Google Just Made AI Visibility a KPI.
Here's What That Changes for Your Brand.

Something shifted at I/O 2026 that most brands missed because it was buried under agent demos and developer tooling. Google quietly turned AI visibility into something you can see, measure, and benchmark inside Merchant Center. 

When the largest ad platform on earth starts exposing your AI share of voice, the category stops being optional.

TL;DR

• Google is rolling out AI performance insights inside Merchant Center, showing brands how they appear across Google AI surfaces and how their share of voice compares to similar brands.
• Conversational attributes are now a feed input, which means product descriptions have to be answer-ready, not just taxonomy-clean.
• Sponsored AI answers are coming to AI Mode and AI Overviews, so the line between earned recommendation and paid placement is going to blur fast.
• Universal Cart and agentic checkout move the goal from citations to actions. Did the AI choose you, explain you correctly, and complete the next step?
• The brands that win will treat AI visibility the way they treated paid search in 2008: a measurable, ownable channel with its own ops, tooling, and weekly review.

Why This Stopped Being Paranoia

At Google I/O 2026, Google made a series of moves that, taken together, formalize AI as a commerce surface with its own measurement, monetization, and action layer. Merchant Center is getting AI performance insights with share-of-voice comparisons against similar brands.

Conversational attributes let retailers update product descriptions for how people actually ask questions in AI shopping. Sponsored AI answers (Conversational Discovery ads and Highlighted Answers) are being tested inside AI Mode, with text and Shopping ads from existing campaigns already eligible to show in AI Overviews.

Underneath all of this, Google said AI Mode has crossed one billion monthly users and queries have more than doubled every quarter since launch. Universal Cart is expanding into Direct Offers, YouTube Shopping ads, and local services. Search agents can now run in the background and ping users when something matches their criteria.

This is happening alongside the rest of the ecosystem. ChatGPT's shopping experience is iterating fast, with public guides now mapping how products get listed and third parties like Minty plugging real-time deals and cashback directly into ChatGPT. The discovery layer is being built in public, by every major model, in parallel.

The real story is this: Google just validated AI visibility as a tracked, ranked, and auctionable surface. Once that view exists in Merchant Center, every executive team in ecomm will demand the same measurement across ChatGPT, Perplexity, Gemini, Claude, and AI Mode. The category is no longer a hypothesis.

What This Looks Like in the Wild

We've seen this across hundreds of brand conversations in the last six months, and three patterns are showing up over and over.

1. The Walmart Pattern: Visible, but the funnel is broken. You're appearing in AI answers, but conversion is a fraction of your owned site. Walmart's ChatGPT checkout reportedly converted three times worse than its own website, and Walmart's response was to embed its own AI assistant inside ChatGPT to own the experience. The lesson for the rest of us: showing up is step one. If the AI cannot explain you, price you, and check you out cleanly, visibility is a vanity metric.

2. The "Hero SKU Trap" Pattern: One product carries you, the rest are invisible. A skincare brand we looked at shows up for its flagship serum in almost every relevant prompt and is essentially absent across the other forty-plus SKUs in its catalog. They have full-catalog SEO. They have zero full-catalog AI presence, because their feed treats every product like a database row instead of like an answer to a question a shopper would actually ask.

3. The Reverse Pattern: Reviewed everywhere, recommended nowhere. A home goods brand with strong off-site coverage (Reddit threads, gift guides, YouTube reviews) shows up in TOFU (top-of-funnel) discovery prompts. The moment the prompt gets specific ("best non-toxic dutch oven under $200"), a competitor with worse reviews but cleaner structured product data wins the citation. Earned trust is real. Eligibility is what gets you selected.

The Four-Layer AI Visibility Audit

Most brands are still running a one-layer check: "Did ChatGPT mention me?" That question was useful in 2025. It is not enough in 2026, because the surfaces are now measured, monetized, and agentic. You need to audit four layers in sequence, because each one feeds the next.

Layer 1: Visibility. "Show me the best [category] for [use case]." Run the same prompt across ChatGPT, Perplexity, Gemini, and Google AI Mode. Are you cited? In what position? With what description?

