A couple of years from now, you tell your phone to organize a trip. It knows your budget, your schedule, and what you consider an acceptable hotel—not just for you, but also for your partner and your eldest daughter, who will make your life miserable if there’s no vibrant nightlife at the destination. The agent compares transportation, verifies live availability, filters out unfavorable conditions, and presents you with a synthesized proposal. You review it, ask a quick question, grant permission, and it books.
Multiple companies participated. All of them performed work. Some got paid.
How many did you actually choose?
You might remember the hotel because you slept there, the last time your family was together under one roof. But you almost certainly will not know which platform sourced the room, which system compared the rates, or which provider managed a specific leg of the journey. And when you need to repeat the process, you will talk to your agent again.
I have spent a long time thinking about this: it is one of the deepest business and societal transformations AI will usher in. We are constantly obsessed with how many tools are launching and which one performs each task marginally better. I find myself asking a very different question: how many of those tools will we actually bother opening? Because a business can preserve its operational activity while losing something infinitely harder to regain: its place in the customer’s conscious decision.
The Discussion Began with Google. The Problem Is Far Bigger
This reflection took concrete shape recently while reading Google’s official guide on optimizing for generative AI features in Search. The documentation argues that SEO fundamentals remain as relevant as ever, and that certain tactics portrayed as mandatory for AI are unnecessary for appearing on Google. It also clarifies that llms.txt does not improve rankings—though it may be maintained for other systems that consume it. (Google Search Central, “Optimizing your website for generative AI features on Google Search”).
So far, standard technical instructions for a specific product. What interests me is the business interpretation we choose to draw from them.
For years, we have engineered budgets, content calendars, internal workflows, and outsized expectations around visibility on Google. It is understandable: where demand gathers, value concentrates. The problem emerges when we mistake a single platform’s access rules for a complete explanation of how our entire market works. The question this unspoken assumption provokes goes far beyond how to secure a fleeting mention in a generative response: What happens if users simply stop going to the place we are optimizing for? We can keep polishing the shop window while the foot traffic moves to an entirely different street.
The SEO War Is Also a Battle to Define What Matters
SEO, AEO, GEO. Rankings, direct answers, AI recommendations. Acronyms are useful when they help isolate technical problems and measure clear outcomes. They are also exceedingly useful for defending agency retainers, justifying new software subscriptions, and claiming commercial categories before a competitor does.
Conflicting incentives surround us on every side. A search engine explaining the digital universe through the prism of its own architecture. An agency defending the indispensability of its craft. A software vendor requiring that the specific metric it tracks remains an urgent source of executive anxiety. And an enterprise that simply wants to know what service to buy, what tool to deploy, and what checklist to execute to avoid being left behind.
I am not claiming everyone is being deceitful. I am stating something far more uncomfortable: much of what is sold today as mandatory remains unproven, and those selling it have no urgent need to prove it as long as the fear of falling behind keeps paying the invoices.
We repeat this cycle tirelessly. A new technology emerges, uncertainty opens up, and immediately a catalog of non-negotiable obligations appears. Occasionally, those obligations solve real operational bottlenecks. Far more often, they merely convert corporate anxiety into vendor billing.
There is serious academic research on this topic. The study “GEO: Generative Engine Optimization”, accepted at KDD 2024, examined targeted content interventions and observed visibility improvements of up to 40% under its test conditions, with wide variations across domains. That percentage belongs to specific experimental setups; it is neither a universal guarantee for any website nor an assurance of revenue. We can respect that research while simultaneously challenging the commercial peddling of universal formulas.
What Happens When Everyone Is Perfectly Optimized?
And right here lies the heart of the dilemma.
Imagine ten companies competing for the exact same pool of clients. All ten utilize state-of-the-art models. All ten identify high-intent search queries, eliminate technical errors, publish well-structured content, and continuously refresh their pages. Their autonomous agents monitor competitors around the clock and counter-move within minutes.
None of them can sustain an enduring competitive edge built on a task that is easily replicable. The optimization still holds baseline utility—a clear website will always outperform an incoherent one. But its capacity to truly differentiate collapses when everyone achieves a comparable standard of excellence. Here emerges a distinction routinely buried in productivity debates: doing something cheaper or faster does not guarantee capturing more economic value than your peers.
If we all drive the marginal cost of content production to zero, the market is under no obligation to multiply its attention span in equal measure. If all ten of us can publish a thousand optimized landing pages, buyers still operate with strictly finite hours in a day. If everyone is mathematically optimized to be recommended, the selecting system is forced to make a ruthless cut.
