Search visibility can no longer be evaluated only through a conventional results page. To monitor link carousels in AI Mode effectively, publishers need to combine Google’s generative-AI performance reporting with direct observation of how sources appear inside AI responses.
The opportunity is growing, but attribution remains imperfect. Google is making AI Mode more link-forward through carousels, inline links, previews, Preferred Sources, subscription links, and other discovery features, while continuing to experiment with their design.
Why link carousels in AI Mode deserve separate attention
Google’s AI experiences do not present links in just one format. A source might appear as an inline citation, inside a prominent carousel, as a highlighted Preferred Source, or through a link associated with a user’s news subscription.
On May 27, 2026, Google said it was bringing Preferred Sources into AI Mode and AI Overviews. It also announced new search carousels intended to highlight those sources.
For some searches about developing topics, Google said AI Mode can show a prominent carousel containing timely articles and different perspectives. That makes carousel visibility particularly relevant to newsrooms, specialist publishers, and organizations covering subjects that change quickly.
- Carousels create an additional discovery surface: A page can be presented among several sources selected to support further exploration.
- Placement can affect prominence: A large carousel may be easier to notice than a less prominent inline link, even when both lead to external pages.
- User preferences can influence presentation: Sites selected through Preferred Sources can stand out inside AI-generated responses and carousels.
- Source context is becoming more visible: Google introduced a Highly Cited badge to identify original reporting and coverage that has influenced other reporting.
- Interaction is changing: On Chrome desktop, an AI Mode link can open beside the AI experience, allowing the user to retain the original context while viewing the page.
These changes do not mean every AI Mode impression involves a carousel. They mean that source visibility now has multiple possible forms, each of which may create a different user journey.
The scale of AI Mode reinforces the need to watch those journeys. In May 2026, Google said AI Mode had surpassed one billion monthly users globally and that queries had more than doubled every quarter since launch. That statement does not establish how much traffic any individual site should expect, but it does show why link discovery inside AI Mode is becoming an important measurement question.
How to monitor link carousels in AI Mode
Direct answer: Monitor link carousels in AI Mode by tracking AI Mode and AI Overview impressions in Google Search Console, recording recurring changes by page and query theme, and manually reviewing representative AI responses for carousel placement. Treat Search Console as the performance record and manual checks as the presentation record, because generative-AI visibility does not automatically prove that a page appeared in a specific carousel.
A reliable process therefore has two layers. The first measures whether Google is surfacing a site in its generative-AI experiences; the second documents the exact link treatment visible to a user.
Layer one: performance monitoring
Google launched Search Generative AI performance reports so site owners can monitor impressions in AI Mode and AI Overviews. The reports support hourly, daily, weekly, and monthly timeframes, making them the starting point for trend analysis.
Use this reporting to answer questions such as whether AI visibility is increasing, which pages are gaining exposure, and whether a change is brief or sustained. Google explicitly encourages site owners to use Search Console to monitor visibility in generative-AI features.
Layer two: interface observation
Search Console tells you about performance in the covered AI surfaces, but carousel monitoring also requires examining presentation. For selected queries, record whether the source appears, where it appears, what type of link treatment is used, and which competing sources appear nearby.
This distinction prevents a common reporting error: equating every AI Mode impression with carousel inclusion. Unless a report explicitly identifies the presentation type, an impression in a generative-AI experience should not be labeled as a carousel impression.
Layer three: outcome analysis
Visibility alone is not the final business result. Evaluate whether pages receiving AI exposure also attract useful visits and support the goals assigned to those pages, such as readership, registrations, subscriptions, qualified inquiries, or deeper content exploration.
The appropriate outcome depends on the site. A news publisher may care about engaged readership and subscriptions, while a documentation site may prioritize successful access to an answer and movement into related guidance.
Build a measurement plan before collecting data
Monitoring becomes noisy when every AI appearance is treated as equally meaningful. Before looking for patterns, define which topics, pages, and outcomes matter to the organization.
- Select priority topic groups. Start with themes for which the site has genuine expertise, original reporting, or frequently updated information. Include developing topics if timely coverage is an important part of the editorial strategy.
- Map the relevant landing pages. List the articles, explainers, category pages, and primary-source materials that could reasonably support an AI-generated response. Separate current reporting from evergreen resources so their performance is not interpreted in the same way.
- Define observable AI treatments. Use a controlled vocabulary such as prominent carousel, standard carousel, inline link, inline-link preview, Preferred Source highlight, subscription link, badge, or no visible source placement.
