Google Ads Sitelinks in AI Mode: What AEO Teams Should Do

4 October 2026 · The Cresia team

Google appears to be testing sitelinks on ads inside AI Mode, and for AEO and GEO teams the practical answer is: don't change your organic plan, but do change how you account for the surface. AI Mode is becoming a place where paid units and cited sources sit in the same answer, and your reporting has to separate them. The pages those sitelinks would point to also need to hold up under a click from someone who just read a synthesized answer.

Key takeaways

Search Engine Roundtable reports Google is testing ad formats in AI Mode, with a screenshot of what may be sitelinks. It is a test and could change or disappear.
The format matters less than the direction: AI Mode is becoming a surface where paid and organic results share one answer.
Citation and ad eligibility are separate systems. Winning one does not win the other, so plan and report them separately.
Sitelinks send people to specific pages, so the pages behind them need to answer a question in the first screen.
Measure ad impressions, citations and branded demand in separate columns before drawing any conclusions about causes.
Do not rewrite your content strategy around one screenshot. Fix structured data, page clarity and tracking first.

What Google is testing with sitelinks in AI Mode

Search Engine Roundtable reported that Google Ads now show up often inside AI Mode, and that Google is trying different ad formats to see which earns the best click-through rate. The post includes a screenshot of what may be sitelinks attached to an ad in an AI Mode answer. The wording matters: may be. This is an observed test, not an announced product, and nothing in the report says it has rolled out widely.

Sitelinks are familiar from classic search ads. They are extra links beneath the main ad, each pointing to a different page on the advertiser's site: pricing, a category, a support page, a store locator. In a standard results page they widen the footprint of the ad and let a searcher skip a step.

In AI Mode, the same idea has a different job. The answer itself is long, conversational and often complete enough that the reader never clicks anything. Search Engine Roundtable's framing is that Google wants to find which ad formats get clicks in that environment, in contrast to the links inside the AI answer. Read that as a hint about the business problem Google is solving: an answer interface that satisfies the reader is hard to monetise, so formats get tested until something works.

For your team the useful part is not the sitelink itself. It is that Google is iterating on paid presentation inside the AI answer, which means the space around organic citations is being actively redesigned. Anything you learn about layout this month may be stale by next quarter.

Why paid formats in AI Mode matter for AEO and GEO

Most AEO and GEO work so far has been framed as an organic discipline: earn mentions and citations in generated answers. A paid unit inside the same answer changes that framing in three ways.

First, attention is shared. If an ad with sitelinks sits near the top of an AI Mode response, the cited sources compete with it for the same click. A brand that is cited but outbid by a competitor's ad may see its citation share hold steady while its traffic falls. Your dashboard would call that stable. Your pipeline would not.

Second, the line between recommendation and advertisement gets harder for readers to draw. That is a trust problem for everyone, and it is a reason to be careful about how you describe AI visibility internally. A citation is an editorial selection by the engine. An ad is a purchase. Reports that blur the two will send the wrong budget to the wrong team.

Third, the ownership question gets awkward. In many companies the SEO lead owns citations and the paid media lead owns Google Ads. Neither owns the combined view of what a user sees when they ask a question in AI Mode. Someone has to, and it is better to decide that now than after a disagreement over attribution.

None of this requires panic. Paid placements on search surfaces are not new. What is new is that the surface is a generated answer whose layout is still in flux.

How AI answer engines choose and cite sources

Sitelinks do not change how an engine picks sources for the answer itself. Ads and citations come from separate systems with separate rules. An advertiser's auction position does not make its page more likely to be quoted, and a well-cited page does not earn ad eligibility.

