Google AI Mode & ChatGPT: bottom citations

Google and OpenAI are currently experimenting with a new way of presenting source citations in their generative search and chat interfaces. Both Google AI Mode and ChatGPT are testing variants in which citation cards no longer appear prominently beside or above the answer, but are deliberately placed at the bottom of the result. For SEO and GEO teams, this is more than a cosmetic detail: the position of citations influences how users perceive sources, click them, and continue their information journey.

What exactly is being tested?

In Google AI Mode, users can be observed being guided via scrolling or anchor behavior directly to the lower section of the answer, where citation cards are grouped together. The answer itself takes center stage, while sources function as a closing block. ChatGPT takes a slightly different approach: when a source reference is clicked, an overlay with the citations box loads. Sources therefore do not remain permanently visible as a bottom card strip, but are loaded interactively.

Both tests aim to improve the readability of the generated answer while preserving the traceability of sources. For publishers and brands, that means visibility inside the AI answer alone is not enough. What matters is whether and how your domain appears in these lower citation elements, and how easily users can move from there to the original page.

Implications for Generative Engine Optimization

Generative Engine Optimization (GEO) focuses on preparing content so that it is cited, summarized, and recommended in AI-powered search and answer systems. Citation cards are the central link between the generated answer and the original source. When their position shifts downward, the attention hierarchy changes as well: the generative answer dominates the viewport, while sources serve more as evidence and entry points.

This creates concrete implications for GEO strategies. Content must be structured clearly and factually robust enough for models to continue selecting it as a citable source—even if the click path is longer or less prominent. At the same time, brand awareness and domain authority gain value: users who only encounter citations at the end of an answer are more likely to choose familiar, trusted domains.

On-page signals and content quality

For content to appear as a citation in AI Mode and ChatGPT at all, classic quality factors remain relevant: clear statements, up-to-date data, transparent sourcing within your own content, and demonstrable author or brand expertise. In short, E-E-A-T-related signals and well-structured sections increase the chance of being listed as a bottom-card source. Precise headings, definitions, and lists also help, because generative systems can extract and assign such building blocks more easily.

Differences between Google AI Mode and ChatGPT

Although both systems move or bundle citations toward the bottom, their interaction patterns differ. Google AI Mode anchors user guidance more strongly at the bottom of the answer and makes citation cards a visible closing element. ChatGPT relies on a click overlay: the source box appears only after interaction. That has consequences for measurability and user behavior.

• In Google AI Mode, citations can be perceived and scrolled as a fixed bottom block.
• In ChatGPT, visibility depends more strongly on an active click on a source marker.
• Both variants reduce visual competition with the generated answer and prioritize answer flow.
• For publishers, the path from citation to landing page becomes a critical conversion point.

SEO teams should therefore not only check whether their domain is cited, but also under which UI conditions users discover the source at all. Search Console observations, log files, and referral traffic from AI surfaces help detect changes early.

Practical levers for SEO and GEO teams

Anyone who wants to benefit from bottom-card citations should deliberately design content for citability. That starts with clear fact blocks and extends to technical prerequisites such as fast load times, mobile usability, and clean indexability. It is also worthwhile to build topic clusters: when multiple pages of a domain consistently contribute to one subject area, generative systems are more likely to treat the brand as a reliable source.

Analyzing snippet and brand signals is equally important. When citations only appear at the bottom, brand recognition and trust signals often decide whether users click. Visible expertise, consistent naming, and solid references in the content support this effect. Teams should also review tracking setups: referral parameters and user-agent patterns from AI clients can indicate which surfaces deliver traffic.

Monitoring and reporting

Because Google and OpenAI continuously adjust UI tests, regular monitoring of citation presentation is recommended. Documenting screenshots, card positions, overlay behavior, and click paths helps identify trends. Organic and AI-related metrics should also be viewed separately so that citation UI changes are not misinterpreted as classic ranking problems.

In the long run, the test underscores a central GEO thesis: in generative interfaces, brands compete not only for rankings, but for citation slots and for attention inside a condensed answer UI. Teams that prepare content to be citable, clear, and trustworthy remain visible even when citation cards move downward or appear only via overlay.

For editorial and SEO leads, the current test run is therefore a reason to reassess content formats, source structure, and brand presence in AI search and chat systems—and to secure visibility along the entire citation funnel.