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

When organic traffic has dropped, the first thing to do is compare, in Google Search Console, the position and the clicks of the pages that dropped. Same position with fewer clicks means the AI Overview took the click. The AI Overview is the AI-written summary that Google shows at the top of many searches, before the links. A falling position means a problem on the site itself or in the content, and the fix is on the page. When the cause is the AI Overview, the work is to get the brand cited inside the answer, in Google AIO and in ChatGPT.

  1. Compare position and clicks in Search Console, page by page, with the same period of the previous year.
  2. Same position and fewer clicks is the AI Overview: Ahrefs measured a click-through rate 58% lower on the first result when the summary appears.
  3. A falling position is the site: too many pages with no value for the reader, a technical problem or a Google update.
  4. Work to be cited inside the summary: in informational searches, the brand cited in the AI Overview receives 120% more clicks per impression than when it is left out, in Seer Interactive’s measurement.
  5. Measure the brand’s presence in AI answers, beyond the click, because Search Console does not separate the traffic that comes from the AI Overview.

How to tell whether organic traffic dropped because of the AI Overview

Organic traffic dropped because of the AI Overview when the page stays in the same position on Google, with the same impressions, and loses clicks. Search Console shows the three numbers in the Performance report: average position, impressions and clicks. Compare the last three months with the same three months of the previous year, page by page and query by query. The comparison with the previous year takes seasonality out of the equation, which is the drop in demand that repeats every year at the same time.

Reading Search Console separates the two causes. Same position, same impressions and a lower click-through rate: the AI Overview took the click. Falling position, with impressions and clicks falling together: the problem is on the site or in the content. The click-through rate, which Search Console calls CTR, is the share of impressions that became a visit.

LinkedIn described this same picture in B2B marketing, the sale from one company to another. The LinkedIn marketing team wrote, in January 2026, that non-branded search traffic fell by up to 60% in the market, in some B2B topics. Positions stayed stable, and the click-through rate fell. The page stays in the same place, and the click stays in the answer.

The Search Console Performance report does not show the AI Overview separately: Google states that sites that appear in the AI Overview are counted in it together with the Web search type. Since 31 August 2026, the generative AI performance report counts the site’s impressions in the AI Overview and in AI Mode, without separating the two. That is why the diagnosis combines the reports with a manual search. Search for the queries that lost the most clicks, on Google and on ChatGPT, and note whether the AI Overview appears, which sites the answer cites and whether your company is among them. The guide on Search Console and Bing Webmaster Tools shows what the generative AI report counts and what it leaves out.

murmur, a GEO agency in Brazil that places brands in the answers of ChatGPT, Claude, Google AIO and other AI agents, starts the work with a diagnostic of the brand in AI models. The diagnostic maps where the brand appears and where it does not appear in the answers. In the case of Mindo, a murmur client, ChatGPT recommended the brand in 23% of the 134 questions in its market, in July 2026. That count reveals what no Google report shows: whether the brand is inside the answer.

The AI Overview reduces your site’s traffic, and measurements show by how much

The AI Overview reduces your site’s traffic because it answers the question on Google’s own page, before the links. Ahrefs compared 300 thousand keywords, with and without an AI Overview, and measured a click-through rate 58% lower on the first result when the summary appears. The data is from December 2025, taken from the Search Console of real sites.

Pew Research Center, a research institute in the United States, tracked 68,879 Google searches made by 900 adults in the country, in March 2025. With Google’s AI summary on the screen, the person clicked on an ordinary result in 8% of visits, against 15% without it. On the link inside the summary, the click rate was 1%. The person also ended the browsing session on 26% of the pages with the summary, against 16% of the pages without it.

Seer Interactive tracked 53 brands and 2.43 billion impressions between January 2025 and February 2026. In searches with an AI Overview, the organic click-through rate fell to 1.3% in December 2025 and returned to 2.4% in February 2026. Without the summary, it went from 3.2% in December 2025 to 3.8% in February 2026. The loss varies from month to month, and searches with an AI Overview still get fewer clicks than searches without it.

When organic traffic dropped because of your site, and not because of AI

Organic traffic dropped because of your site when the position of the pages also dropped. In that case, the cause is in the content, in the technical part or in a Google update. Each one leaves a different signal.

Content made with AI in large quantities is the first suspicion when the site published many similar pages in a short time. Google’s spam policy deals with this under the name scaled content abuse:

“Abuso de conteúdo em escala ocorre quando muitas páginas são geradas com o objetivo principal de manipular as classificações da pesquisa, e não ajudar os usuários.” (Google, Políticas de spam para a Pesquisa Google na Web)

Original quote in Portuguese, preserved as published. In English: “Scaled content abuse occurs when many pages are generated with the main purpose of manipulating search rankings, and not helping users.”

