Google appears to be testing a larger, more descriptive source panel in AI Mode while simultaneously moving those citations farther away from the generated answer. The experimental layout places citation cards toward the bottom of the page and gives each source more editorial context, including a title, description and image.
The behavior was documented by Sachin Patel in a September 10 report for Search Engine Watch. Patel says he tested the interface across different browsers and saw the bottom citation panel appear consistently. He had previously spotted a more limited version of the behavior in July, when interacting with free product listings in AI Mode.
The change has not been formally announced by Google, so it should be treated as an interface experiment rather than a confirmed redesign. There is also no public click-through data showing whether the larger cards compensate for their lower position. That tension—more information for each cited publisher, but more distance between the answer and the source—is what makes the test important for anyone measuring visibility in AI search.
The citation cards are bigger and more descriptive
According to Search Engine Watch, the tested panel is longer than AI Mode's standard citation presentation and gives individual sources substantially more visual information. Instead of relying primarily on a favicon or compact link treatment, the cards can display the publisher, page title, descriptive text and an image.
That can make a cited page easier to evaluate before a click. A user who reaches the sources can see not only which domain Google used but also what the page is about and whether it appears relevant to the original question. For publishers, the extra space creates a stronger opportunity to communicate why their source is worth opening.
It also changes the role of search-result presentation. In traditional organic results, the title and snippet have long helped users decide which page to visit. AI answers can compress that discovery step by synthesizing information directly. Rich citation cards partly restore it by giving the underlying pages a small preview surface inside the AI experience.
But the sources are being pushed below the answer
The potential drawback is placement. A larger citation card is only useful if the user reaches it. Moving the source panel to the bottom means people may need to consume or scroll through more of the AI-generated response before encountering the full citation treatment.
This creates two competing UX effects. Richer cards may be more attractive and more informative once visible, potentially improving the quality of outbound clicks. Lower placement may reduce the number of users who ever see those cards. Without controlled data from Google, there is no basis for declaring which effect wins.
The distinction matters because citation presence and referral traffic are not the same metric. A publisher can be cited prominently in the model's evidence set while receiving few visits if the interface makes the source difficult to notice or if the generated answer satisfies the user's need before a click becomes necessary.
The bottom-card idea was already spotted in July
The September observation is not the first sign that Google is experimenting with this direction. Search Engine Roundtable reported on July 22 that Google AI Mode was testing citation cards at the bottom of the experience. That version was spotted by Patel while opening a free listing product and used an anchor behavior that moved the user down to the source cards.
The newer Search Engine Watch observation is notable because Patel says the format now appears more broadly in AI Mode and was reproducible across multiple browsers. That suggests the experiment may have expanded beyond the narrow scenario observed in July, although it still does not establish that Google has begun a general rollout.
Google routinely runs search-interface experiments, and not every test becomes a permanent product feature. The correct status is therefore that a broader test has been observed—not that Google has officially moved all AI Mode citations to the bottom.
Google has been experimenting heavily with citation placement
The bottom-panel test fits a longer pattern of experimentation around how AI-generated search experiences expose the web sources behind their answers. During 2026, observers have documented tests involving citation favicons, result counts, sliding citation carousels and different panel dimensions across AI Mode and AI Overviews.
In May, for example, Google was spotted testing a mobile carousel that could slide source cards along the bottom of the screen after a citation interaction. In June, another test moved citation counts and favicons relative to link cards. More recently, Google has been testing a thinner citations panel in AI Overviews.
These experiments should not be conflated because AI Mode and AI Overviews are distinct surfaces and the tests do not necessarily share the same objective. Taken together, however, they show that source presentation remains an active design problem for Google's generative search products.
Google has publicly said links remain an area of focus
The broader context is Google's attempt to balance generated answers with pathways back to the web. Publishers have repeatedly questioned whether AI search products will provide meaningful referral traffic when users can get synthesized answers directly on the results page.
Google has maintained that links and exploration remain important to its search experience. That does not tell publishers whether any individual citation layout will generate more clicks, but it explains why the company continues testing different ways to expose sources alongside AI answers.
