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

The loss from not appearing in ChatGPT is measured in answers, not in reais: the buyer questions where your company is missing. The sale that went to the competitor does not pass through the company’s website and does not become a number in any report. Two other numbers complete the picture. One is the visits that AI models already send to the site, in Google Analytics. The other is the click that Google search stopped delivering when it shows the AI Overview, the AI-written summary at the top of the page.

  • The loss is counted in answers: in how many answers to buyer questions the company appears, and whom the AI recommends in its place.
  • Visits from AI models are in Google Analytics 4, in the AI Assistant channel, which shows only those who clicked a link.
  • The missing click is measured in Google search. In the US, the Pew Research Center found a click on a regular link in 8 out of every 100 visits with an AI summary, against 15 without it.
  • The sale that went to the competitor does not appear in any report, and that is why no loss figure in reais holds up.
  • On leads, the answer has three parts: the effect without a click, the click in Google Analytics and the presence in answers.

murmur is a GEO agency in Brazil that places brands in the answers of ChatGPT, Claude, Google AIO and other AI agents. Google AIO is what murmur calls the AI Overview. Its work starts with the answers where the company is missing. A dedicated specialist studies the company and the sector and sees, question by question, whom the AI models recommend in its place. murmur produces the content that fills these gaps and tracks, week after week, in how many answers the brand appears alongside competitors.

The loss from not appearing in ChatGPT has no value in reais

The loss from not appearing in ChatGPT has no value in reais because the buyer who read another name in the answer leaves no trace. They ask the question, read the companies recommended, look for one of them and close. The company that was left out receives no visit, no form and no call, and never finds out that this buyer existed. Every loss calculation in reais depends on a number that nobody observes: how many buyers read an answer without the company and bought from another one.

Olivier Martinez reviewed 45 studies on GEO, the work of making a brand appear in AI answers, published from November 2023 to July 2026. For the author, content the AI has already found can change the answer.

When a ready-made loss calculation shows up, in a proposal or on an agency page, ask where each number came from: the lost leads, the traffic and the conversion. The four questions to check an AI traffic number also apply to these calculations.

Instead of a value in reais, the company counts three numbers, each in a different place. The sale is tracked in the CRM, the system where the sales team records each contact and each deal. Each number has a blind spot:

what is countedwhere it is readwhat the number does not say
in how many answers the company appears, and who appears in its placein the buyer questions, asked to AI models in roundsa fixed position in the answer, which the AI does not have
the visits that come from ChatGPT and other AI modelsin Google Analytics 4, in the AI Assistant channelwho read the answer and did not click
the click that Google search stopped deliveringin Search Console, when the page keeps its position and loses clickswhat happens in ChatGPT, outside Google search
the sales from those who arrived through an AI modelin the CRM, with the source of each contactthe sale that went to the competitor

The three numbers are read together, with the CRM sale alongside, and none of them shows the sale that went to the competitor. That is the number missing from every loss calculation in reais.

The first number: in how many AI answers your company appears, and who appears in its place

The first number is presence: in how many answers to buyer questions your company appears, and whom the AI recommends in its place. Presence is how many answers the brand appears in, divided by the total number of answers. It is the number closest to the loss, because each answer without your company shows what a buyer read in its place. The questions with the competitor in the answer and your company out are the ones that count most.

Presence is valid for one AI model, one list of questions and one period, and it can only be compared with the same basis. The AI answer changes from one conversation to another, and that is why there is no fixed position to lose: the number adds up many answers, collected on different days. How many questions the list asks for, so the number does not swing for no reason, is explained in the guide on how many questions to use to measure brand presence in AI models. To build the spreadsheet, follow the step-by-step guide on how to measure brand presence in ChatGPT. Anyone who wants to appear in these answers starts with how to make your company appear in ChatGPT, covering what to do on the site and outside it.

