One of the AI concepts currently reshaping the digital marketing industry is agentic commerce. It’s similar to the shift we’ve lived through in search over the past four years, where AI has reduced the friction needed to find things.
Search used to mean digging through 10 blue links to assemble your own answer. Now you get the answer handed to you.
Agentic commerce does that to buying. Instead of the many steps of “search, find, click, create an account, buy,” you tell an agent what you want, and it handles it for you.
The friction is obvious when discovering a new brand: You find a product you like, but are immediately blocked by a forced account creation or an unwanted “20% off” email prompt. This behavior drives customers away; according to Baymard Institute, forced account creation is a top reason for cart abandonment, cited by roughly 19% of shoppers who walk away.
Stack the cookie banners, pop-ups, and manufactured urgency on top, and you have a buying experience built to extract an email and an impulse, not to get someone the thing they came for.
That’s what makes handing the job to an agent so appealing. “It knows who you are and what you like. It knows your card and your address. It can just get you the thing you want.”
The technology is ready, the platforms are behind it, and consumers want it. That’s why this is moving fast and worth getting ahead of.
The question for those of us who manage Google Ads is: what should we change this year to keep showing up? Here’s what’s happening, and a four-part checklist you can start this week.
The Shrinking Ad Landscape
This is the part that freaks us all out in digital advertising. AI is reducing ad impressions, our opportunities to connect with prospective buyers.
AI Mode and AI Overviews have eaten into the space where ads used to live. There are simply fewer impressions to go around. We saw this in our own data, where we noticed an 11% decline in impressions YoY.
Google is running ads inside AI Mode and rolling out formats like Direct Offers. But the canvas is smaller, and the bar is higher.
The audience on the other side of that canvas is changing, too. Microsoft frames it as three eras of the web at once: “help me find it,” “help me choose,” and “do it for me.”
That last one is growing fast. Microsoft cites data showing automated traffic growing roughly eight times faster than human traffic.
Agents don’t scroll, and they don’t respond to a clever headline. They evaluate, select, and act.
Retail media strategist Roger Dunn has a useful name for the result: the “shortlist economy.”
When a shopper turns to ChatGPT, Gemini, or Copilot for a recommendation, they get three to five options back. That shortlist becomes the entire consideration set.
This is already mainstream behavior. A December 2025 Semrush survey of 1,030 U.S. shoppers found that 43% had discovered a new brand through AI, and 47% said they notice AI-mentioned brands often or very often.
If you’re not on that shortlist, you don’t get to make your case. A strong brand campaign counts for nothing if the agent never surfaces you.
So, the impression squeeze isn’t really about fewer ad slots. It’s a shortlist of five, and the work is making sure you’re on it.
You can now measure whether you’re making that cut. Google’s AI performance insights in Merchant Center show your share of voice on AI surfaces against similar brands – the closest thing we have to a rank report for the shortlist economy.
When the screen had ten organic results and four ads up top, a “pretty good” match could still earn a click. That cushion is gone.
By the time your ad shows up now, it had better be the exact product the user wants – and the agent had better know they can transact on it right now. There’s no room left for “close enough.”
Confidence: The Third Pillar Of The Auction
For two decades, we’ve viewed the auction as a balance between bid and quality. But now agents are representing their humans, acting on their behalf only when they are certain of the outcome. In this new landscape, confidence becomes an equal participant alongside your bid and your quality score.
Think about what an agent is actually checking. Is this product available? Does it match what my human asked for? Is it at the price they said they’d pay?
If all three answers are clean and verifiable, it buys. If anyone is uncertain, it hedges.
That’s the part advertisers need to be concerned with.
If your product is the better fit but the agent isn’t sure it can transact at the price it sees, it will recommend a competitor it can transact on. Not because your product was worse, but because the other one was a known quantity.
The agent’s job is to deliver a successful purchase. Uncertainty is the enemy of that job.
So, your real competitive lever isn’t only a higher bid. It’s higher confidence.
The cleaner and more trustworthy your data, the more often an agent picks you – sometimes even over a rival with the marginally better product.
And the lever really is data, not position. In Semrush’s survey, only 21% of shoppers said a brand stood out because it appeared earlier in an AI answer – whereas 43% pointed to a clearer, more detailed description and 39% emphasized price and value context. Poor product data used to mean lower conversion. In the agentic world, it means you’re never in consideration at all.
Rethinking Promos And Offers
Here’s where a lot of conversion rate optimization playbooks will soon break.
A huge amount of retail pricing is psychologically aimed at humans. Anchor prices. Three options sized to nudge you to the middle. The Prime Day classic where the price sneakily goes up the week before, so the “37% off” looks bigger.
Those tricks work because we’re human and we’re susceptible to them.
The agent is not susceptible. It looks at specs and the real number, not the story around the number.
Whether something is 43% off, 35% off, or full price matters far less when the buyer is software with a price ceiling. It just compares the actual cost to what it’s human authorized.
This is where it gets interesting, because the offer logic starts to look like a stock trade.
Google showed an example where you tell Gemini to watch for a fragrance and buy it the moment it drops below $15. That’s a limit order. You set the price, and when the market hits it, the transaction fires in the background.
So discounting doesn’t disappear, but its job changes. A discount that interrupts a human mid-scroll still works as attention bait. A discount aimed at an agent only matters if it clears the price the agent was told to hit.
But your brand reputation still matters too. Most shoppers still verify an AI’s shortlist before buying: Semrush found 86% double-check AI recommendations at least sometimes, validating on Google (68%) and brand websites (48%) before they commit.
The catch: that verification is a confirmation exercise, not an open search. They’re checking the brands the AI already named, and the one they trust gets the click.
So clean data gets you onto the shortlist, and brand equity closes the sale. You need both.