The new bottom layout is interesting because it appears to increase the amount of information devoted to publishers while decreasing the proximity of those publishers to the claims they support. From a transparency perspective, a detailed source card can be useful. From an attribution perspective, separating evidence from the sentence it supports may make the connection less immediate.
Citation UX is becoming a ranking-independent traffic variable
For SEO teams, the experiment highlights a new source of volatility. A publisher's underlying inclusion in AI Mode could remain unchanged while its referral traffic changes because Google altered where and how citations appear. The content may still be selected as evidence; the user interface around that evidence may simply create a different likelihood of clicking.
That means AI visibility reporting should distinguish at least two outcomes: whether a source is cited and whether that citation generates measurable traffic. Treating the first as a proxy for the second can hide interface effects. A site might maintain stable citation share while AI referrals fall, or receive more traffic from fewer citations if the presentation becomes more compelling.
The same principle applies to tests that add images and descriptions. Those elements can change user behavior without any change in retrieval or source selection. In AI search, the path from being chosen by the system to receiving a visitor increasingly includes a presentation layer that publishers do not control.
The richer cards could reward stronger titles and visual assets
If Google ultimately adopts a citation format that consistently displays page titles, descriptions and images, source presentation may become more important to AI-search traffic. A citation would function less like a tiny attribution marker and more like a compact recommendation card.
That does not create a reason to rewrite pages around an unconfirmed test. Google has not documented exactly how the card title, description or image is selected, and the experiment may change or disappear. But publishers should already be maintaining clear titles, useful page summaries and strong visual assets because those elements serve users across search, social sharing and other discovery surfaces.
The deeper opportunity is differentiation after the answer. If AI Mode has already summarized the basic fact, the source needs to promise something the summary cannot fully provide: original evidence, deeper analysis, interactive data, a primary document, expert interpretation or a useful tool. A richer citation card can communicate that value more effectively than a favicon alone.
Moving citations downward raises a transparency question
Citation design is not only a publisher-traffic issue. It affects how easily users can evaluate the provenance of an AI-generated claim. Inline or nearby citations make the relationship between a statement and its supporting source relatively visible. A consolidated panel at the end can make the evidence easier to browse as a group but less directly connected to individual claims.
Different research and writing interfaces have long made this tradeoff. Footnotes preserve the reading flow but require movement away from the sentence. Inline links make verification immediate but can visually interrupt the text. Google's test appears to be exploring a similar design problem at search-engine scale.
The ideal implementation may depend on the query. A short factual answer could support a bottom source panel without imposing much friction. A long, multi-part response drawing on many sources may make claim-level attribution more important. The current observation does not establish how Google will handle those cases if the experiment progresses.
Publishers should avoid drawing CTR conclusions from screenshots
The visual difference makes the test easy to debate, but screenshots cannot answer the most commercially important question: whether the layout sends more or less traffic to publishers. Search Engine Watch's Patel says he personally prefers the larger bottom cards and believes their added detail could help featured sites. That is a reasonable interface judgment, not evidence of measured click behavior.
Users may scroll farther because the cards look useful. They may stop earlier because the AI answer already resolved the query. Images may increase attention, while the lower position reduces exposure. Those effects can only be separated with behavioral data.
For SEO teams, the better response is to watch first-party referral trends alongside citation tracking. If a layout change becomes widespread, shifts in AI Mode traffic should be evaluated against stable citation visibility where possible. That can help distinguish a retrieval change from a presentation change.
This is a test, not a new citation standard
The most important qualification remains the simplest one: Google has not officially announced that AI Mode citations are moving to the bottom. Search Engine Watch has documented a reproducible test, and earlier reporting shows that a related bottom-card treatment was already appearing in limited scenarios in July.
That is enough to show Google's design exploration is continuing, but not enough to declare a permanent new interface. Publishers should monitor it without turning the experiment into a new SEO rule or assuming that larger cards automatically mean higher referral traffic.
If the format survives testing, it could produce a meaningful tradeoff. Sources would receive more visual real estate, with titles, descriptions and imagery that make each citation easier to understand. At the same time, the full source panel would sit farther from the answer that used those sources. Google may be making citations bigger while making users work harder to reach them. Until the company rolls out the design or releases engagement data, both sides of that equation remain open.