Google Built a Protocol for AI Agents to Buy B2B Software. Your Brand Wasn't in the Room.
Google's Universal Commerce Protocol lets AI agents execute B2B purchases without human involvement. If your brand isn't already in the models, no salesperson closes that deal.
In January, at the National Retail Federation conference, Google announced the Universal Commerce Protocol — an open standard that lets AI agents execute purchases across multiple vendors, handle payments, manage post-sale logistics, and complete B2B buying cycles without a human touching the keyboard. Shopify, Etsy, Wayfair, Target, and Walmart backed it at launch. It stacks on top of Google's existing Agent Payments Protocol and Agent2Agent standard.
The headline got some coverage. The implication mostly didn't.
An AI agent executing a purchase doesn't research vendors the way a human does. It doesn't Google your company, scan your homepage, or read your case studies. It queries its training data and retrieval layer, assembles a shortlist from what it already knows about trusted players in your category, and executes. The human who delegated the task never sees the alternatives that didn't make it.
Your brand either lives in that shortlist or it doesn't. There is no "we'll get them next time."
The shortlist already closed
Most B2B marketing teams are still optimizing for the moment a buyer types a query into a search bar. That buyer is increasingly not the one making the call.
Forrester's 2026 State of Business Buying research, drawn from nearly 18,000 global business buyers, found that 94% now use AI during their purchasing process. The more pointed finding: buyers increasingly lean on AI for speed and breadth of initial research, then validate the AI's shortlist against trusted external sources. They are not building the shortlist themselves and then checking with AI. They are taking the AI's shortlist and looking for reasons to confirm it. As Christian Lehman documented in his breakdown of how B2B buying decisions are made before companies know they're being evaluated, the shortlist is invisible to the brands that didn't make it.
The qualification round is happening before your SDR gets a call. And now, for a growing class of purchases, the qualification round and the purchase itself are handled without a human in the loop. AI agents don't browse — they recall. The shortlist is assembled from memory, not from a fresh search.
Harvard Business Review's March 2026 piece on agentic AI brand strategy put it plainly: AI agents interact autonomously on both sides of transactions, with human preferences "pre-filtered using algorithms." The brands whose positioning survives that filter make the shortlist. The ones who don't, don't get notified they were dropped.
What the agent is actually reading
When an AI agent evaluates vendors, it's not reading your marketing copy. It's reading what third parties have said about you.
Ahrefs published data on 75,000 brands showing that brand web mentions — the direct output of editorial coverage — correlate 0.664 with AI visibility, while backlinks correlate 0.218. That's a 3x signal difference. The thing that drove SEO for a decade is one-third as useful as the thing that PR programs produce.
AI engines are built on the same trust hierarchy that shaped human brand perception for decades: authoritative publications, independent editorial coverage, named citations from credible sources. Earned authority is the term for this layer — and it's the one most B2B brands have underbuilt. A Forbes article about your company isn't just a traffic source. It is the signal the model uses to determine whether your company is credible enough to recommend.
When an AI agent executes a purchase under the UCP, it is drawing on that exact layer. Not your blog. Not your paid media. What Forbes, TechCrunch, and the trade press have said about you over the past two years.
Muck Rack's Generative Pulse report found that over 85% of non-paid AI citations come from earned media placements. There's no shortcut here. The models already decided what counts.
The gap most companies haven't noticed yet
Here's the math that makes this uncomfortable.
| Channel | AI citation rate | Source |
|---|---|---|
| Earned editorial coverage | 85%+ of non-paid AI citations | Muck Rack Generative Pulse, Q4 2025 |
| Brand web mentions | 0.664 correlation with AI visibility | Ahrefs, 75K brands, May 2025 |
| Backlinks (traditional SEO signal) | 0.218 correlation with AI visibility | Ahrefs, 75K brands, May 2025 |
| Wire distribution (PRNewswire, etc.) | 0.21% of 4M citations analyzed | BuzzStream, 2025 |
| Syndicated press releases (Yahoo, MSN) | 0.04% of 4M citations analyzed | BuzzStream, 2025 |
| SEO top-10 overlap with AI Mode citations | 12% | Moz, 40K queries, 2026 |
Most B2B companies are spending brand budget on the bottom rows of that table. The traditional PR playbook — wire a release, hope for pickup — is statistically invisible to AI engines. Strong Google rankings and the AI citation lists are almost entirely separate. They're built from different signals, measured by different models.
The brands that make it into AI agent shortlists have editorial coverage in publications AI engines treat as authoritative. They've built that coverage over time, not in a sprint before a product launch.
And this plays out at the founder level too. AI is already eroding the human judgment layer that would have caught a bad vendor recommendation — which means the model's initial shortlist carries more weight than any single decision-maker's override.
The window you're inside right now
A compounding gap is opening between companies that have established editorial authority in AI-indexed publications and companies that haven't started.
Every month that passes, the training data and retrieval signals accumulate. The brands consistently covered in TechCrunch, Forbes, the Financial Times, and the relevant vertical press become the default answers for their category queries. The brands not in that coverage don't just miss individual recommendations — they get systematically excluded from a growing share of the buying process.
This is not theoretical. Perplexity's enterprise Computer agent, which crossed 100 enterprise customers in a single weekend after launch, is already executing research and workflow synthesis on behalf of its users. Google's agent protocol stack — UCP, A2A, AP2 — is already live with major commerce partners. What has been true in research-phase AI queries is now true in transaction-phase AI execution.
The companies building editorial authority now are establishing the signal that will feed AI agent shortlists for years. The companies waiting to see how this plays out will find the shortlist was already finalized.
That's what Machine Relations describes as the central shift: AI engines cite the same publications that shaped human brand credibility for decades. Earned media placements in those publications are now the foundation of AI citation authority. The mechanism never changed. The reader did.
The share of citation your brand holds today — how often you appear in AI visibility queries across the platforms buyers use — is the clearest leading indicator of whether you'll be on the AI agent's shortlist tomorrow. It's measurable. And for most B2B companies, it's lower than they think.
If you want to see where your brand stands in AI answers before the next agentic purchase protocol goes live, run the visibility audit.
Related Reading
- AI Visibility for HR Tech Companies: How People Platforms Get Cited in Enterprise AI Search
- AI Visibility for Consumer Brands: The 2026 Earned Media Playbook
Sources: Forrester State of Business Buying 2026, Google Universal Commerce Protocol announcement, TechCrunch, January 2026, Harvard Business Review, Preparing Your Brand for Agentic AI, March 2026, Muck Rack Generative Pulse Q4 2025, Ahrefs brand visibility study 75,000 brands May 2025, BuzzStream AI citation analysis 4M data points, Moz 2026 AI Mode analysis 40,000 queries