Two new AI Mode ad formats are in testing in the US right now: Conversational Discovery Ads and Highlighted Answers. Both share a property that distinguishes them from every Google ad format that came before. Gemini writes the actual creative. You supply the inputs; Google's AI builds what your user sees.
This isn't a bidding change or a new placement. It's a shift in who composes the ad.
What launched and where it stands
Google announced both formats at Marketing Live on May 20. As of late May they're in live testing on US mobile and desktop. There's no confirmed broad rollout date, but the testing footprint is expanding and early-access accounts are already running them.
Eligibility requires AI Max for Search, AI Max for Shopping, or Performance Max. Standard Search campaigns don't qualify for either format. This is part of Google's consistent direction: AI Mode participation lives behind AI Max and PMax. If your campaigns still run on old Standard Search with no AI Max toggle, you're sitting outside the formats where AI Mode impression share is growing fastest.
How Conversational Discovery Ads work
When a user types a specific question into AI Mode — "what's the best protein powder for someone who works out after dinner and hates chalky texture?" — Gemini reads that query, pulls from your product data, landing page content, and uploaded creative assets, then composes tailored ad copy for that specific question. The user doesn't see the headline you wrote. They see what Gemini generated for their query.
That's the change. Previously: you write ad copy, Google matches it to searches. Now: you provide raw material, Gemini generates the actual headline in real time.
The ad carries a Sponsored label. Gemini also generates an independent explainer alongside the creative — a short note explaining why this result fits the user's question. Both the headline and the explainer are Gemini's output. Neither came from your Ads account.
For advertisers who've spent years A/B testing headline combinations, this takes some adjustment. The static copy you upload is no longer the ad. It's the brief.
How Highlighted Answers work
Highlighted Answers sit inside AI Mode's recommendation lists. When someone asks for suggestions — "running shoes for overpronation under $150" or "best budgeting apps for freelancers" — AI Mode returns a curated response with multiple options. Your ad can appear as one of those items, formatted as a recommendation with a Gemini-written explanation of why it matches the query.
You're not sitting beside the results. You're inside them. If the AI surfaces five options and your product is one, the Sponsored label is the only signal separating your placement from the organic recommendations. For product categories where people research before buying — apparel, supplements, software, consumer electronics — this is a meaningful placement to understand.
What determines whether your product makes the list is the same thing that determines Conversational Discovery quality: how much specific, structured information about your product Google's AI can actually find and process.
The creative brief has inverted
The structure has always been: you write headlines and descriptions, Google selects and rotates combinations. Responsive Search Ads gave you 15 headline options; Google mixed and matched. Dynamic Search Ads pulled copy from your landing page. In all of these, the final ad was assembled from language you created.
Conversational Discovery inverts that. You don't write what the user sees. Your landing page content, product descriptions, and supplemental feed attributes become source material Gemini draws from when composing ad creative.
If your landing page is thin, Gemini has thin inputs. If your product description is a feature list with no context about who the product is for, the AI-generated creative reflects that gap. The quality ceiling on your ad is now the quality of your content and data — not your headline copy.
Most paid search teams spend significant time crafting and testing headline variations. For these two formats, that work doesn't transfer. The new leverage points are landing page depth, product feed attribute coverage, and asset variety.
What to do before the rollout widens
Your landing page content needs to answer specific questions. Not just describe the product — explain who uses it, in what context, for what problem, and how it compares to the obvious alternative. AI Mode queries are question-shaped, longer and more specific than traditional searches. Your page needs enough structured information for Gemini to compose a confident, relevant answer.
Your product feed should go deeper than title optimization. The product_faq and product_use_cases attributes that Google launched in Merchant Center at GML are the direct inputs Gemini pulls from when building AI Mode ad creative. If you haven't added those fields yet for your top revenue SKUs, that's the first practical step — not a campaign rebuild, just a supplemental data source addition.
If your full catalog runs on standard Search with zero PMax or AI Max presence, you're not eligible for either format today. That's not a reason to panic-migrate everything. It is a reason to identify which products you'd want appearing in AI Mode recommendation lists and verify they have PMax coverage before these formats expand.
The formats are in testing. Save your campaign restructuring for when they're out of beta. Use the time to treat your product data and landing page copy as creative assets, not operational housekeeping. That improvement compounds across every AI Mode surface, not just these two formats.
If you want a quick read on where your account stands on feed quality and landing page depth right now, the free audit at Gromerce surfaces the gaps in a few minutes.
The headline your customers see in AI Mode is no longer yours to write. Give Gemini better material to work with.
Sources: Search Engine Land, Google Blog (Google Marketing Live 2026), Search Engine Journal, May 2026
