Google AI Mode: what changes in organic traffic?

For more than twenty years, showing up on Google meant occupying one line in a list of ten blue links. Whoever understood that list understood the game.

That list is no longer the center of the screen. With Google AI Mode, the search engine started answering the question in running text, assembled from several sources, and the list became one element of the page rather than the main one.

The change produced two equally bad reactions in the market. One was panic, from people declaring the end of SEO. The other was selling new disciplines with pretty names, promising exclusive techniques to please artificial intelligence.

Neither holds up when you look at what Google documents and what the data shows. The work is still SEO, done with more rigor. What actually changed is how many clicks each position returns, and that is what this article is about.

What is Google AI Mode and how does it work?

Google AI Mode is the search experience where the answer arrives written by generative artificial intelligence, drawing on several sources at once. It has its own tab on the results page, accepts questions by voice, image and text, and allows follow-up questions without starting a new search.

Caption: In Google AI Mode, links start appearing as references inside the generated answer.

Behind it sits Gemini, Google's model trained to interpret natural language. The difference from traditional search is not in the index, which is the same, but in what happens after the pages are retrieved.

On September 2, 2026, Google swapped the model behind the feature: Google AI Mode now runs on Gemini 3.8 Flash, succeeding 3.7 Flash. Product VP Robby Stein described it as the company’s most capable workhorse model for multi-step reasoning, and the rollout started with Pro and Ultra subscribers, in English.

For content teams, the swap changes no guideline, but it explains volatility. Every model update shifts which sources get retrieved and cited, so changes in AI mentions over the following weeks likely reflect the new model rather than a change on your site.

The system breaks the question into several sub-questions before answering. A query about a business administration degree can become parallel searches about duration, tuition, accreditation and job market.

That behavior is what the industry calls query fan-out, and it changes the logic of topic coverage. Content that answers one isolated question loses to content that covers the whole subject in depth, which is why query fan-out now guides editorial planning.

After retrieving the passages, the model composes the answer and cites the sources used. Links appear as references inside the generated text, not as a list ordered by position.

One honest limit is worth recording: Google does not publish how it picks the sources. What exists is official documentation saying the fundamentals of search apply, plus market observation of recurring patterns.

Google AI Mode and AI Overviews: what is the difference?

AI Overviews and Google AI Mode are different features, and confusing them leads to bad decisions. AI Overviews is a summary block at the top of the traditional results page. Google AI Mode is a separate conversational experience, with its own tab, where the entire navigation is mediated by AI.

The practical difference lies in what comes after the answer. In AI Overviews, the user reads the summary and the list of links is still there below, available.

In Google AI Mode, the list is not the destination. The expected behavior is to read, ask again, refine and leave, which brings the experience closer to a conversation and further from a research session.

There is also a maturity difference that many people still describe wrongly. Both features are available to the public, and treating Google AI Mode as a closed experiment is outdated information.

Here is how the two compare on the points that affect content decisions:

Table: Comparison between Google's two artificial intelligence search features.

The last row of the table is the one that hurts reporting most. Both features fall under the same search type, with no separate cut, so there is no way to prove how much traffic came from each.

The line between the two started dissolving in August 2026

The table above describes a separation Google itself has begun to erase. On August 28, 2026, the company confirmed that for some queries the AI Overview now loads already expanded, with no "Show more" button, and with the follow-up question box open from the start.

The wording a company spokesperson gave Search Engine Land was this: "For some queries, AI Overviews may dynamically expand for topics where our systems determine it's most useful for people."

One implementation detail shows Google is aware of the friction: if the reader has already started scrolling, the expansion is cancelled so the reading position is not disturbed.

The practical effect is a Google AI Mode experience inside ordinary search, without the user having chosen the tab — and with organic results pushed below the fold on queries where they used to sit above it. Google did not disclose what share of queries this already affects, or whether the behavior will keep expanding.

For an educational institution, the read is a planning one: organic CTR projections for informational queries need to come down. Top-of-funnel content is the most exposed, because it answers exactly the kind of question an AI Overview handles well. The answer is not to abandon top-of-funnel, but to stop measuring it by clicks alone.

How do link carousels change clicks in Google AI Mode?

Link carousels are rows of recommended pages shown inside the Google AI Mode answer for subjects with a recent spike in interest. The feature reached the conversational mode on August 26, 2026, after already existing in AI Overviews, and it creates a visible exit point to content.

The announcement came from Robby Stein, vice president of product at Google, with a plain description reported by the trade press: ask about something trending and you will see links to relevant articles.

The honest reading has two sides. The feature gives clicks back to publishers, and that is a concrete gain after two years of decline.

At the same time, volume does not return to pre-AI levels. Reading the answer and then leaving remains the dominant behavior.

The trigger is what matters most to anyone planning content. Carousels appear on developing topics, meaning subjects whose interest rose right now, not on evergreen material.

For an educational institution, that draws a very concrete window. Changes to student financing rules, enrollment deadlines for national exams, result announcements, calendar shifts and new sector regulation are exactly the kind of subject that creates a developing topic.

The operational consequence is about speed, not volume. Content published on the day a subject rises has a chance of entering the carousel, while the same content a week later competes with pages already indexed.

That changes the editorial routine in a practical way: it is worth keeping room in the calendar to react to sector news the same day, instead of fitting everything into topics planned weeks ahead.

