Ads in Google AI Mode: What's Actually Buyable in 2026

Google announced two ad formats inside AI Mode at Google Marketing Live on May 20, 2026: Conversational Discovery ads and Highlighted Answers. Both are labeled Sponsored. Both rely on Gemini to build creative tailored to the specific query a user typed. Neither is targeted by keywords in the way you've spent the last decade buying search.

If you run paid search for an SMB lead-gen business and you're trying to decide whether to care about this, the honest read is: care, but slowly. Most of what was announced isn't broadly buyable yet. The behavioral data behind it, though, is the part that should change how you plan Q3 and Q4 budgets.

TL;DR

• Google announced two AI Mode ad formats on May 20, 2026: Conversational Discovery and Highlighted Answers, both labeled Sponsored.

• Gemini generates the creative per query — you give signals, the model writes the ad.

• Pew found AI Overviews cut traditional-result clicks from 15% to 8%, and only 1% click sources cited inside the answer.

• AI Mode, including AI agents in Search, rolls out in the US in summer 2026.

• For SMBs: keep your performance accounts tight, opt into AI surfaces deliberately, and rebuild measurement before scaling spend.

1. What AI Mode is and where the rollout stands

AI Mode is the conversational layer Google is bolting onto Search. Instead of returning a ranked list of links, it returns a synthesized answer the user can talk back to — ask a follow-up, refine the constraint, request a recommendation. Per Press Gazette's coverage of the announcement, the expansion — including AI agents working inside Search — rolls out in the US in summer 2026.

For a working definition: AI Overviews are the summary box you've already seen above the blue links. AI Mode is the full conversational surface, a separate experience where the entire session is a dialogue with Gemini. The ad formats announced in May live inside that surface, not inside the classic SERP.

This matters for one practical reason: the inventory is new, the auction dynamics are still being calibrated, and the share of US queries flowing through AI Mode is small but climbing through the back half of 2026. You're not late. You're not on time either. You're in the window where opt-in decisions get made before the data exists to validate them.

2. The two new ad formats explained

Conversational Discovery ads and Highlighted Answers are different products solving different moments in the same session.

Conversational Discovery ads appear inside the back-and-forth. A user asks AI Mode for help picking a CRM for a 12-person services firm. They get a synthesized recommendation. They ask a follow-up — "which of these has the best email integration?" — and the next turn can include a Sponsored placement Gemini built specifically for that refined intent. The ad is part of the conversation, not a banner pasted next to it.

Highlighted Answers sit inside the recommendation set itself. When AI Mode produces a list of options — vendors, products, services — one slot in that list can be a Sponsored entry, again with creative Gemini assembled from the advertiser's feeds, assets, and landing pages. The user sees an answer that includes a paid recommendation, labeled.

Both formats, per Google's announcement, carry the Sponsored label. Neither is keyword-bid in the traditional sense. You're not buying "best crm for small business." You're authorizing Google to enter your assets into a model-driven match against any conversation where the intent fits.

3. How matching works when there is no keyword

This is the part most SMB operators get wrong on the first read. There is no keyword list. There is no negative keyword list in the classic sense. The match is built from three inputs: the user's current query, the conversational context (what they asked two turns ago, what they refined), and your asset library — landing pages, product feeds, campaign signals, performance history.

Gemini reads all of that and decides, per turn, whether your offer fits and what the ad should say. The creative is generated on the spot. You don't write 15 headlines and 4 descriptions and let the model recombine them. You give it source material — pages, products, brand language — and the model writes the placement.

If this sounds familiar, it should. It's the same operating model behind Performance Max and the Ask Advisor layer Google has been rolling into account setup; we wrote about that in Google Ads Ask Advisor. The implication is the same here: your control surface is your feed and your landing pages, not your match types. If your product pages are thin, the model has nothing to work with. If your feed has missing attributes, you don't show up for the conversations where those attributes are the deciding factor.

Where the Sponsored label appears in each AI Mode ad path

4. The click economy these ads live in

The behavioral data is the part you can't ignore, regardless of whether you opt into AI Mode placements this quarter.

Pew Research's July 2025 study found that when an AI Overview is present, only 8% of visits result in a click to a traditional result, compared to 15% when no AI summary appears. Clicks to sources cited inside the AI answer itself? About 1%.

Read that 1% twice. The links Google puts directly inside the synthesized answer — the ones that look like the obvious destination — get clicked roughly one time in a hundred. The organic click pool is shrinking on queries where AI summaries appear, which is most informational queries and a growing share of commercial-research queries. We covered the upstream version of this dynamic in AI Overviews' impact on Google Ads.