Layer 2: Share of voice. "Compare [Your Brand] to [Top 3 Competitors]." Who is mentioned more often? Whose attributes are described more accurately? Whose reviews are quoted? This is the exact comparison Google is about to start showing inside Merchant Center, so build a baseline now.

Layer 3: Eligibility. "What is the return policy on [Your Hero SKU]? Is it in stock in size medium?" This tests whether your feed, schema, and on-site content give the model enough structured data to answer specific questions. If the AI hedges or pulls from a third party, you have a feed problem, not a visibility problem.

Layer 4: Actionability. "Add [Your Hero SKU] to my cart and check out." This is the agentic test. With Universal Cart, ChatGPT checkout, and Sparky-style embedded assistants live, the question is no longer whether you are mentioned. The question is whether the AI can transact you. If the agent stalls, you lose the sale to whoever it can complete. AI shopping agents are already reshaping how consumers find deals and complete purchases.

Run It: Practical Checklist

Here is what to actually do in the next two weeks. This is a workstream, not a strategy deck.

1. Stand up an AI visibility baseline with Discover. You need a fixed set of prompts covering TOFU (category discovery), MOFU (comparison), and BOFU (purchase intent), spanning your hero SKU, your weakest SKUs, and your top competitors, run consistently across ChatGPT, Perplexity, Gemini, and Google AI Mode.
   
   Building that by hand and rerunning it every week is exactly the kind of thing that gets done once and abandoned. Yotpo Discover runs it for you: it tracks where you show up, in what position, and how your share of voice moves against competitors over time. That tracked baseline is what turns "are we in AI" from a quarterly guess into a number you watch every week.
   
   
2. Audit your product feed for conversational readiness. Open your Merchant Center feed. For each of your top 20 SKUs, ask: does the description answer a real shopper question, or does it just list specs?
   
   Conversational attributes are now a Google input, which means feed copy is creative work, not catalog work. Rewrite titles and descriptions in the language shoppers actually use in prompts.
   
   
3. Know your citation surface, not just your ranking. Showing up is only half the picture. You need to know where the AI is pulling from when it cites you. Your own site? A Reddit thread? A retailer PDP? A review aggregator?
   
   That source tells you whether your gap is on-site structured content, off-site presence, or both, and it points to completely different fixes. Discover surfaces the source behind each citation alongside your visibility, so instead of guessing why a competitor wins a prompt, you can see whether they beat you on owned content or earned presence and go fix the right one.
   
   
4. Run the agentic checkout test. Pick three of your top SKUs. In ChatGPT (with shopping enabled) and via Google's Universal Cart surfaces where available, try to complete a purchase. Note every point of friction: wrong price, missing variant, broken offer, no buy button. Each of those is a revenue leak you can fix before the surface goes mainstream. The Walmart and OpenAI partnership has been iterating on exactly these friction points, which tells you how early and how fixable this still is.
   
   
5. Assign one owner. This is the step nine out of ten brands skip. AI visibility is not an SEO subtask, a paid media subtask, or a content subtask. It is a horizontal channel that touches feed, schema, content, reviews, offers, and checkout. Pick one person with the authority to coordinate across those teams, and put AI share of voice on the weekly scorecard. The brands moving fastest right now have already done this.

The Bigger Picture:

Google making AI visibility measurable inside Merchant Center is the moment the category becomes real for the enterprise. Once a CFO can see "AI share of voice" next to "impression share" in a Google dashboard, it stops being a 2026 conversation about whether AI matters and starts being a 2026 conversation about whose share is growing and whose is shrinking.

The brands that figure this out early will treat AI the way smart brands treated paid search in 2008 and influencer in 2017: as a measurable channel with its own playbook, its own ops, and its own line in the budget.

The brands that wait for Google to roll the dashboard out to their region will already be behind on the baseline, behind on the feed work, and behind on the off-site presence that drives citations they cannot buy. Adoption of AI shopping agents is already surging, and the gap between early movers and laggards is going to widen fast.

This is the work. Make sure the AI can find you, explain you accurately, trust your data, and complete the sale. Discovered, explained, trusted, and chosen.

Hope this was helpful. If it was, feel free to share it with a friend.

— Tomer

– Tomer

If you haven't assessed how your brand shows up in AI-powered discovery, the tool we launched covers that: https://commerce-gpt.yotpo.com