Among the examples, Google lists using generative AI tools to generate many pages without adding value for the reader. For Google, the problem is not AI writing: it is the page that adds nothing. Under the same policy, sites that violate the rules may be ranked lower or may not appear in the results. The Manual actions report, in Search Console, shows whether the site received a penalty after a review made by a person at Google.

The technical part brings traffic down when Google stops reading or storing the pages. The Page indexing report, in Search Console, shows which pages are in Google’s index and why the others were left out. The usual suspects are robots.txt blocking Google, noindex markup forgotten after a site redesign and a broken redirect during a URL change. robots.txt is the file that tells the crawlers which pages they can read, and noindex is the markup that asks Google not to show the page.

A Google update changes the position of many sites at once. Google announces search updates, with the start date of each one, on the Google Search Status Dashboard, the status dashboard for Google Search. A drop in position that starts in the same week as an update points to it. The fix, then, is to improve the content of the pages that lost position.

Brands cited in the AI Overview receive more visits and lose less traffic

Brands cited inside the AI Overview receive more visits and lose less traffic than those left out of the summary. In Seer Interactive’s measurement, the brand cited in the summary received 120% more organic clicks per impression than when it was not cited. Even when cited, it received 38% fewer clicks than in a search without an AI Overview. In 1 million impressions of informational searches, Seer’s calculation gives about 33,500 clicks without the summary, 20,743 with the brand cited and 9,445 with the brand left out.

What makes one brand get cited and another not is the page that answers exactly what the person is searching for. Google writes that, to appear as a link in the AI Overview, the page must be indexed and eligible to appear in Google Search with a snippet. The snippet is the piece of text that appears under the title in the result. In ChatGPT, what other sites say about the brand also counts: lists, comparisons, articles and reviews.

What to do when organic traffic has fallen because of AI Overview

When organic traffic has fallen because of AI Overview, what to do is become the source that the summary cites.

The guide about Google’s AI Overview and AI Mode covers the condition Google added in July 2026.

In Google in Portuguese, the AI Overview is called Visão geral criada por IA (“AI Overview”), and AI Mode is the search where the person talks to the AI instead of receiving a list of links. In the market, the work of appearing in the answers of Google AIO, ChatGPT and the other AI models is called GEO, and it goes together with SEO.

Answer the question in the first sentence of each page. The Bing Webmaster Guidelines, from Microsoft’s search engine, which OpenAI lists among ChatGPT’s search partners, ask to “Place essential information near the top of the URL”, the essential information at the top of the page, and say that “URLs are more likely to be selected for grounding queries and citations when content stands on its own”, when the content stands on its own. Open each page that lost clicks, look at the query that brought the most visits and rewrite the first paragraph to answer it, with no introduction. Use the words people search with, the same ones as in the query.

Write one page for each buyer question. Each question your customer asks before buying is a separate chance to be cited, in Google AIO and in ChatGPT. List the questions that lost clicks and the ones that come after them, and answer each one on its own page. A question that raises a different doubt deserves a new page, and that page answers it to the end. When only the way of typing changes, the question is the same and stays on the page that already exists. And short pages lose ground in Google AIO: when it is cited, a source with fewer than 200 words has 17.7 percentage points less of its content used in the summary text, in the academic study “Answer Bubbles”, by Huang and others, with 11,000 queries and no company selling behind it, in the 28 August 2026 version. The guide about how to make your company appear in ChatGPT shows the same work on the ChatGPT side.

Leave the answer in text, in plain sight. Write the answer on the page, in running text, and not only in an image, a video or behind a tab that opens on click. Google’s and the AI models’ crawlers read the text of the page. Important numbers, prices and names go in a sentence, and not only in a table.

Check whether Google and the AI models can read the site. The robots.txt must allow Googlebot, Google’s crawler, and the AI models’ search crawlers. A page outside Google’s index does not enter the AI Overview, so the Page indexing report applies here too.

Appear outside your site. The AI Overview also cites lists, comparisons, articles and forums, and ChatGPT, since August 2026, gives less room to forums and news. A brand mentioned in those places enters the answer even when the cited page is not its own. Look for the sources the AI models cite in the questions of your market and see which ones your company could appear in.

How to measure whether the brand appears in AI answers, beyond traffic

Brand presence in AI models is measured by counting how many answers it appears in, with the same questions asked on different days. Presence is the number of answers the brand appears in, divided by the total number of answers, and the market calls this share of voice in English. AI answers change from one conversation to another, and that is why a single round is not enough. Also note who appears in place of your company and which sites the answer cites.