Mindo, a motion design studio in São Paulo and a murmur client, shows what this number looks like when the company is out. On 29 May 2026, murmur measured Mindo in three AI models. Mindo did not appear in any of the 63 answers, in any of the three AI models. The AI crawlers found Mindo’s page empty, because the site was an application that only renders in the browser. The guide that murmur wrote for Mindo was published at guia.mindo.com.br and was fully live on 22 June 2026. On 13 and 14 July 2026, ChatGPT put Mindo in first place in 16% of the 134 questions in its market, against 3% on 25 and 26 June.

The second number: the visits that already come to the site from AI models

The second number is the visits that already arrive from AI models, and Google Analytics 4 groups them in the AI Assistant channel. According to Google Analytics help, the channel gathers sources such as ChatGPT, Gemini, Deepseek, Copilot or Grok. Google Analytics help says the AI Assistant channel excludes Google’s AI Overviews and AI Mode. AI Overview is Visões gerais criadas por IA (AI-generated overviews) in Portuguese. AI Mode is the conversation with the AI inside Google search. Anyone who clicks a link from these two features enters the report as an organic search visit.

The number in the AI Assistant channel is a floor, not the total. Anyone who read the company name in an answer and came back later, typing the brand into Google or the site URL, appears as organic search or direct access. An empty channel does not mean an absent buyer: it means that Google Analytics did not recognize any visit coming from a link in the answer. The guide on how to measure presence includes the step-by-step guide to find AI visits in Google Analytics.

Leadster, which sells lead qualification tools, compared conversion by channel across 2,425 Brazilian websites, with 2025 data. Those who arrived through AI models converted at a median of 7.80%, and those who came from organic search, at 2.39%. For Leadster, conversion is any relevant action by someone visiting the site, from a quote request to a sign-up. The page does not say how much of the traffic came from AI models or how that source was recognized. The number is about those who arrived, not about those who read the answer and stayed in it.

The third number: the click that Google search stopped delivering

The third number is the click that Google search stopped delivering to sites when it shows the AI Overview above the links. The studies that measure this click are all about Google. Elizabeth Reid, Google’s vice president of Search, wrote in May 2026 that the number of searches hit a record in the last quarter. What decreased was the click on some of them.

In ChatGPT, the calculation is different. There the company has no position from which the click comes: the loss is the answer where it is missing, the first number. The Google click studies do not apply to ChatGPT.

The Pew Research Center tracked 900 American adults in March 2025, and their Google searches totaled 68,879. The analysis compared visits to the results page with and without the AI summary. With the summary, a click on a regular link came in 8 out of every 100 visits; without it, in 15. Links inside the summary itself received a click in 1% of visits with the summary.

Google’s AI Overview does not appear in every type of search. Seer Interactive, a search marketing agency, tracked 53 brands from January 2025 to February 2026, and writes that the data shows a trend. In 49,358 searches from February 2026, the AI Overview appeared in 36% of informational ones, in 8% of commercial ones and in 5% of transactional ones. Transactional, in Seer’s classification, are purchase searches, those whose results page shows a map, route, local result, store, jobs or events.

Being cited in Google’s AI Overview changes the brand’s calculation. In informational searches, Seer counted 120% more clicks per impression for the brand cited in the AI Overview than for the one not cited. In informational searches, the click of the cited brand remained 38% below that of a search without the summary. An impression is each time the page appears in a search result.

Search Console is Google’s dashboard for the site. Its generative AI report, opened to all sites on 31 August 2026, counts impressions. Each impression is one time the site’s links appeared in Google’s AI features. An impression is not a visit. The report shows that the link appeared, not that someone visited the site.

Click studies talk about other sites and other audiences, and your company’s calculation comes from its own Search Console. A page that kept its position and lost clicks is the sign of the AI summary. The step-by-step instructions are in the guide on organic traffic that fell because of AI.

What is not measured: the leads that went to the competitor

The leads that went to the competitor are not measured, because anyone who read another name in the answer never reached your site. The question about how many leads AI models bring, on the other hand, has an answer, in three parts.