I am not suggesting that everyone will have identical data, models, or capital. In fact, those disparities will become profoundly decisive. My core thesis is that widespread automation will steadily erode superficial operational advantages, shifting the competitive battleground toward assets that cannot be synthesized: Proprietary Offer. Authentic Reputation. Owned Distribution. Private Data. Hard Trust. Verifiable Fulfillment Capacity. And direct access to the intermediary that decides.
The Obsession with Metrics Can Outlive Its Utility
There is a insidious form of organizational addiction that takes root when we stop asking what a metric actually achieves. Impressions climb, and we celebrate. AI mentions rise, and we celebrate. We appear in a synthesized answer, and someone takes a screenshot for LinkedIn. Has anyone paused to ask what fundamentally changed in the business?
An AI citation can project authority without ever generating a visit. A visit can land without leaving behind a shred of enduring value. A transaction can materialize without the customer ever registering our brand name. The hotel from our opening scenario will bill that night, and inside its dashboard the reservation will be logged as an undeniable triumph.
In an agent-dominated ecosystem, we may well witness fewer web sessions and higher absolute transactions. Yet we may simultaneously experience more transactions accompanied by tighter margins, because another entity owns and taxes the distribution layer. We could even report revenue growth while our existential vulnerability to a single intermediary compounds exponentially. I suspect much of today’s euphoric rush to measure «new visibility» will eventually collide with an ancient truth: measuring what is easily accessible is far simpler than measuring what is genuinely important.
The real strategic inquiry is determining whether we are cultivating an enduring relationship or merely being utilized as swappable inventory. Both can generate invoices today. Five years from now, their business realities will bear no resemblance to each other.
The More Tools That Emerge, the Fewer Interfaces We May Need
Today’s chaotic tool proliferation seems to contradict the idea of disappearing interfaces. I see a direct causal relationship between the two.
First, discrete human capabilities are digitized. Next, APIs connect them. Finally, someone orchestrates the entire chain behind a unified interface that spares the user from hopping between fragmented dashboards. Today, an operator opens an analytics tab, interprets the dataset, drafts a strategic proposal in a document, and logs into a campaign tool to execute it. Each step requires its own software subscription and its own cognitive overhead.
Meta introduced its Creator Assistant in June 2026, translating raw account performance metrics into natural conversational recommendations without requiring creators to traverse analytics menus. It is a quiet yet revealing milestone: foundational platforms are actively absorbing the cognitive burden of interpretation directly into their core experiences.
The decisive leap ahead is full-journey orchestration: understanding the strategic intent, querying background systems, proposing a concrete intervention, and executing it upon human authorization. Once that operates with bulletproof reliability, opening five disparate SaaS tabs to execute a routine workflow will feel as obsolete as manually transcribing database records onto paper. Software applications may well continue to exist as back-end APIs. Our daily habit of visiting them as interactive websites is what will quietly evaporate.
My Prediction on Google Comes with a Date, a Threshold, and a Condition
I predict that before September 2029, Google will cease to be the primary default interface for a decisive share of commercial search queries.
I am not using the word “might.” I am stating it unequivocally, and I want the scale of this wager to be crystal clear. Google currently commands roughly 91% of global search market share. I am betting against the single most impenetrable digital monopoly of the past two decades, and I am putting a calendar stamp on it.
I am specifically referring to the commercial journey that initiates with an intent and concludes after navigating through dozens of blue links, price comparison tabs, and decision hurdles. If an autonomous agent can handle the bulk of that cognitive legwork while strictly honoring our individual constraints, the traditional search engine forfeits its role as a mandatory gateway.
This does not mean Google’s web index, cloud infrastructure, or corporate enterprise will vanish. Google can readily power the back-end infrastructure of countless third-party experiences, monetize computational capabilities, pioneer new revenue streams, and defend critical assets. But being an infrastructure utility is vastly different from controlling the consumer front door.
The explicit condition of my prediction is that users must delegate commercial transactions with routine frequency, and that a decisive portion of that relationship must materialize outside of Google’s proprietary perimeter. If that fails to happen, my timeline is wrong.
In September 2029, I will write about this again and openly assess whether I was right or wrong. Setting a firm date forces honest intellectual debate. Doing otherwise is merely announcing a company’s death and continually postponing the funeral every time the patient walks across the room.
