On July 24, 2026, Google reported quarterly earnings that most SEO teams skimmed past. That's a mistake. Revenue grew 24% year-over-year. Google Cloud revenue grew 87%. And free cash flow went negative, on purpose, because Google is spending roughly $6 billion a quarter building server and data-center capacity for AI agents. SEO and AI-search consultant Marie Haynes flagged the real story buried in that call: Google isn't just building a better search engine anymore. It's building the infrastructure for agentic commerce, a world where an AI agent doesn't just read your site, it acts on it. That's not a messaging shift. That's a capital-spending shift, and capital spending is the one signal a company can't fake on an earnings call.
Google's $6 Billion Earnings Call Signal for Agentic Commerce
Earnings calls aren't normally content-bot material. Executives read prepared remarks, analysts ask about margins, and nobody says anything an SEO team can use. This one is different, because the company running the world's largest AI search product just told its investors exactly where it's putting its money, and the money points at agents, not answers. Marie Haynes, who has been tracking Google's AI moves closely at mariehaynes.com, published Marie Haynes's read on Google's earnings call on July 24, 2026, and the numbers she pulled out are the ones that matter here: 24% year-over-year revenue growth company-wide, 87% year-over-year growth in Google Cloud, and negative near-term cash flow tied to close to $6 billion in server and data-center capital investment for that quarter alone.
Read those three numbers together and a pattern falls out: a company doesn't spend $6 billion a quarter building infrastructure for a feature it plans to quietly retire. It spends that kind of money on the thing it believes is the next decade of its business. Haynes's thesis, and it's a sound one, is that Google is repositioning itself from a search engine you query into an AI-agent platform you delegate tasks to. That's a bigger claim than another product update. It's a statement, backed by capital, that the company expects users to increasingly ask an agent to do something rather than ask a search box to find something. Negative near-term cash flow, in this context, isn't a warning sign for investors watching Google's AI ambitions. It's the cost of laying track ahead of demand the company is confident is coming, and enterprise sites that wait for that demand to arrive before preparing for it will be preparing during the surge instead of ahead of it.
What Agent-Actionable Actually Means for Agentic Commerce
'Agent-actionable' sounds like a buzzword until you unpack what it requires. A page that reads well to an AI agent and a page an AI agent can complete a task on are two different engineering problems, and most enterprise sites have only solved the first one. Readable-for-AI means your content is well-structured, crawlable, and summarizable, the baseline GEO work of clean headings, schema, and extractable facts. Actionable-by-agent means an agent arriving at your site can actually finish something: check real-time pricing, add a specific configuration to a cart, confirm availability, complete a form, or hand back a structured answer it can act on without a human reading the page first. Turning a page into an AI agent website that can finish a task is different engineering work than turning it into a well-cited source that merely summarizes well. Agentic commerce lives in that second category, and right now, almost nobody's site does.
| SITE CHARACTERISTIC | READABLE FOR AI | ACTIONABLE BY AI AGENTS |
|---|---|---|
| Pricing | Listed in prose, sometimes behind a 'contact sales' page | Returned as structured, current data an agent can quote and compare |
| Product or plan selection | Described across marketing copy | Selectable via structured options an agent can pass to checkout |
| Forms and checkout | Built for a human with a mouse | Exposes steps an agent can complete without a rendering-dependent UI |
| Inventory and availability | Static or delayed updates | Live, queryable, and consistent with what a human sees |
| Machine access | robots.txt allows crawling for summarization | APIs or agent-readable endpoints that go beyond simple crawl access |
That's Something Inc.'s own framework for sorting site readiness, and most enterprise sites clear the left column and stall on the right one. We've published research on exactly this gap. In our look at why AI agents can't find your price, we found agents routinely fail the simplest commercial question a site can be asked, not because the price is hidden, but because it isn't structured anywhere an agent can parse it. And in our four-dimension framework for enterprise AI-agent readiness, we mapped what it actually takes for an agent to complete a buying task end to end; most enterprise sites clear one or two of the four dimensions and stall on the rest. Google spending $6 billion to make agents faster and more capable doesn't close that gap. It widens it, because the agents arriving at your site next quarter will be better at trying and no more forgiving about failing.
Why Agentic Commerce Raises the Stakes on Site Structure
Here's the part that should change budget conversations this quarter, not next year. If Google is building the rails for agents to browse, compare, and transact on a user's behalf, then AI search visibility stops being only about whether your brand gets named in a summary. It becomes about whether the agent that named you can then go complete the task it was sent to do. An agent that finds your brand but can't get a straight price, can't confirm a plan fits a use case, and can't complete a next step will do exactly what a frustrated human does: leave, and quote a competitor instead. The visibility work and the completion work are now the same budget line, whether your reporting treats them that way or not. Enterprise teams that keep those two line items separate on a roadmap are going to keep shipping half of what agentic commerce actually requires, and wondering why citation gains never show up as pipeline.
This is also where AI Mode's shift toward in-conversation actions matters. Google has already started building app actions directly into AI Mode results, and our breakdown of AI Mode's agentic commerce shift covers what that means for GEO strategy specifically: the surface where your brand gets discovered and the surface where the transaction happens are collapsing into one interface. An enterprise site built only to be summarized, not transacted with, is building for last year's version of that surface.
The Next Investment After Citation-Readiness
None of this replaces the work GEO practitioners have been doing. Citation-readiness, clean structure, extractable facts, demonstrated authority, machine access via robots.txt and sitemaps, is still the floor. An agent that can't parse your page in the first place never gets far enough to attempt a task on it. But the floor stopped being the ceiling the moment Google put $6 billion behind agents instead of another summarization feature. Agent-actionable design, structured pricing, completable forms, live inventory, transaction-ready data, is the next line item, and it sits on top of the citation-readiness work, not instead of it.
Do this next: pull your last quarter's GEO roadmap and add one column to it, can an agent complete a transaction here, not just read about one. Score your top ten commercial pages against that column this week. If the honest answer is no across the board, that's not a failing grade, it's the actual starting line for agentic commerce readiness, and Google just told you, with $6 billion, that the clock on it already started. Our generative engine optimization work is built around closing exactly this gap, from citation-readiness through to agent-actionable structure, before your competitors close it first.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Tyler leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.