TL;DR

Recent reporting indicates Google switched on news carousels inside AI Overviews this week, with AI-generated images reportedly on the way. Separately, AI Overviews and AI Mode reportedly now show pop-up preview links when you hover over citations in the AI-generated answer. We're hedging both of those because this is fresh, fast-moving reporting, not a Google announcement with a spec sheet.

Here's the part that matters if you publish anything time-sensitive: a news carousel is a third, distinct slot inside the AI Overview box — separate from the citation links already sitting in the answer text, and separate from whatever a hover-preview surfaces. Three slots means three things to check, not one. If you've been tracking "am I cited in the AI Overview," that question just got more specific: cited in the text, or featured in the carousel, or neither?

ScrapeMaster is the tool for the tracking half of this — point it at the AI Overview box after you search your own queries, let it read the rendered structure (it works fine on JS-heavy, dynamic UIs like this one), and log what's in the main answer, what's in the carousel, and what's hover-only. Do that on a schedule and you'll actually see if your site starts showing up in the new slot, instead of guessing from a single screenshot.

What reportedly changed this week

To be precise about what we do and don't know, going only on recent reporting:

  • Google reportedly switched on a news carousel inside AI Overviews this week — a horizontal, card-style module of news sources surfaced for a given query, sitting inside the AI Overview box itself.
  • AI-generated images are reportedly on the way for AI Overviews, though not confirmed as live yet.
  • AI Overviews and AI Mode reportedly now show hover-preview pop-ups when a user hovers over a citation link inside the AI-generated answer — a preview card rather than an immediate navigation.

That's the full list of confirmed-by-reporting facts. We're not going to attach a rollout percentage, a precise launch date beyond "this week," or additional feature claims to any of this — nobody outside Google has that level of detail yet, and anyone quoting a specific number is getting ahead of the reporting. Treat the shape of the change as reliable; treat any specifics beyond it as unverified until Google documents it directly (which, per our experience with the July 9 AI performance reports, it may never fully do).

The existing AEO conversation — the one we've written about before — is about the citation links embedded in an AI Overview's generated text: the numbered footnotes and the source list underneath the answer. That's one surface. It has one selection logic (whatever sources the model drew on to write the paragraph), one visual location (inline and below the text), and one way to check it (read the answer, note the sources).

A news carousel is architecturally different, even though it lives in the same box:

  • Different selection criteria. A news carousel is very likely pulling from a curated set of news sources for time-sensitive queries — closer to how Google News or Top Stories selects than how an LLM picks citations for a generated sentence. Being cited in the text and being featured in the carousel are probably two separate qualification bars.
  • Different visual location. It's a horizontal card row, not an inline footnote. Someone glancing at the AI Overview box will notice the carousel before they read footnote 4.
  • Different competitive set. The carousel likely favors publishers with recent, dated content on the specific query — which may include outlets that don't appear anywhere in the AI Overview's text citations at all.

That means "am I cited in AI Overviews" is no longer one yes/no question. As of this week, it's at least three:

  1. Am I in the main answer text's citation list?
  2. Am I in the news carousel for this query?
  3. Am I only reachable via the hover-preview, meaning a user has to actively hover to even see my link exists?

Treating these as one bucket means you'll miss the case where your competitor gets carousel placement on every breaking story in your category while you're stuck arguing about footnote position. If you publish news, product updates, or anything with a publish date attached, the carousel is the slot worth watching first — it's the one built for exactly your kind of content.

Why this is worth tracking starting now, not after it settles

The honest caveat up front: this is a brand-new feature, reportedly live for less than a week as of this writing. The UI you see today may not match the UI a reader sees next month. Carousel position, card count, what triggers it, and how the hover-preview behaves are all things Google will very likely keep adjusting.

That's an argument for starting to track now, not for waiting until it's stable. Two reasons:

  • You can't retroactively reconstruct a "before" state. If the carousel changes shape in three weeks, the only record of what it looked like at launch is whatever you captured this week. Every one of our posts on AI-surface tracking says some version of "start the time series today" — this is that advice applied to a feature that's days old.
  • Early movers get a cleaner baseline. If your site starts appearing in the carousel in week three, you want a week-one and week-two record showing it wasn't there, so the change is visible in the data rather than something you vaguely remember.

None of this requires waiting for the feature to "finish rolling out." Track what's in front of you today; update your method when the UI updates.

The mechanics are close to what we've described for general AI Overview tracking, with one addition: you're now logging three zones inside the box instead of one.

Step 1 — Fix a query list that's actually relevant to a news carousel. General AEO watchlists (best-X, X-vs-Y) mostly won't trigger a news carousel — it's tied to timely, newsy queries. Build a separate short list: your brand name, your product launches, your industry's current events, breaking topics you publish about quickly. 15–30 queries is plenty to start; this is a narrower, faster-moving list than a general citation watchlist.

Step 2 — Search each query and let the AI Overview render fully. Give the box a moment to finish loading before extracting — these are dynamic, JS-rendered UIs, and grabbing a half-rendered state gives you an incomplete row.

Step 3 — Open ScrapeMaster and read the structure, not just the text. ScrapeMaster's auto-detect reads the page as rendered — it works on JS-heavy, dynamic interfaces because it looks at the DOM after the page has finished doing its thing, not a static fetch. Point it at the AI Overview box and let it find the repeating structure. Rename the columns it suggests to something that will still make sense to you in a month: something like zone (main_text / carousel / hover_preview), source_domain, source_title, position_in_zone, query, date.