Google Is Actively Racing to Occupy That Very Future
The strongest counter-argument to my thesis is hiding in plain sight: Google itself could become the default agent. On January 11, 2026, at the NRF Big Show, Google unveiled the Universal Commerce Protocol (UCP), an open technical standard co-developed alongside Shopify, Etsy, Wayfair, Target, and Walmart designed to allow autonomous agents to discover merchant inventory, negotiate terms, and complete cross-platform checkout. This is no theoretical whitepaper: it operates natively inside AI Mode and Gemini, and was expanded in May 2026 to support multi-merchant carts and deferred financing. This decisively shatters the naive narrative that Google is merely observing its own displacement from the sidelines.
Yet there is a parallel fact rarely mentioned in the same breath: OpenAI and Stripe had launched their own agentic commerce protocol in September 2025, three months prior. Major global retailers are already supporting both frameworks concurrently. This is the structural reality that must be understood: this is not exclusively about Google. It is about the fundamental standardization of the decision-making layer, and multiple titans are aggressively vying to govern it.
My assessment is that Google possesses the engineering prowess to secure a prime seat in the architecture that replaces traditional search. But technological capability and economic model continuity are entirely separate crises. Google must figure out how to extract massive commercial value from an agentic flow without eroding the consumer trust required for a user to delegate decisions in the first place.
The strategic question is no longer where to place a sponsored search ad. It becomes: what degree of commercial bias can be covertly injected into an autonomous decision that the user believes was executed purely in their personal interest? That structural conflict haunts every agent provider, not just Google.
The Consumer’s Agent and the Merchant’s Agent Want Different Things
Suppose my personal agent is dispatched to book a hotel. As the consumer, I demand a luminous room, transparent cancellation policies, and the optimal equilibrium between cost and comfort. The hotel wants to preserve room margins. The intermediary booking platform demands its percentage rake. The agent platform provider must monetize its compute infrastructure.
None of those underlying economic tensions vanish simply because the conversational interface feels smooth and human-like. What happens when those incentives collide?
An agent can seamlessly compare ten thousand options while quietly restricting its final selection to a pre-approved commercial catalog. It can deliver a beautifully articulated explanation justifying its recommendation while omitting superior alternatives it was never programmed to consider. It can strictly respect my financial ceiling without uncovering the absolute best value on the open market. Interface elegance is not evidence of algorithmic independence. User trust in autonomous agents will ultimately hinge on governance: who funds the model, which suppliers are indexed, what revenue-sharing agreements dictate the preference weights, and which decisions can be audited by the user.
Automated marketplace competition is not entirely novel. Behind programmatic ad placement lies an elaborate economic architecture of real-time auctions and mechanism design. With autonomous agents, however, that machinery is pushed infinitely further away from consumer visibility.
If We Stop Fighting Manually, the Battle Simply Shifts to Code
Exhausted from chasing every algorithmic update, every margin squeeze, and every volatile lead, business owners will inevitably hand daily operational execution over to AI. One specialized agent will dynamically calibrate ad bids. Another will reprice inventory. Another will field inbound customer queries and negotiate within predetermined guardrails. This is already unfolding.
Executive leadership will step away from hundreds of micro-decisions. That will unlock unprecedented operational bandwidth. But it does not eliminate the brutal underlying struggle for customer attention, finite resources, and gross margin. We could easily find ourselves paying proprietary AI systems to battle around the clock against the proprietary AI systems financed by our competitors. Or, far more likely, we will all end up paying subscription fees to the exact same cloud provider.
From the outside, business will look frictionless. Beneath the surface, the contest will escalate. If detecting market shifts and responding to demand carries near-zero cost, competitive velocity explodes: every unique advantage decays faster, and systemic behavior becomes far harder to decipher. The role of executive leadership shifts fundamentally: instead of micromanaging operational maneuvers, leadership must establish non-negotiable boundaries, strategic intent, and audit mechanisms. Delegating «maximize revenue» to an agent without specifying risk tolerance, fulfillment capacity, brand equity guardrails, or client quality is merely an exceptionally automated way to chase catastrophic outcomes.
A Sold Room Can Conceal a Lost Customer Relationship
Consider an independent boutique hotel in Ciudad Real. A traveler’s AI agent scans local availability, confirms secure private parking, verifies room dimensions, and processes the booking. The local business welcomes the guest, provides great hospitality, and deposits the payment.
The economic transaction was real. The service delivered was genuine. Yet the customer attributes the entire seamless experience to their personal AI assistant. When they travel six months later, they repeat the exact same prompt to their phone. They have no recollection of which booking platform was utilized, nor do they feel any organic compulsion to book directly with the hotel. The independent lodging must battle to be algorithmically selected all over again from scratch.
