when the number is reviewed, on the same questions as the first measurement. The guide How long it takes to show up in ChatGPT: from first citation to presence in answers explains why one company takes longer than another to reach the first citation.

What murmur tells the boss: who does it, how much it costs and how to leave

At murmur, a dedicated specialist takes care of the brand, the work starts at R$ 5,000 per month and the contract has no minimum term. The content comes from the study of the client’s sector. Ongoing monitoring shows the brand’s presence next to that of competitors, in a proprietary software dashboard, included and not sold separately.

The value of murmur’s work changes according to the number of business lines, languages and brands, and according to complexity. There is no cap on questions, hours or pieces, and market prices are in the guide on how much a GEO agency costs.

murmur’s contract has no minimum term, and the company can end it from the first month. That is why the test’s stop criterion applies on the date the number is reviewed. The case study pages for Mindo, FotoRegistro and Visio stay open for the boss to check before saying yes.

Frequently asked questions

We already pay an SEO agency. Isn’t that the same work?

It is not the same work, although SEO and GEO partly overlap. The SEO agency takes care of the site’s position on Google. Ask it for the number with the math shown: the total number of questions and answers, in how many answers the brand appeared, in which AI model and on what date. Also ask for a published case study, with the client named. murmur works on the brand’s presence in those answers: it produces the content that answers the buyer and tracks in how many answers the brand appears, and in murmur’s case studies the client appears with name, numbers and dates. The guide on GEO and SEO shows where the two kinds of work meet and where they separate. To split the budget with SEO, see how to add GEO without cutting SEO.

Can it be done with the in-house team, without hiring an agency?

The recommendation to take to the boss is to hire a GEO agency, such as murmur, instead of leaving the work only in the hands of the in-house team. What supports the request to the board comes from murmur’s data engineering: in-depth information about the company, a comparison with competitors and reliable measurement, plus the result that follows. Without data or measurement, an effort from inside the company alone can help in part, but it is unlikely to reach the same result. To choose the agency, see what to ask before hiring a GEO agency.

If the AI model changes, do we lose what was done?

There is no fixed position to lose: the same AI model already changes the list of brands from one answer to the next, with the same question. In the model change of May 2026, 47% of ChatGPT citations in German went to other sites within 48 hours, in a correlation measured by SISTRIX that counts sites, not brands. That is why presence is measured again in every round. The extra round after a model change is in the guide on how often to measure the brand’s presence in AI models. The published pages remain on the company’s site.

If we stop after a few months, do we lose the results?

The pages stay with the company: at murmur, the content all belongs to the client, even after the contract ends. It was published at a URL on the company’s own site, not on murmur’s. Presence in the answers is not stored as a position, because the AI model does not repeat the same list of brands from one answer to the next.

Bring the boss a picture of your market in AI answers

The picture of your market in AI answers comes from murmur’s free diagnostic, with no sign-up and in less than two minutes. You enter your website and see how AI models answer people looking for what your company sells. To see who AI models recommend in other sectors, there are murmur’s sector studies. Each study reports the number of questions, the AI models consulted and the month, and shows a single round. With the picture in hand, the conversation about appearing in ChatGPT starts from the answers your buyer receives today, not from an acronym.

Read also

References