Google AI Mode is testing at least three new query expansion (ways to extend or refine a search beyond the original question) interfaces as of September 14, 2026: multi-selection follow-up tools, query refinement options, and paginated responses. Spotted by Sachin Patel and reported by Search Engine Roundtable’s Barry Schwartz, these formats go beyond standard follow-up questions.
What new AI Mode interfaces is Google testing?
Three distinct interface formats have been spotted in active AI Mode testing. Each changes how users navigate from an initial question toward more specific information, with screenshots shared by Sachin Patel via X and Search Engine Watch.
- Multi-selection follow-up tools: Follow-up questions are presented with multi-selection inputs, letting users choose more than one suggested topic at once rather than selecting a single path forward.
- Query refinement options: A dedicated refinement interface lets users adjust or narrow the direction of their current query.
- Paginated responses: AI Mode answers are broken across multiple navigable pages rather than appearing as a single continuous response. Schwartz notes this format has surfaced recently in testing.
Barry Schwartz wrote that he is “not sure how new these are” but expects “a lot of these tests from Google” as AI Mode continues to develop. The report appeared in Search Engine Roundtable on September 14, 2026.
Are these formats close to launch, or still early experiments?
Based on Schwartz’s report, the specific start dates of these tests are not confirmed. What is clear is that Google is running multiple query expansion experiments in parallel inside AI Mode, not converging on a single format yet. This is consistent with how Google typically develops major search features before broader rollout.
What this means for AI-search visibility
More interface options inside AI Mode means more possible paths a user can take before they reach a final answer. Each path is a potential citation point. The following is Hingewise’s analysis of what this shift signals, separate from the reported facts above.
The plain-language version first: right now, AI Mode mostly works like a conversation with one follow-up question at a time. These tests suggest Google is building toward a mode where users can branch in multiple directions at once, narrow their search midway through, or scroll across several pages of results. For any brand trying to appear in AI Mode answers, that means the “shape” of a winning piece of content may need to change.
Multi-selection: If users can pick multiple follow-up topics simultaneously, AI Mode may need to surface content that covers a cluster of related sub-questions rather than one narrowly focused answer. A page about “email marketing” that also has clear, distinct sections on automation and segmentation may perform better than a page optimized for just one of those sub-topics. That is a different calculation than standard single-query SEO.
Query refinement: When users refine mid-session, the content cited in the refined search may differ from what surfaced for the original query. Content structured to serve both a broad topic and its narrower variants, with clear internal organization, may be better positioned at both stages of that journey.
Paginated responses: This introduces a variable that does not exist in single-response AI answer formats. Whether citations on page one of a paginated response carry more brand visibility than later pages is not yet known. But the format itself creates a sequence, and sequence implies priority.
One detail the source does not address: it is not clear whether Google is testing these formats to eventually pick one winner, or to match different interface types to different kinds of queries. There may not be a single AI Mode format to optimize for. Different search intents may route to different expansion interfaces, which means blanket tactics may underperform compared to topic-specific content structuring.
The Hingewise position: these tests collectively suggest Google is building AI Mode to support longer, multi-step research sessions rather than single-query resolution. That shifts the relevant question from “does my content rank for this query?” to “does my content serve a user across several steps of a research journey?” Content built around one isolated question may serve AI Mode progressively less well if multi-step, multi-branch sessions become the default.
What to check before these formats expand more broadly
- Does your key content address a cluster of related questions on the same topic, or only one narrow angle?
- Are your most important answers placed in the first one to two paragraphs? AI systems typically pull citations from early in the page.
- Are distinct subtopics within a page clearly separated with headings, making them easier for AI to identify and cite individually?
- Does a subtopic section hold up if a user arrives there directly, without reading the full page from the start?
- Do you have a baseline tracking which AI Mode queries currently surface your brand versus competitors? Establishing that now makes it easier to detect shifts if new interface formats change citation patterns.
Whether these three formats progress to wider rollout or get dropped in favor of something else is not yet clear. Google runs many interface experiments simultaneously, and not all advance. But the direction of the tests, more ways to expand, refine, and paginate searches, points toward AI Mode becoming a more layered, multi-step research tool. That trajectory is worth tracking closely as the feature continues to develop.
Source: Search Engine Roundtable, Barry Schwartz, September 14, 2026. Original observations by Sachin Patel via X and Search Engine Watch.
Lam Nguyen · Hingewise
This article was drafted with AI assistance from the sources cited above and checked against them before publishing. The "What this means for AI-search visibility" section is Hingewise's editorial analysis, built on the frameworks and positions we use in our GEO work. How we work: Editorial policy.