Vulnerability to digital aggregators is already familiar to hoteliers and merchants. My contention is that autonomous agents will aggressively extend that dynamic across virtually every industry, compressing the fleeting moments where a brand has the opportunity to introduce itself, articulate its unique difference, and build durable emotional preference.
There is a strategic way forward. An extraordinary physical experience can compel a consumer to issue explicit instructions to their software: “Book the exact same hotel we stayed at last time” is a profoundly different prompt than “Find me a hotel.” That is precisely where the modern strategic function of Brand resides: compelling the human user to inject your company’s name as an immutable constraint that the AI agent is forbidden to override. As my esteemed colleague Matt Murphy would put it: “That’s a win!”
Returning to Fundamentals Will Not Be as Simple as It Sounds
If easily replicable growth hacks cease to provide defensible moats and algorithmic systems become better at auditing actual offerings, economic value may well concentrate back into bedrock business fundamentals: product quality, contractual fulfillment, operational availability, fair terms, and rigorous problem resolution. The product. The service. One’s word.
Yet that return to fundamentals is not automatic. An autonomous agent requires structured evidence to evaluate those virtues. Where does it pull that data? Who verifies that a provider consistently honors its word? How does a brilliant newcomer establish visibility without an extensive historic audit trail? If the agent leans exclusively on legacy authority signals, it entrenches entrenched incumbents. If it relies on third-party verification platforms, those platforms become rent-seeking gatekeepers. If it ingests unverified marketing claims, it reopens the floodgates to manipulation.
A business could provide an extraordinary service and still face the technical imperative of proving that reality across the specific registries, data formats, and protocols that the selecting agent trusts. The analog world will not return pristine; it will be mediated by whoever certifies that a vendor deserves consideration.
This is why I reject the defeatist claim that autonomous intermediation will turn every business into an undifferentiated commodity. But it will punish businesses that were already interchangeable. A company whose sole differentiation was generic copywriting will struggle to justify an agent’s recommendation. A company anchored to a distinctive, tangible experience possesses something far more durable.
This fundamentally transforms the purpose of content creation. Publishing blog posts purely to satisfy an editorial calendar is an exercise in futility. Conducting rigorous original research, explaining complex technical trade-offs, demonstrating verified field experience, and staking out an unapologetic viewpoint creates the intellectual footprint that makes humans remember who is behind the brand. Combine that with the ingenuity of real entrepreneurs—which is formidable—and many will carve out dominant positions in this reality. It does not guarantee an algorithmic citation, but it builds human brand preference that precedes and overrides the algorithm.
Infrastructure Will Wield Power. The Selector Will Wield Leverage
Immense commercial fortunes will be accumulated in the infrastructure tier: compute, energy, high-bandwidth connectivity, distributed storage, payment rails, and execution runtimes. But owning infrastructure does not automatically secure the highest margins. The entity that owns the direct, trusted relationship with the end user occupies the most defensible economic position because it understands commercial intent, stores years of contextual preferences, orchestrates vendor calls, and continuously learns from transaction outcomes.
If changing your primary AI agent requires re-educating a system on years of intimate personal context, switching costs skyrocket and convenience hardens into lock-in. We could easily arrive at a landscape with thousands of technically accessible API endpoints, yet with real-world consumer selection concentrated in a tiny handful of foundational agents. An open protocol does not guarantee an equitable distribution of demand. Making connectivity seamless solves one engineering problem; earning customer selection solves an entirely different commercial challenge. The front door can be open to all, while the selector inside chooses who gets summoned.
What Happens to Analytics, Search Console, and Daily Dashboards?
I do not expect the necessity of measurement to disappear. I expect the volume of manual human labor required to extract actionable answers to drop precipitously. A business owner will stop endlessly navigating reporting dashboards and simply ask: Why did gross margins compress this week? Which marketing channel brings high-retention enterprise accounts? Where are operational handoffs stalling? Behind the scenes, robust data architectures, clean definitions, and structured tracking remain essential. Verifying that the agent’s analytical explanation is mathematically correct becomes the real human job.
Certain software interfaces will see usage evaporate. Others will evolve into specialized cockpits for governance, anomaly detection, and exception handling. SaaS products will have to defend their value when staring at charts is no longer the default way humans interact with business data. The same fate awaits mobile apps: utility apps designed purely as transactional overhead are vastly more vulnerable than software built for immersive experiences, collaborative design, deep conversation, or creative exploration.