- Choose a review cadence. Use hourly or daily reporting when evaluating a fast-developing event, and weekly or monthly views for broader strategic trends. The cadence should match the speed at which the underlying topic changes.
- Assign business outcomes. Decide what useful performance means for each page group. Avoid assuming that an impression, a visit, and a conversion represent the same level of value.
- Create an annotation process. Record major publication updates, corrections, redesigns, migrations, and known changes to AI link presentation. Annotations make later interpretation more disciplined.
The controlled vocabulary is especially important because Google is still testing link designs. In June 2026, Google told site owners that it was continuing to experiment with a range of new link designs in AI experiences, including AI Mode.
A monitoring system should therefore describe what was observed rather than assume that one layout will remain permanent. Labels based on function are usually more durable than labels tied to a button position, card shape, or color.
Separate leading and outcome indicators
AI impressions are a leading visibility indicator. They show that a page or site was surfaced within the relevant Google experience, but they do not by themselves establish that a person opened the result or completed a valuable action.
Manual carousel observations add presentation context. Site analytics then help evaluate what happened after a person arrived, subject to the tracking and attribution available to the organization.
- Visibility indicators: AI Mode and AI Overview impressions, pages receiving exposure, and movement over time.
- Presentation indicators: Carousel inclusion, source order as observed, inline-link placement, preview availability, badges, and adjacent sources.
- Engagement indicators: Meaningful on-site activity appropriate to the content and audience.
- Business indicators: Subscriptions, registrations, leads, purchases, or other defined outcomes when applicable.
Keeping these layers separate makes the report more trustworthy. A visibility gain can be positive even before a business effect is clear, but it should not be presented as proof of revenue or audience growth.
Use Search Console to track generative-AI visibility
Search Console is the primary Google-provided source for monitoring performance in AI Mode and AI Overviews. Begin with a consistent reporting window and preserve comparable views so that a short-lived spike is not mistaken for a durable trend.
Review more than one timeframe
Hourly views are useful when a topic is developing quickly. They can help teams see whether exposure follows the publication or update of timely coverage, although timing alone does not prove why Google selected a source.
Daily and weekly views provide more stable context. Monthly views are better suited to identifying whether generative-AI visibility is becoming a recurring part of a site’s search presence.
Each timeframe answers a different question:
- Hourly: Is visibility emerging or fading during an active story or rapidly changing topic?
- Daily: Did exposure persist beyond a brief moment?
- Weekly: Which content groups are repeatedly being surfaced?
- Monthly: Is AI visibility expanding, contracting, or shifting across the site?
Analyze page groups, not only individual URLs
Individual pages matter, but grouping related pages can expose broader editorial patterns. A publisher might separate original reporting, live coverage, analysis, explainers, reviews, and evergreen reference material.
If a group gains AI Mode impressions while another loses them, inspect differences in topic freshness, depth, sourcing, and update practices. That comparison can generate useful hypotheses, but it should not be turned into an unsupported claim about Google’s ranking systems.
Compare AI visibility with the publication record
Maintain a simple record of when important pages were published, substantially updated, corrected, consolidated, or redirected. This history helps explain whether visibility followed a real editorial change or merely coincided with one.
For developing topics, note when the underlying facts changed and when the page reflected those changes. Google describes prominent carousels as a way to surface timely articles and different perspectives, so update discipline is a sensible area to evaluate.
Investigate changes without overdiagnosing them
A rise or decline may reflect demand, changing source selection, a new interface treatment, topic seasonality, content changes, or ongoing product experiments. Search Console can reveal the movement, but it may not identify the cause.
When a material change appears, review multiple pieces of evidence before acting:
- Check whether the shift affects the entire site, one directory, or a small group of pages.
- Compare fast-moving topics with stable evergreen topics.
- Look for publication, technical, or template changes made during the same period.
- Review representative AI Mode responses to see whether the link format has changed.
- Confirm whether the apparent change persists across longer timeframes.
This approach is slower than assigning an immediate explanation, but it reduces false conclusions. It is particularly important while Google continues to experiment with AI link designs.
Audit carousel placement and source presentation manually
Manual review fills the gap between an impression report and the experience visible on screen. It is not a replacement for Search Console, and it should not be treated as a complete census of all users or queries.
Results can depend on the query, timing, user context, preferences, subscriptions, device, and interface changes. Preferred Sources and news subscriptions make that context especially important because two people may not receive identical source emphasis.
Create a representative query set
Choose queries based on real editorial and business priorities rather than searching only for terms where the site already appears. Include broad questions, specific questions, developing topics, and queries that could benefit from original evidence or multiple perspectives.