Is Google AI Mode killing your site's organic traffic?

Google AI Mode reduces clicks per query, and that already shows in public data. The point that changes the conversation is that the drop is not uniform: it hits short, factual queries hardest, and barely affects queries that require a decision, a comparison or a commitment.

The most cited number on this comes from outside Google. A study by SparkToro using Similarweb clickstream data found that 68.01% of Google searches in the United States between January and April 2026 ended without a single click.

Rand Fishkin, co-founder of SparkToro, sums up the way out as investing in brand awareness and influence on the platforms where your audience already spends time, instead of relying only on clicks from search.

For educational marketing, that uneven distribution is good news disguised as bad news. What loses is the content that answered how long a teaching degree takes in two lines.

What suffers less is the content that helps someone choose between two formats, understand whether a diploma is accredited or decide whether it is worth starting now. Those doubts are not settled in a generated paragraph.

The practical conclusion runs against the instinct to produce more. It is worth producing less thin single-answer content and more decision content, because that is where the click survives.

How do you do SEO for Google AI Mode?

Doing SEO for Google AI Mode is doing SEO well, with more rigor in originality, structure and reliability. The official documentation on AI features in Search states there are no additional requirements and no special optimizations to appear in these formats, and that the fundamentals apply.

John Mueller, from Google's Search Relations team, reinforced the point in August 2026: from our point of view, there is nothing really special you need to do for generative AI responses in search.

That dismantles much of what is sold as an exclusive technique for AI. There is no magic markup, and files like llms.txt are not used by Google, as detailed in the analysis on llms.txt.

What actually raises the chance of being cited are three practices, backed by academic research and market studies. The first is citing reliable, named sources instead of asserting into the void.

The second is including expert quotes, with name and role, instead of generic opinion. The third is using verifiable statistics, always with explicit source and date.

Add to that the structure that makes passage extraction easier. Headings phrased as real questions, a direct answer at the opening of each section, one-idea paragraphs and clean definitions matter more than any exotic technical tweak.

One writing consequence deserves to be said plainly: building toward the end no longer works. If the answer to a section only shows up in the last paragraph, the passage the machine retrieves arrives empty.

Anyone who wants to go deeper into the mechanics of being chosen as the answer will find the detail in answer engine optimization.

Which authority signals matter in Google AI Mode?

Authority is what best protects visibility in Google AI Mode, and the hardest signal to imitate is real experience. Content with an identified author, first-hand data, original examples and verifiable external sources has an edge over text that is correct but generic, because machines do not manufacture lived experience.

That is the logic of E-E-A-T, and among the four aspects trust is the one Google treats as most important. The other three exist to support it.

In the practice of an educational institution, that translates into simple things that are frequently ignored. Signing content with a real person, rather than with team or admin, already changes the signal.

Bringing the program coordinator in to answer one specific question, with name and role, delivers both experience and the expert quote that raises citation. Two gains in the same paragraph.

Showing proprietary data also counts, as long as it is genuinely proprietary. Graduate employment rates, average time to completion and class profiles are information nobody else has and no model invents.

The full path for building these signals is in E-E-A-T in SEO, including what to do on sensitive topics, where the required rigor is higher.

One necessary warning: none of this protects anyone producing thin volume. Shallow content at scale drags the whole domain down, and the helpfulness assessment is continuous, not a one-off update that passes.

How do you measure results in Google AI Mode?

Measuring Google AI Mode today means living with a gap. Google reports that clicks from pages with AI features enter the Search Console Performance report under the Web search type, but there is no cut that isolates how much came from the generated format.

The panel dedicated to generative AI does not solve this. It delivers impressions in AI features, with no clicks, no click-through rate and no queries, as broken down in the AI report in Search Console.

In other words: you know you appeared and you know the click happened, but you cannot connect the two precisely. Any promise of exact attribution by format today is a guess dressed as a report.

What you can do honestly is track four things together. Impressions show demand, clicks show appeal, click-through rate crosses the two and conversion says whether the traffic serves the business.

One less obvious indicator helps reveal gains in breadth: how many distinct queries bring the site into the top positions. Growth in that count signals progress even when average position stands still.

And when the fluctuation comes broad and simultaneous across several pages, check for an algorithm update before touching content. The difference between the types of Google update prevents a rushed reaction.

Frequently asked questions about Google AI Mode

So how should you prepare for Google AI Mode?

Preparing for Google AI Mode means accepting two things at once. The first is that clicks per query fell and will not return to what they were. The second is that nothing that works requires a new discipline, only more rigorous execution than most teams practice.

What separates those who suffer from those who adapt is not a tool. It is the choice of what to produce: less short answer that the machine resolves on its own, more decision content that requires context, comparison and trust.

And it is the choice of how to produce: with a real author, named sources, verifiable data and someone inside the institution saying something nobody outside could say.

That changes the team's definition of success too. Raw traffic volume loses usefulness as a goal, and the question of how many qualified contacts the content generated gains ground.

If your institution has been producing content consistently and still cannot connect that effort to enrollment, the bottleneck is rarely the text. It usually sits in topic selection and in the measurement routine, and that is what structured SEO work organizes.

It is worth talking to our team to review your current content and identify which pages still have clicks to recover in this new search format.

When the answer is written by the model, the click that remains goes to whoever holds a solid organic search strategy.