What does this mean for an SMB lead-gen advertiser? Two practical things. First, the ceiling on free traffic from informational content is dropping faster than most blogs are updating their forecasts. Second, paid placements inside AI surfaces aren't a vanity buy — they're increasingly where the click is, because the click outside the answer is disappearing.

That doesn't mean you should pour budget in tomorrow. It means the calculation of "can I afford to ignore this" has a different answer in 2026 than it had in 2024.

5. What SMB advertisers should do now vs deliberately wait on

The honest split:

Do now. Audit your landing pages and product feeds as if they were the entire ad. Because under model-built creative, they basically are. Fix thin pages, missing schema, sparse product attributes, weak comparison content. If Gemini is going to write your ad from your assets, your assets need to be worth writing from.

Do now. Decide your opt-in posture for AI surfaces inside your existing campaigns. Performance Max and broad-match search campaigns already serve into AI Overviews and adjacent AI experiences. If you've left expansion settings on default, you're already in. Read your campaign settings; don't assume.

Do now. Tighten conversion tracking. Server-side, enhanced conversions, offline conversion import if you do phone-led sales. Measurement inside AI surfaces is murkier than classic search, and the worst position to be in is scaling spend into a black box.

Wait on. Building bespoke campaigns specifically optimized for AI Mode placements. The inventory is too new, the reporting too thin, and the auction dynamics too unsettled to justify a separate campaign structure in June 2026. Revisit in Q4 when more accounts have meaningful AI Mode impression data.

Wait on. Repricing your entire offer around "conversational commerce." Conversations on AI Mode are still mostly research, not transaction. The transaction still happens on your site, your booking flow, your sales call. Build for that.

If you want a deeper read on where paid budget sits across these new surfaces, we keep a running framework on the paid ads service page.

6. Measurement caveats inside AI surfaces

A few things to know before you read any AI Mode performance report at face value.

Attribution windows behave differently when the user's path includes a multi-turn conversation. A user who refines their question three times before clicking is, behaviorally, a different prospect than a user who clicks on turn one. Current Google Ads reporting collapses both into the same impression-and-click frame. Expect that to evolve through late 2026; don't expect it to be solved in June.

Click quality is genuinely different. A click from inside a conversation where the user has already specified budget, team size, and use case is qualified in a way a broad-match search click rarely is. In our client work, early Performance Max accounts with strong feeds have seen better-than-average conversion rates on AI-surface-attributed clicks — but the sample sizes are small and we'd hedge anyone who claims a clean benchmark today.

View-through and assist credit get noisier. When the synthesized answer mentions your brand without a click, that's brand impact you can't directly measure in the ads UI. Build a habit of reading brand-search lift alongside your AI surface spend, not after it.

7. FAQ

Are AI Mode ads opt-in or do my existing campaigns serve into them automatically?

Existing Performance Max and broad-match search campaigns already serve into AI-adjacent surfaces by default. The new Conversational Discovery and Highlighted Answers formats announced May 20, 2026 are rolling out through existing campaign types — meaning if your expansion settings are on, you're likely eligible. Check your campaign-level settings rather than assuming.

How are AI Mode ads labeled, and will users actually notice?

Both Conversational Discovery and Highlighted Answers carry a Sponsored label per Google's announcement. The label sits on or near the ad placement inside the conversation or recommendation list. Whether users notice is the open question — disclosure research consistently shows in-line labels get lower recognition than separated ad blocks, and that's worth factoring into your brand-safety posture.

What about brand safety — can I control which conversations my ad appears in?

The brand-safety controls available at launch mirror what exists for Performance Max: account-level brand exclusions, negative keywords at the account level, and content-type exclusions. There is no per-conversation control. If brand adjacency matters to you, lean on placement exclusions and audit reports rather than expecting query-level transparency.

Can I bid specifically for AI Mode placements?

Not as a separate inventory in June 2026. Bidding remains campaign-level under Smart Bidding, with the model allocating across surfaces including AI Mode. You set the target CPA or ROAS; Google decides where the impression serves. Separate AI Mode bidding controls have not been announced.

Does Gemini-generated creative mean I lose control of my ad copy?

You lose direct control of the final string. You keep upstream control through asset quality: landing pages, product feeds, brand guidelines you upload, and the campaign-level assets you provide. Treat your asset library as the editorial layer. If you wouldn't approve a sentence pulled from your landing page, fix the landing page.

What to do this week

Pick one campaign. Pull the landing pages and feeds it relies on. Read them the way Gemini will — as the only source material the model has to build your ad from. Fix the three weakest pages by Friday. That's the work that pays off whether you opt into AI Mode placements in July or wait until November.

When you want a second set of eyes on the asset-and-feed audit, tell us what you're running and we'll take a look.