Google is testing a version of AI Mode that makes a search session look less like a results page and more like a conversation with pages.
A new experiment adds pagination to successive AI Mode responses and gives users a “Skip” button for moving through them, according to Search Engine Roundtable. The test, spotted by Sachin Patel, also illustrates a deeper change in the mechanics of Google Search: when the user replies inside the conversation, Google performs another search while keeping that interaction inside the same conversational thread.
The interface is still a test rather than a confirmed global rollout, and Google has not published an official announcement explaining this exact pagination experiment. But the behavior fits the direction Google has already documented for AI Mode, where users can ask complex questions, continue with follow-ups and let the system perform additional retrieval as the conversation develops.
Search is no longer necessarily one query followed by one results page. It can become a sequence of searches hidden inside one continuous conversation.
The Skip button makes AI answers behave like pages
The most visible part of the experiment is pagination.
Instead of treating each generated answer as one long block that must simply be scrolled, the test divides the experience into successive responses that users can move through. A Skip control lets them advance without necessarily engaging deeply with the current response.
That may sound like a minor interface decision, but pagination changes how information is consumed. It creates a defined sequence: current answer, next answer, next stage of the conversation.
Traditional Google pagination separated groups of search results. AI Mode’s version appears to paginate the evolving dialogue itself.
The unit being navigated is no longer a page of ten links. It is a step in an AI-mediated research process.
Every follow-up can become another search
The more consequential detail is what happens when the user responds.
Search Engine Roundtable reports that each reply triggers a new search while remaining part of the existing AI Mode conversation. The user therefore experiences continuity even though Google is retrieving new information behind the interface.
Official Google Search Central documentation describes AI Mode as particularly useful for questions that require further exploration, reasoning or complex comparisons. Google says users can ask nuanced questions that previously might have required multiple searches and receive a comprehensive response with supporting links.
The key phrase is “multiple searches.” AI Mode is increasingly turning that sequence into an invisible part of one conversational workflow.
One conversational turn can involve more than one backend query
There is an important technical distinction between saying a follow-up initiates a new search and saying it maps to exactly one search query.
Google says AI Mode can use a technique called query fan-out. Its systems generate multiple related queries across subtopics and data sources, retrieve relevant information and use those results to construct the response.
Google’s 2026 generative AI optimization guide explains that fan-out queries can run concurrently to gather additional information related to the user’s original need.
A single conversational reply can therefore be much more complex than a conventional search box submission.
The user may type one sentence. AI Mode can interpret that sentence in the context of everything already discussed, formulate additional retrieval tasks and return a synthesized next step.
The search query is becoming a conversation state
Traditional SEO assumes a relatively discrete unit of demand: the query.
A person searches for “best laptop for video editing,” Google processes that phrase and a results page appears. If the person then searches “best laptop under $2,000,” marketers can treat that as a second query with its own rankings and search demand.
Conversational AI complicates that model.
The second question might simply be “What about under $2,000?” Its meaning depends entirely on the previous turn. The literal words no longer contain the full intent.
AI Mode can preserve the context and search again without requiring the user to restate the subject.
That makes search demand harder to understand through isolated keyword strings. Part of the query now exists in the conversation history.
Pagination could encourage deeper exploration without returning to the SERP
The Skip control also suggests that Google wants users to navigate AI Mode internally rather than repeatedly returning to a conventional results page.
If one answer is not useful, the next action does not have to be reformulating a query from scratch. The user can advance, ask a follow-up or continue the same research session.
That reduces friction for complex tasks such as planning a trip, comparing products, learning a topic or evaluating several alternatives.
It can also increase the distance between the user and the conventional organic results beneath or outside the AI experience.
The longer a task remains inside AI Mode, the more important supporting citations become as the route from the generated answer to the open web.
Publishers may compete across multiple turns, not one results page
For publishers and SEO teams, conversational search changes the competitive frame.
A website that does not support the first AI response could still become relevant during a follow-up. Another source may appear early in the conversation and disappear as the user narrows the question. A third could be retrieved only after Google fans the query out into a more specific subtopic.
Visibility therefore becomes temporal as well as positional.
The question is no longer only whether a domain appears for the initial prompt. Marketers may also need to understand whether it appears during the second, third or fourth stage of the user’s exploration.
That creates a much larger measurement space than conventional rank tracking.
A “position” inside AI Mode becomes difficult to define
Pagination makes the ranking problem especially clear.
In classic Search, position can be described with a number. A result is first, second or tenth.