Organize them into stable groups. A useful set might contain recurring core topics, current developments, brand-related searches, and non-brand searches where the site has substantive material.
Record the observation consistently
For every review, capture enough context to make the observation interpretable later. A basic record should include the query, review time, device or environment, observed link treatment, destination page, nearby sources, and any visible badges or previews.
If a carousel appears, note whether the site is present and how the card represents the content. If there is no carousel but an inline link appears, record that separately rather than classifying it as a failure.
- Query context: What exactly was asked, and was the subject developing at that moment?
- Surface: Was the observation made in AI Mode or an AI Overview?
- Link format: Carousel, inline link, highlighted source, subscription link, or another visible treatment.
- Page selected: Which URL did Google surface, and was it the most suitable page?
- Source context: Which other publishers or websites appeared nearby?
- Quality markers: Was a Highly Cited badge or another visible source treatment present?
- Interaction option: Was a hover preview available, or did the page open in the side-by-side Chrome experience?
Screen captures can preserve visual context for internal analysis, while a structured log makes patterns easier to review. Document observations in accordance with the organization’s privacy, security, and record-retention practices.
Test personalized treatments carefully
Preferred Sources allow users to label sites they value in settings, and Google says those sources will stand out in AI responses and search carousels. This introduces a meaningful distinction between general visibility and preference-influenced visibility.
When reviewing results, note whether the observer has marked the site or a competitor as a preferred source. Do not report a preference-influenced result as though it represented a neutral experience for every user.
News subscriptions create a similar issue. Google is rolling out a feature that highlights links from a person’s news subscriptions inside AI Mode and AI Overviews, so subscription status may affect which links receive emphasis.
Observe the full click experience
The link itself is only one part of the journey. Google announced in April 2026 that, on Chrome desktop, clicking an AI Mode link opens the destination beside AI Mode so the user can continue exploring without losing context.
In May 2026, Google also said desktop users could hover over inline links in AI Mode and AI Overviews to see a quick website preview. A title, image, or opening section that communicates value clearly may therefore shape the decision to open the page, even though Google has not supplied a guaranteed formula for preview selection in the facts provided here.
During an audit, evaluate whether the destination immediately fulfills the promise implied by its presentation. If the relevant evidence is buried, outdated, or obscured, earning visibility may not translate into a satisfying visit.
Improve the content signals that carousel monitoring reveals
Monitoring should lead to editorial decisions, not just a dashboard. When particular pages repeatedly appear,or repeatedly fail to appear,review whether they provide the qualities a person would need when exploring a topic through AI Mode.
This does not justify manufacturing content solely to trigger a carousel. Google has described these surfaces as ways to make timely articles visible, present different perspectives, and help people explore the web, so the durable goal is to publish material worth discovering.
Prioritize original contribution
The Highly Cited badge is intended to make original reporting and influential coverage more visible. For publishers, that reinforces the practical value of clearly distinguishing first-hand findings from summaries of work produced elsewhere.
Original contribution may include reporting, direct documentation, expert analysis, primary data owned by the publisher, or a clear synthesis that helps readers understand complicated evidence. The exact form depends on the organization and topic.
- Identify the source of factual claims clearly.
- Link to primary materials when they are available and relevant.
- Explain what the publisher independently verified or contributed.
- Distinguish reporting from opinion, interpretation, and prediction.
- Correct substantive errors transparently and update developing coverage promptly.
These practices support reader trust even when they do not result in a badge or carousel placement. No badge or link format should be treated as guaranteed.
Make timely pages genuinely current
For a developing subject, a page should make its state clear. Readers benefit from knowing what is confirmed, what has changed, and which details remain unresolved.
Update timestamps alone are not a substitute for meaningful revisions. When adding new information, integrate it into the page so that the line, introduction, key facts, and supporting context remain aligned.
Offer a perspective that adds value
Google says prominent carousels can surface different perspectives. That creates room for expert interpretation, regional knowledge, specialist reporting, and other useful viewpoints, but difference should come from substance rather than artificial controversy.
A strong perspective explains its evidence and limitations. It helps the reader understand why the analysis differs, which assumptions are being made, and what new information could change the conclusion.
Strengthen source identity
Preferred Sources let users actively choose sites they want to see highlighted. Publishers cannot force that choice, but they can make their identity, coverage area, and editorial value easy to understand.
Consistent expertise and a recognizable publishing proposition give readers a reason to return. If asking loyal readers to select a site as a Preferred Source fits the audience relationship, explain what the feature does without promising that every article will appear prominently.