  1. Most of the effect comes without a click. In Similarweb’s May 2026 data, fewer than 7% of questions to ChatGPT brought any cited source. Without a link in the answer, that influence does not pass through Google Analytics.
  2. When there is a click, it stays in the company’s Google Analytics. The visit lands in the AI Assistant channel, and the lead source can carry over to the CRM. In the median of the 2,425 Brazilian sites in the Leadster study, in 2025, those who arrived through AI models converted more than those who came from organic search. A “how did you hear about the company” field in the form, with ChatGPT among the possible answers, captures part of those who arrived without clicking.
  3. Presence counts what the click leaves out. It counts in how many answers the company appears, out of the total answers, side by side with each competitor, round after round. That is the part murmur delivers, read alongside Google Analytics and the company’s own CRM.

murmur writes the pages for the questions where the company is missing, and tracks presence

murmur writes the pages that answer the questions where the company is missing and tracks, week after week, the brand’s presence. Each page goes live at a URL of the brand itself, a subdomain such as Mindo’s guia.mindo.com.br, and the content is all the client’s, even after the contract ends. A dedicated specialist, the same person from the diagnostic onward, closely tracks the result. The dashboard, included in the service, puts the brand’s presence side by side with the competitors’, on the same questions, round after round.

murmur’s published case studies show the result with name and date. At Mindo, ChatGPT recommended the brand in 12% of the 134 questions in its market on 25 and 26 June 2026. On 13 and 14 July, it recommended it in 23% of them. At Visio, operations software for retail, the case study page counts the appearances in the same 135 questions, from June to July 2026. In Google AIO, Visio went from 0% to 45% of answers; in ChatGPT, from 2% to 22%; in Claude, from 1% to 7%. FotoRegistro, a photo printing e-commerce, and Fly Vet, a veterinary marketing agency, are also clients with a published case study.

The click and the sale that come from AI answers appear in Google Analytics and in the company’s own CRM. Alongside them, murmur’s dashboard shows the brand’s presence, the part the click does not show.

Frequently asked questions

If my audience does not use AI yet, am I losing anything?

The company loses less, and only it finds out how much: by asking the buyer. Ask ChatGPT the questions your customer would ask before buying and ask new customers how they heard about the company, with AI models among the options. An AI Assistant channel with no visits does not prove that the audience does not use AI, because it only shows those who clicked. To learn how much Brazilians already use AI, broken down by income and education, see is it worth investing in GEO now.

Is it worth investing in GEO now?

It is worth investing in GEO when the answers to your buyer’s questions point to competitors and leave your company out. In the market, this work of getting into the answers of ChatGPT and the other AI models is called GEO. The cases where it is worth waiting are in the guide on is it worth investing in GEO now, along with the cost of waiting.

Can you calculate the return on GEO in reais?

The return on GEO in reais cannot be calculated before starting, because most of the influence of AI models happens without a click. A number in reais would be a guess. What is written down beforehand is the cost of the test, the success criterion and the stop criterion. The guide how to explain GEO to your boss or the board shows how to put that request together. Agency prices and the ROI math, which compares what the work brought with what it cost, are in the guide on how much a GEO agency costs.

Why doesn’t my company appear in ChatGPT?

The cause changes from company to company, and each one has a test that is done before any fix. The test for each cause, and who in the company solves each one, comes in why ChatGPT does not recommend your company. The mechanism behind them comes in why AI recommends some companies and not others.

How do I take this number to my boss?

Take today’s presence, with the sample and the date, and the names that appear in place of the company. The script for the conversation, with the test request and what finance shoots down in a slide, is in the guide on how to explain GEO to your boss or the board.

Before calculating the loss, see what your buyer reads in the answers

Before calculating any loss, ask ChatGPT your buyer’s questions and note who appears in each answer. Repeat the list on another day, because the answer changes, and count in how many your company came up. Without building the list by hand, murmur’s free diagnostic gives a first reading based on your site. With no sign-up and in less than two minutes, it shows how AI models answer the questions of the market in which your company sells. To see which brands AI models indicated in other markets, in a round of questions by sector, there are murmur’s sector studies.

Read also

References