Inside a paginated AI conversation, several dimensions could matter simultaneously: which conversational page contains the citation, where the citation appears within that answer, whether it remains visible while the user moves forward and whether the user ever reaches that stage of the dialogue.
A citation on the first response and a citation on the fifth response cannot automatically be treated as equivalent exposure.
Yet current AI visibility tools often compress these complex appearances into a simpler metric such as “mentioned” or “cited.”
As interfaces become more conversational, those summaries may need more context.
The Skip button introduces another behavioral signal
A Skip button can also generate a new class of interaction data for Google.
If users consistently skip certain response types, Google can learn that those stages are not satisfying. If users stop, follow a citation or ask a related question, those behaviors can indicate a different level of engagement.
There is no evidence from this test that Google uses Skip behavior as a ranking signal, and it would be speculative to claim that it does.
But at the product level, explicit navigation controls give Google more information about how people move through AI-generated search journeys.
That feedback could help the company refine the interface even if it never becomes part of web ranking.
Search Console still does not expose the full conversational journey
The test also highlights the widening gap between the complexity of AI Search and the reporting available to publishers.
Google says traffic from AI Overviews and AI Mode is included in the overall Web search performance reported in Search Console. The company does not provide a dedicated AI Mode performance report that reconstructs conversational turns, fan-out queries or the exact stage at which a publisher was cited.
A publisher may therefore receive a click from the third response in a multi-step AI Mode conversation without seeing the full path that produced it.
The observable event is the visit. The preceding discovery journey remains largely inside Google.
That makes attribution more difficult precisely as the search process becomes more complex.
Follow-up questions are becoming a core Search behavior
Google has been signaling this direction for some time.
Its Search Central guidance says AI experiences allow users to ask longer and more specific questions and then use follow-up questions to dig deeper. Google has positioned that behavior as one of the ways generative AI expands what people can ask Search.
Pagination gives that conversational behavior more visible structure.
Instead of follow-ups feeling like separate queries stacked in a chat transcript, the interface can present the research process as a sequence users navigate deliberately.
This is a meaningful departure from the search session model that shaped decades of SEO analytics.
Keyword research cannot see the whole query anymore
The implications extend beyond rank tracking.
Keyword research tools depend on the idea that users submit identifiable query strings that can be aggregated into demand estimates. Conversational follow-ups can be semantically rich while being linguistically incomplete.
“Which one is cheaper?” is a meaningful search only when the system knows which products were being compared in the previous response. “What about Italy?” may completely change a travel recommendation while containing almost none of the original subject.
As more searches happen this way, marketers will need to model topics, tasks and decision journeys rather than relying exclusively on standalone keyword phrases.
The keyword does not disappear, but it becomes one piece of a larger context window.
Google still says conventional SEO fundamentals apply
Despite the interface change, Google’s official guidance does not prescribe a separate optimization system for AI Mode.
The company says AI features are grounded in its core Search ranking and quality systems and that pages need to be indexed and eligible for normal Search snippets to appear as supporting links. Google continues to recommend useful, original, people-first content rather than special AI markup or pages manufactured around every possible fan-out query.
That last point becomes especially relevant in conversational search.
It would be tempting to create separate pages targeting thousands of hypothetical follow-up questions. Google explicitly warns against generating large quantities of pages for query variations primarily to manipulate Search or generative AI responses.
The more durable strategy is comprehensive content that can remain relevant as the conversation branches into related needs.
The user journey is moving from query-result-query to one continuous session
The larger significance of the test is not pagination itself. It is the workflow pagination reveals.
Classic web search often looks like this: query, results, click, back button, new query, new results.
AI Mode can compress much of that into one session: question, generated answer, follow-up, new retrieval, new answer, skip, another question, another retrieval.
From the user’s perspective, it is one conversation. From the information-retrieval perspective, it can contain many searches.
That distinction will matter increasingly for anyone trying to measure search demand, rankings, citations and traffic.
Google is testing a new unit of Search: the conversation
For more than two decades, the basic unit of Google Search was the results page generated by a query.
AI Mode is testing something different. The unit can become the conversation itself, with multiple searches, responses and source sets contained inside it.
The new pagination and Skip experiment makes that transition unusually visible. Each response can become another stage to navigate, while each user reply can send Google back out to retrieve more information without breaking the conversational flow.
This is still an experiment, and Google may redesign or discard the pagination interface. But the underlying behavior is already part of the company’s documented vision for AI Mode: complex questions, follow-up exploration and query fan-out inside a generative search experience.
Google Search is not simply adding chat to the SERP. It is turning multiple searches into a single conversation—and now it is testing pages for navigating that conversation.