The PPC Manager’s Agentic Checklist
Enough theory. If you manage accounts, here are four things to start now.
The good news: most of this is unglamorous data hygiene, which means it’s well within your control.
1. Unblock The Bots
For years, the default instinct was to keep bots out. A bot was probably a competitor scraping your site, so you blocked it.
That instinct is now actively costing you sales.
Today’s shopping agents legitimately represent real buyers trying to purchase from you. Blocking them in 2026 is roughly what blocking Googlebot was in 2010: you disappear from the channel that’s becoming your next acquisition surface.
Here’s how to check in ten minutes:
- Pull up
yourdomain.com/robots.txtin a browser. Look forDisallowrules denying any AI user-agents. - Make sure the live, user-triggered agents are allowed:
OAI-SearchBotandChatGPT-User(OpenAI),PerplexityBot,Google-Extended, and Anthropic’sClaude-Web. - Know the difference between a training crawler and a live agent.
ClaudeBottrains models;Claude-Webfetches pages to answer a live request. Many sites block both with one blunt rule and shut out the shopper-facing one. - Don’t stop at robots.txt. Check your WAF, CDN, or Cloudflare bot rules and any rate limiting — they can block agents even when robots.txt says “come on in.”
On Shopify, much of this is handled for you. Eligible products are syndicated to AI channels through Shopify Catalog by default, and there’s a hub in Shopify Admin to toggle channels and see what each one drives.
2. Obsess Over Data Cleanliness
Agents don’t reward storytelling. They reward data they can trust.
We’ve had Merchant Center feeds for years. They just became far more important.
An agent needs to know an item is actually available, that the price it sees is the price it’ll pay, and what shipping will cost and when it arrives. Any gap between your feed, your site, and reality is a reason to hedge and go elsewhere.
Practically, audit your feed for stale availability and price mismatches. Turn on automated item updates so Merchant Center reconciles price and availability from your site in real time.
Then make sure your on-page structured data matches your feed. When a bot reads the page instead of the feed, it should see the same numbers.
3. Expand Your Product Attributes
I still see brands pour all their energy into the product title and stop there. In an agent-driven world, that’s a miss.
Agents love data more than imagery or emotional copy. So flesh out descriptions and attributes: materials, sizing, compatibility, use cases, the answers to the questions a shopper would actually ask.
Agent queries are specific, which is why this pays off. Semrush found 52% of shoppers state their constraints upfront – a budget, a required feature, a compatibility need — so the listing that answers those constraints is the one that gets surfaced.
Google added conversational attributes in Merchant Center for exactly this – answers to common product questions, compatible accessories, and substitutes, built for how people and their agents really query.
The brands with the richest, most accurate attributes are the ones agents surface, because the agent has more to verify against. Title-only listings are invisible to a buyer like an agent who reads everything.
4. Embrace The Protocols
The plumbing is being standardized right now, and there are two distinct technologies worth knowing.
Agentic Commerce Protocol (ACP)
Co-developed by OpenAI and Stripe, ACP is designed for AI-agent-driven checkout, specifically within conversational interfaces like ChatGPT. Think of it as a chat-to-buy protocol. It focuses on product discovery within a conversation, cart creation, and delegated payments, where the AI acts on your behalf.
Currently, ACP has tilted toward deep partnerships with large retailers; while big names get full in-chat checkout, smaller merchants often get product discovery with a link back to their own site.
Universal Commerce Protocol (UCP)
UCP is an open commerce interoperability standard supported by a broader ecosystem including Google, Shopify, Visa, Mastercard, and Stripe. Think of UCP as a search/discovery-to-buy protocol that aims to work across many different AI agents and platforms. It focuses on identity linking, order tracking, and payment token exchange across the web.
UCP is built for absolute scale because it rides on platforms that already host millions of catalogs. When a platform like Shopify supports the protocol natively, every merchant on it gains agentic capability through simple configuration rather than expensive custom engineering.
That’s good news if you’re not an enterprise brand. You sell through Target, Walmart, Amazon, or Shopify, and those platforms do the protocol-heavy lifting. Your job is to match their data feed formats exactly and ensure you are transactable through at least one protocol rather than betting everything on one ecosystem. You don’t have to build your own shopping agent; you just need to ensure that whatever agent shows up at your store knows precisely what it can do for its user.
And you don’t lose the customer relationship in the bargain. Across these protocols, the retailer stays the merchant of record – you still own the transaction, the customer data, and the returns, even when discovery and checkout happen on someone else’s surface.
There are two agents worth switching on: Google’s Business Agent – and Microsoft Clarity’s equivalent Brand Agents – let you put a brand-voice assistant in front of shoppers to answer product questions right in Search, and you activate and customize it from Merchant Center. That’s a setting rather than a six-month build: the difference between engineering your own agent and making sure the one representing you speaks in your voice and knows your catalog.
Humans Still Drive
None of this changes the fundamentals of our job. It just changes the details yet again.
I’ve said for years that humans plus machines beat machines alone. Agentic commerce is a clean example of why.
The agent handles the search, the comparison, and the checkout. It does not decide what your brand stands for, what your margins can absorb, or whether your feed tells the truth. That’s still us.
The advertisers who win the next year won’t be the ones with the cleverest promo psychology. They’ll be the ones whose data is so clean and trustworthy that an agent picks them without hesitating.
Clean the feed. Open the door to the bots. Tell the truth about price and availability. Then let the machine do what it’s good at.
More Resources:
- Agentic Commerce And The New Rules Of Google Ads
- Selling To AI: The Complete Guide To Agentic Commerce
- Agentic Commerce Optimization: A Technical Guide To Prepare For Google’s UCP
Featured Image: Summit Art Creations/Shutterstock