The Bottleneck Lies in Edge Cases and Who Carries Liability
A software demo concludes the moment a task executes successfully on stage. A real-world business must continue functioning when data feeds break, pricing APIs fail, duplicate transactions occur, or a premium client demands a bespoke exception. For enterprise commerce, operational reliability matters infinitely more than technological spectacle.
An AI agent that executes flawless reservations nine times out of ten but triggers a catastrophic commercial failure on the tenth demands too much human supervision to be economically viable. The architecture of Google’s UCP explicitly accounts for exception states requiring human intervention when automation exhausts its parameters.
My prediction depends on whether these edge cases become manageable and whether the net cost of auditing an agent remains lower than the human labor it replaces. If supervising an autonomous workflow becomes more burdensome than performing the task, the productivity promise collapses. Adoption will advance through discrete operational categories and measured autonomy thresholds. Searching for alternatives requires far less trust than committing corporate funds without confirmation.
Agencies and SaaS Companies Are Caught in the Same Dilemma
It would be easy and comfortable to write this essay as if the disruption threatened Google alone. I will state this in the most uncomfortable terms possible: I run a digital agency that sells services directly impacted by what I am questioning.
If a significant portion of our agency’s value proposition consists of manually shuttling data between disparate marketing dashboards, what happens when that integration is automated by default? If a SaaS product merely provides a pretty visual interface over commodity APIs, why would an enterprise continue paying for it? If a technical SEO audit merely enumerates syntax errors that an autonomous agent can diagnose and patch autonomously, what additional strategic decision does that audit unlock?
I do not hold packaged, convenient answers to those three questions. And I know my industry prefers not to utter them out loud.
Specialized technical work will undoubtedly persist; the real question is how much of it will remain a defensible differentiator versus a commoditized background utility. I see durable competitive moats in deep institutional context, proprietary data sets that cannot be scraped, tight integration with physical business operations, and ultimate accountability for business outcomes after a recommendation is executed. Engineering an operational system that gracefully handles real-world business failure demands vastly more depth than prompting a model to generate a polished answer. Businesses will always require strategic partners who know the difference between the two.
How I Will Know If My Thesis Is Unfolding
I will not measure my prediction by tracking the velocity of PR announcements or AI assistant rollouts. I will watch for verifiable behavioral shifts: consumers routinely delegating end-to-end purchasing, commercial transactions settling with zero human interface interaction, and merchants receiving revenue without visibility into the prior consideration funnel. I will monitor whether market dependence consolidates into a tiny oligopoly of consumer agents, whether brands receive fewer explicit organic choices, and whether vendors are coerced into accepting aggressive commercial terms just to remain discoverable.
Counter-signals matter just as much to me. If consumers use AI agents purely for research but consistently return to traditional web interfaces to execute final purchasing decisions, the disruption will be far milder. If Google successfully retains user loyalty across its proprietary apps, my timeline regarding its loss of gateway dominance will need revision. If human supervision of agents remains prohibitively expensive, the timeline will stretch significantly.
I maintain my forecast for rapid structural transformation. Precisely for that reason, I want to frame it with falsifiable clarity, rather than insulating it behind an amorphous, indefinite future.
In the End, Every Road Converges on Choice
The war over SEO, AEO, and GEO acronyms. The corporate obsession with vanity metrics. The industrial automation of web copy. The explosion of SaaS interfaces. Autonomous agents negotiating with other autonomous agents. Every single one of these threads converges on a single question: Who actually decides?
Perhaps we will eventually cease battling for every fractional ranking position, every marginal click, and every technical tweak. Perhaps that endless sprint loses its meaning once those technical capabilities become ubiquitous background utilities. Yet someone will still govern the allocation of economic opportunity. We can keep working, shipping products, and serving clients while the fundamental relationship that makes our business possible is quietly captured elsewhere. Disruption does not always arrive with dramatic collapses; it frequently marches in quietly alongside record revenue—until the day a business wakes up to realize that while it remains undeniably useful, it has also become terribly easy to replace.
In Blade Runner (Ridley Scott, 1982), Roy Batty says farewell with a line that has stayed with me my entire life: “All those moments will be lost in time, like tears in rain.” I think of that line when looking at the countless digital apps, software brands, and consumer habits that feel permanent today. Many will survive for decades as automated subroutines deep inside software systems, long after they have vanished from human memory.
Just like the hotel in our opening scene. The room was sold. The family trip went wonderfully. And nobody remembers who made it possible.
Jesús Lacera C.
Director of Digital Strategy & Founder of SEO-Invoke · September 2026