What decides citations is a mix of things that have been consistent across engines, though each vendor weights them differently and none publishes the full recipe. In practice, these are the factors teams can influence:

Retrievability. The page has to be crawlable, render its main content without needing a script to run, and load reliably. An engine cannot cite what its fetcher cannot read.
Passage clarity. Engines tend to lift short, self-contained passages that answer one question. A paragraph that opens with the answer and then qualifies it is easier to quote than one that builds slowly to a point.
Entity clarity. Who the company is, what the product does, who wrote the page and when it was updated should be stated plainly, not implied.
Structured data that matches the page. Search Engine Journal has covered the structured data mistakes that hurt AI visibility; the recurring theme is markup that contradicts or outruns what the visible page says.
Corroboration. Claims that appear consistently on your site and on independent pages are safer for an engine to repeat than claims that appear once.

Search Engine Journal has also published a piece on finding the exact phrases that get content cited in AI search. The useful habit it points toward is working from the language real prompts use, rather than from the language your product team prefers. If your category page says 'unified engagement platform' and buyers ask for 'tools to test landing page variants', the second phrase is the one to be findable under.

Vendors change retrieval behaviour without notice. For the current crawler names, controls and documented behaviour, check each vendor's own documentation, and for Google specifically the Search Central documentation.

What to change in practice

The sitelink test does not demand a new playbook. It sharpens a few existing priorities, and it adds one new question: what happens when someone lands on your page from an AI answer, with an ad or without?

Start with the pages. A sitelink is a promise that a specific page answers a specific need. The same is true of a citation. Someone who clicks through from an AI Mode answer has already read a summary, so they arrive with a narrower question than a typical search visitor. They want the detail the summary left out.

That shapes a few concrete changes:

Lead each key page with its answer. The first screen should state what the page covers, who it is for, and the one fact a skimming reader needs. A hero image and a slogan do not do that job.
Audit your top ten landing pages for the post-summary visitor. Ask what a person who has already read a three-paragraph answer would still need. Put that first.
Check markup against the visible page. If your product schema lists a price or availability the page does not show, fix one or the other. Mismatches are exactly the type of structured data error Search Engine Journal describes.
Name your people and dates. Authors, reviewers and last-updated dates should be visible and match your structured data.
Make the paid and organic landing pages consistent. If the ad sitelink says 'Pricing' and the page it leads to hides prices behind a form, the click is wasted whichever system sent it.

Teams that run many landing pages face a version-control problem here, because a page that is accurate in March can be wrong in September. A shared system for creative and content assets helps; the Mosaic product page describes how Cresia approaches creative management. If you want to test whether a clearer first screen changes outcomes, what is A/B testing covers the method, and the Evolve product page covers the tooling.

How to measure AI Mode visibility when ads and citations mix

The measurement trap is simple: one number for 'AI Mode performance' that adds up paid and organic. Do not build it. Keep the columns apart, because the decisions they drive are different.

Signal	Where to look	Owner	What it tells you	What it does not tell you
Ad impressions and clicks in AI surfaces	Google Ads reporting; check the vendor's documentation for how AI placements are labelled	Paid media lead	Whether your ads are being served in the new format	Whether your content is being cited
Citation presence for a fixed prompt set	Manual checks or a visibility tool, repeated on a schedule	SEO or AEO lead	Whether engines name or link your pages for the questions you care about	Traffic or revenue
Referral sessions from AI engines	Web analytics, segmented by referrer	Analytics lead	Whether cited pages produce visits	Whether the visits came from a citation or an ad
Branded search and direct traffic	Search Console and analytics	Growth lead	Whether being mentioned lifts demand you never see as a click	Which engine caused it
Landing page conversion	Analytics and experimentation tooling	Growth or CRO lead	Whether the page works for the post-summary visitor	Why a visitor arrived

Two habits make this table useful. First, fix your prompt set. Pick thirty to fifty real questions your buyers ask, phrase them the way buyers phrase them, and run the same list every month. Changing the list each time makes trends meaningless.

Second, tag your landing pages consistently, so a visit from a paid sitelink and a visit from a cited link do not collapse into one bucket. If your tracking is inconsistent across properties, a written tracking specification is the cheapest fix, and the OmniSpec product page shows how Cresia frames analytics governance.