Prepare pages for preview-based decisions
Hover previews reduce the distance between seeing a link and assessing its likely usefulness. Pages should therefore communicate their purpose quickly through accurate titles, clear openings, and content that directly addresses the expected need.
Avoid making the introduction so vague that a preview cannot reveal the page’s value. At the same time, do not compress every nuance into the first sentence; the opening should orient the reader and establish why the page is credible.
Interpret carousel visibility without overstating results
AI Mode monitoring can answer several important questions, but not every question. A careful report distinguishes observed facts, data-supported trends, and working hypotheses.
What the available evidence can show
Search Console’s generative-AI reporting can show performance trends for AI Mode and AI Overviews over the supported timeframes. Manual audits can show how a source appeared for a specific query in a specific observed context.
Together, these sources can identify pages gaining exposure, topics associated with recurring appearances, and link treatments that are visible during sampled reviews. They can also help teams find mismatches, such as Google surfacing an older page when a newer resource would better serve the reader.
What it cannot safely prove
An observed carousel does not reveal the complete selection logic behind it. A Search Console impression does not necessarily identify whether the source appeared in a carousel, inline link, or another generative-AI link treatment unless the reporting itself provides that distinction.
Likewise, correlation does not prove causation. If visibility rises after an update, the change is worth documenting, but other factors may have contributed.
- Do not estimate total carousel market share from a small manual query sample.
- Do not call all AI impressions carousel impressions.
- Do not present a personalized Preferred Source result as universal.
- Do not promise that original reporting will receive a Highly Cited badge.
- Do not attribute a performance change to one optimization without supporting evidence.
- Do not assume today’s visual layout will remain stable while link designs are still being tested.
Use confidence levels in internal reporting
A simple evidence scale can improve decision-making. Label a finding as confirmed when directly supported by reporting or a recorded observation, probable when supported by several consistent signals, and exploratory when it is only a hypothesis requiring more review.
For example, “AI Mode impressions increased in the weekly report” can be a confirmed measurement. “The increase came from a newly introduced carousel” would remain a hypothesis unless the available data or repeated observations directly support that explanation.
Create an operating cadence for AI Mode carousel monitoring
A repeatable cadence prevents teams from reacting to every isolated result. It also creates a history that becomes more valuable as Google changes link designs and expands connections between AI Mode, websites, subscriptions, and apps.
During active, developing coverage
Review hourly or daily Search Console data when available and appropriate. Check a small set of high-priority queries, confirm that the strongest current page is accessible, and update the content when the underlying facts materially change.
Record observations without repeatedly rewriting the strategy around each fluctuation. Developing-topic results can change as new information and perspectives become available.
Every week
Review AI visibility by major page and topic group. Compare newly visible pages with the editorial calendar, identify recurring carousel observations, and inspect pages that receive exposure but do not meet expected on-site outcomes.
Use the review to produce a short action list. Suitable actions might include updating a stale explainer, consolidating overlapping pages, strengthening primary sourcing, or improving the opening of a page that appears in previews.
Every month
Assess whether AI Mode and AI Overview visibility is becoming more concentrated or more distributed across the site. Review the balance among timely reporting, original work, analysis, and evergreen material.
Also revisit the manual query sample. Add emerging topics that matter, retire irrelevant queries, and retain a stable core so month-to-month observations remain comparable.
When Google changes the experience
Treat product changes as measurement events. Google’s 2026 updates included Preferred Sources, search carousels, inline-link previews, subscription links, a side-by-side Chrome experience, and continuing tests of new link designs.
In July 2026, Google also said users could securely connect services such as Instacart, Canva, and YouTube Music directly inside AI Mode. These app connections are not the same as web link carousels, but they show that AI Mode is becoming a broader exploration and action environment.
When a new feature appears, ask three practical questions:
- Does it change visibility? Determine whether the feature introduces a new way for the site or its content to be surfaced.
- Does it change interaction? Evaluate whether users preview, open, compare, or act on a source differently.
- Does it change measurement? Check whether reporting can distinguish the new experience or whether manual observation is still needed.
The strongest monitoring program will not depend on one screenshot, one query, or one metric. It will combine Google’s Search Console reporting, a disciplined observation sample, editorial records, and outcome data while clearly stating what each source can and cannot establish.
Start by defining priority topics, preserve a baseline of AI Mode and AI Overview impressions, and audit representative link treatments consistently. As Google continues making AI Mode more link-forward, this process will help teams recognize meaningful visibility changes without mistaking product experiments or personalized results for universal performance.