Step 4 — Capture all three zones as separate rows, not one blended row. This is the one change from a standard AI Overview citation extraction. For a given query, you might get: 4 rows from the main answer's citation list, 3 rows from the carousel, and however many domains show up only when you hover a citation (log those too — a hover-only source is meaningfully different from one confidently sitting in the visible text). Tag the zone on every row.

Step 5 — Export and append, don't overwrite. CSV, XLSX, or JSON, or straight to clipboard for Google Sheets. Keep one running sheet with a date column — the date is what turns a snapshot into a time series. Do this weekly for your news-relevant query list; a fast-moving feature rewards more frequent sampling than a slow AEO watchlist would.

Step 6 — Watch the carousel column specifically. Once you have three or four weekly samples, you can ask questions the main-text citation data alone can't answer: Does your site ever make the carousel, even on queries where it doesn't make the main citation list? Which competitors show up in the carousel consistently? Does carousel presence correlate with how recently you published on that topic?

Because ScrapeMaster reads whatever is rendered in front of you — logged in or out, whatever layout you're served — it doesn't need updating every time Google tweaks the carousel's card count or the hover-preview's trigger delay. The column-by-column structure adapts to what's on the page; you just keep sampling.

What this doesn't tell you (and isn't meant to)

Worth being direct about the limits, the same way we've been on the AI performance reports gap: this workflow shows you what's in the box, not how many people saw it or whether anyone clicked the carousel card. Google Search Console's AI performance reports don't break out carousel clicks separately from AI Overview impressions generally, and nothing here changes that. Pair your carousel-tracking sheet with whatever GSC gives you, and be honest in any report you write that this is a convenience sample of your own queries — not a market-wide panel study of carousel behavior.

It also won't tell you why a source made the carousel and another didn't — Google hasn't published the selection logic, and this week's reporting doesn't include one either. What you get is an observational record: presence, position, and change over time. That's still meaningfully more than nothing, and it's the only version of "ground truth" available until Google documents the feature itself.

The hover-preview deserves its own line in your log for a reason beyond completeness: a source that only surfaces on hover is functionally less visible than one sitting in visible text or a carousel card. If a user has to deliberately mouse over a citation marker to discover your link exists, that's a weaker placement than a carousel card competing for attention on page-load. Tracking hover-only appearances separately lets you see whether your presence in AI Overviews is concentrated in the low-visibility tier — worth knowing even if you can't yet quantify how much less traffic it drives.

Frequently asked questions

They're reportedly distinct — Top Stories is its own established SERP module that's existed for years, separate from AI Overviews entirely. The one described in this week's reporting sits inside the AI Overview box specifically. Don't assume tracking one gives you data on the other; log them as separate zones if you're watching both.

The reporting we've seen doesn't specify a login requirement, and Google search results, including AI Overviews, are publicly viewable either way. Track both logged-in and logged-out states if you suspect personalization is affecting what you see — AI Overview content is known to vary by account and location.

Can ScrapeMaster click through the hover-preview automatically?

No. ScrapeMaster extracts what's visible and rendered on the page — it doesn't simulate hover states or drive interactions to reveal hidden content. If a source only appears via hover-preview, you'll need to hover it yourself before capturing that row, the same way you'd need to scroll a page before ScrapeMaster can extract lazy-loaded content further down it.

Recent reporting ties it to news-relevant, timely queries — it's not described as a universal AI Overview element. Build your tracking query list around topics where a news carousel would plausibly make sense (breaking news, product launches, current events in your category) rather than your full general-purpose AEO watchlist.

Different product, different mechanism. Our ChatGPT/Perplexity citation post covers those surfaces' citation panels. This carousel is a Google-specific, AI-Overview-specific module tied (per this week's reporting) to news content — it doesn't have a direct equivalent in Perplexity's sources panel or ChatGPT's browsing citations. Track them as separate rows in separate query lists; conflating them will blur what's actually a Google-only feature change.

Should I change my query watchlist just for this feature?

Add a short, separate list rather than replacing your existing AEO watchlist. Your general "best X" and comparison queries mostly won't trigger a news carousel. A focused 15–30 query list of your brand name, launches, and timely industry topics will surface the carousel far more reliably.

Likely, honestly — this is days-old reporting about a feature Google is actively iterating on. The mindset matters more than the exact steps above: identify the distinct zones inside whatever the AI Overview box looks like when you read this, tag each row with which zone it came from, and keep sampling on a schedule. ScrapeMaster's auto-detect adapts to whatever structure is actually on the page, so the workflow holds up even if the visual layout changes.

Bottom line

Recent reporting says Google turned on news carousels inside AI Overviews this week, with AI images reportedly coming and hover-preview citation links reportedly already live. The practical takeaway isn't another "get your content cited" lecture — it's that the AI Overview box just grew a third zone, with its own selection logic and its own visibility profile, sitting alongside the citation text you may already be tracking.

If you publish anything time-sensitive, that carousel is worth a dedicated, narrow watchlist and a weekly sample — logged separately from your main-text citation tracking, tagged by zone, dated every time. ScrapeMaster reads the rendered AI Overview box as-is, handles the fact that it's a dynamic JS interface, and exports a structured row you can append to week over week — free, local, no login bypass, no scraping of anything you couldn't already see with your own eyes on a public search results page.

The feature is new enough that the screenshot in this post would be stale by the time you read it. The habit of tracking it systematically won't be.