Be honest in reports about what you cannot see. Engines do not expose full citation data, and ad reporting for AI placements may lag the format changes. A gap in the data should appear as a gap in the report, not as an estimate presented as a measurement.

What not to do

When a new format appears, the pull is to react to it. Most reactions are wasted effort.

Don't redesign your content strategy around a screenshot. One reported test of an ad format tells you where Google is experimenting. It does not tell you what will ship, to whom, or for how long.
Don't treat ad spend as a citation tactic. The systems are separate. Spending more on ads will not make an engine quote you.
Don't write pages for sitelinks. Write pages for the questions buyers ask. Sitelinks, citations and organic listings will all pick from the same good pages.
Don't stuff pages with FAQ markup in the hope of being quoted. Markup that the visible page does not support is a liability, not a lever.
Don't chase a single engine's quirks. Behaviour shifts without notice. Durable fixes, such as clear passages, accurate markup and fast pages, help everywhere.
Don't merge paid and organic into one KPI. You will lose the ability to tell what is working.

And a note on pace. Teams under pressure to 'do something about AI Mode' often ship low-quality pages quickly. A smaller set of accurate, well-maintained pages beats a larger set of thin ones, both for readers and for the engines selecting among them.

A practical plan for the next month

If you need something to bring to a planning meeting, here is a sequence that fits in four weeks and does not depend on the sitelink test going anywhere.

In the first week, agree who owns the combined view. A paid media lead and an SEO lead sitting down for an hour to define shared terms will prevent months of confused reporting. If your team is structured around channels rather than outcomes, the media teams solution page describes how Cresia thinks about that coordination, and the MediaPilot page covers its media and AI-search visibility focus.

In the second week, build the prompt set and run the first baseline. Do not interpret it yet. You need at least two or three cycles before a movement means anything.

In the third week, audit your ten most important landing pages for first-screen clarity and markup accuracy. Fix the obvious mismatches. Leave stylistic debates for later.

In the fourth week, review the baseline, the page fixes and the tracking, and decide what to test next. If Google has widened or dropped the sitelink test by then, you will have lost nothing, because none of the work depended on it.

Frequently asked questions

Are sitelinks in AI Mode ads live for everyone?

Not as far as the public reporting shows. Search Engine Roundtable described a screenshot of what may be sitelinks and said Google is testing formats, which points to an experiment rather than a launch. Check Google's own Ads documentation and your account for the current status.

Does running Google Ads make my content more likely to be cited in AI Mode?

There is no public evidence that it does, and the two systems serve different purposes. Citations depend on how well your pages answer a question and how easily an engine can retrieve them. Treat ad spend and citation work as separate budgets with separate goals.

What should an AEO team change first because of this?

Start with the landing pages your best prompts should lead to. Make the first screen answer the question, make sure structured data matches the visible page, and keep authors and dates current. Those changes help citations, organic listings and any paid sitelink that points at the same page.

How do I report AI Mode performance when ads and citations both appear?

Keep them in separate columns: ad impressions and clicks from your ad reporting, citation presence from a fixed prompt set, and referral visits from analytics. Add branded search as a third lens on demand. Avoid a single combined score, because it hides which lever moved.

Should I wait to see what Google does before acting?

Wait on anything specific to the sitelink format. Do not wait on page clarity, markup accuracy or tracking hygiene, because they pay off whatever Google decides. A month of that work leaves you ready for the format if it ships and no worse off if it does not.

Sources

https://www.seroundtable.com/google-ads-sitelinks-ai-mode-42206.html
https://www.searchenginejournal.com/what-are-common-structured-data-mistakes-that-hurt-ai-visibility-ask-an-seo/589924/
https://www.searchenginejournal.com/where-to-find-the-exact-phrases-that-get-your-content-cited-in-ai-search/591562/