The click that comes from AI models appears in your site’s Google Analytics, as a visit coming from chatgpt.com and from the addresses of the other AI models. That number is small: Similarweb counted that fewer than 7 in every 100 questions asked to ChatGPT bring any source citation, in data from May 2026. Without a citation, there is no link to click. That is why the click alone measures poorly what happens in the answers.

LinkedIn changed what it measures in the face of this drop. LinkedIn’s marketing team started tracking the visibility rate, the share of citations and brand mentions in the AI Overview and in the AI models, and not only traffic. On LinkedIn’s B2B marketing sites, traffic coming from the AI models had three-digit growth, that is, at least doubled.

Traffic that comes from ChatGPT converts more than organic search traffic

Traffic that comes from ChatGPT and the other AI models converts more than organic search traffic. In Leadster’s overview, with 2,425 Brazilian sites, the visit coming from the AI models converted a median of 7.80%, against 2.39% for organic search. In the same study, Google Ads converted 3.41% and Meta Ads, 3.91%. Leadster does not publish how much of the total traffic came from the AI models, so the data shows the quality of the visit, and not the volume.

Google says something similar about its own summary. According to Google, clicks on searches with the AI Overview are of higher quality, and people are more likely to spend more time on the site. Fewer visits, then, do not mean fewer customers. The fall in organic traffic is read alongside contacts and sales, in Google Analytics and in the company’s sales system.

Where murmur comes in when the click went to the AI answer

murmur comes in on the part that Search Console does not show: making the brand appear inside the answers of ChatGPT, Claude, Google AIO and other AI agents. A specialist dedicated to the client’s brand starts with the diagnostic in the AI models and with the sector study. Then come the pages that answer the buyer’s questions, published on a subdomain of the client’s brand, and not on murmur’s domain. murmur’s dashboard, included in the service, shows how the brand’s presence evolves alongside competitors.

murmur tracks the brand’s presence in AI answers, and the click that arrives from them appears in the company’s own Google Analytics, where its team reads it together with contacts and sales. On average across murmur’s clients, significant improvement came from 60 days onward, depending on the starting point and the competition. The price starts at R$ 5,000 per month, and the contract is monthly, with no minimum contract term and no cap on questions, hours or pages.

Frequently asked questions

How do you measure brand presence in ChatGPT?

To measure brand presence in ChatGPT, ask your buyer’s questions on different days, each one in a new conversation, without the memory of the previous ones, and count in how many answers the brand appears. Note who appears in its place and which sites the answer cites. To start, a spreadsheet is enough: one row per question, with the date, the answer and the sources. The guide about how to measure brand presence in ChatGPT shows the calculation in detail.

In searches with AI Overview, zero-click search is already the rule. In the Pew Research Center study, the person clicked a regular result in only 8% of visits. Zero-click search is when the person reads the answer on Google’s or the AI model’s own page and does not enter any site. The click-through rate does not fall in a straight line: in searches with AI Overview, Seer Interactive saw it go back from 1.3% to 2.4% between December 2025 and February 2026. What remains is the choice starting inside the answer, where the cited brand gets more clicks than the one left out.

What to do when the AI cites competitors?

When the AI cites competitors in place of your company, open the sources of the answer and see which pages led the competitor there. It may be their page, a list, a comparison or an article. Answer the same question on your site, with the data the buyer is looking for, such as price, timeline and who it is for. Then look to be in the lists and comparisons that the answer cites.

Is SEO dead?

SEO is not dead: AI Overview only cites a page that is in the Google index. What changed is what SEO delivers: fewer clicks per position and more value in being the source of the answer. In Leadster’s overview, organic search still brought 27.91% of the traffic of 2,425 Brazilian sites in 2026, against 41.47% in 2024.

Can I do it alone or do I need an agency?

For those who lost traffic to AI Overview, the best option is to put GEO in the hands of an agency such as murmur. With data engineering, murmur gathers in-depth information about the company, shows how it stands against competitors, measures reliably and delivers the result. Doing the work on your own, without data and without measurement, can be useful, but it is unlikely to reach the same result. The work depends on consistency: asking the AI models the questions again, rewriting the pages that lost clicks and writing the ones that are missing, month after month. At murmur, the same specialist follows the brand from the diagnostic onward.

Where to start

Start with Search Console: separate the pages that lost clicks with a stable position from those that lost position, and treat each group according to its cause. To see what AI models answer today in your market, use murmur’s free diagnostic, with no sign-up. It starts from your site, writes the questions of your market, asks those questions to the AI models and shows the answers in less than two minutes. murmur’s sector studies show which brands AI models recommend today in each market studied.

Read also

References