AI agents are becoming the Internet’s second audience, forcing Google, publishers, businesses, and search platforms to rethink traffic, attribution, access, payments, and who gets the lion’s share of value from the next Internet.
DUBAI, United Arab Emirates, October 6, 2026: For decades, science fiction imagined the first serious confrontation between humans and machines as something physical.
Metal skeletons. Ruined cities. Resistance fighters. Machines hunting people through landscapes that looked as if civilization had already lost.
The Terminator movies gave that fear one of its most recognizable forms.
Reality, unsurprisingly, has arrived differently.
There are no T-800 endoskeletons marching through Silicon Valley. There is no Skynet declaring war on humanity. Nobody is fighting over a weapons factory.
But a much quieter battle between humans and machines may already have begun.
And one of its first major battlefields is Google Search.
Not a battle in the military sense, of course. This is a fight over something less dramatic but potentially far more consequential to the digital economy: visibility, access, traffic, information, attribution, computational resources, commercial value, and control.
For roughly 25 years, publishers, businesses, journalists, marketers, ecommerce companies, creators, and SEO professionals competed against other humans and organizations for the same limited territory on Google.
Rank higher, win attention.
Win attention, earn clicks.
Turn enough clicks into subscriptions, customers, advertising revenue, leads, sales, influence, or brand recognition.
That model is now being challenged by something fundamentally different.
AI agents have entered the same information ecosystem.
And machines do not use the Internet the way humans do.
The Internet Was Built for Humans. That Is No Longer the Whole Story
For most of the commercial Internet’s history, the economic relationship was reasonably easy to understand.
A person opened a browser.
They searched for something.
They clicked a website.
They read an article, compared a product, watched a video, saw an advertisement, filled out a form, purchased something, subscribed, or simply left.
Bots were always around. Search engines crawled websites. Monitoring tools checked servers. Scrapers collected information. Automated systems indexed enormous amounts of content.
But those machines were generally intermediaries.
The human remained the audience that ultimately mattered economically.
That distinction is beginning to disappear.
Cloudflare now describes AI agents as the Internet’s “second audience.” Its network data shows why the phrase matters.
At the end of 2024, Cloudflare says it was handling around 63 million HTTP requests per second. That figure has since risen to approximately 115 million per second, with peaks above 150 million.
More strikingly, Cloudflare says daily requests from AI agents across its network increased by more than 1,700% in one year, while more than half of the traffic reaching sites on its network is now automated rather than human.
Cloudflare’s own analysis of the emerging agentic web argues that the human Internet did not shrink. Instead, another audience arrived beside it.
Machines.
Or, more precisely, software acting increasingly on behalf of people.
Some agents research.
Some compare products.
Some monitor prices.
Some summarize journalism.
Some analyze financial data.
Some search for jobs.
Some inspect technical documentation.
Some transact.
Some retrieve information for AI assistants.
Some train models.
Others simply extract information at extraordinary scale.
The web did not lose its first audience.
It gained another one.
That changes far more than SEO.
Google Spent Decades Being Watched
SEO has always involved watching Google.
An enormous global software industry grew around observing its results.
Rank trackers queried Google repeatedly to determine whether websites moved up or down.
Keyword databases attempted to map search demand.
Competitive intelligence platforms estimated which queries sent traffic to rival websites.
SEO teams tracked visibility indexes, featured snippets, local results, shopping listings, news carousels, ranking changes, and almost every measurable part of the SERP.
This became so normal that people rarely stopped to consider how unusual it was.
Millions of automated searches could be performed, stored, analyzed, compared, and ultimately sold as commercial intelligence.
Google has never treated unrestricted automated querying as an entitlement.
Its current Search spam policies explicitly describe machine-generated traffic as automated queries sent to Google, including scraping search results for rank-checking purposes without express permission.
Google states directly that automated rank-checking queries can violate its policies.
That policy is not new in principle.
What has changed is the scale and nature of machine demand.
Traditional rank trackers are no longer the only systems trying to interrogate the world’s largest search engine.
AI agents have arrived too.
Rank Trackers May Be Only the First Visible Casualty
The immediate SEO discussion focuses on a familiar concern.
What happens if it becomes increasingly difficult for third-party tools to collect Google results reliably at industrial scale?
That matters because the SEO industry has spent years turning observed SERPs into apparently precise metrics.
Position 4.
Visibility 22.6%.
Estimated monthly traffic 38,000.
Keyword volume 4,400.
Competitor traffic 120,000.
These numbers can be useful.
But they are not all measurements of reality.
Many are estimates produced by observing a search engine from outside.
A ranking database does not contain Google.
It contains what the provider was able to observe from Google under particular technical conditions, locations, devices, query sets, and collection methods.
If automated observation becomes less reliable, the gap between estimated search visibility and actual website performance matters more than ever.
That may eventually force SEO teams back toward something they should arguably have prioritized all along:
first-party truth.
Actual Google Search Console impressions.
Actual clicks.
Actual landing pages.
Actual user engagement.
Actual conversions.
Actual revenue.
AI Has Changed the Economics of a Search Query
There is another reason Google has stronger incentives to distinguish human activity from automated machine activity.
Search itself has become computationally more complicated.
A conventional search engine retrieves and ranks indexed documents associated with a query.
An AI-powered search experience can expand the query, conduct multiple retrieval steps, interpret multimodal information, reason across sources, synthesize information, and generate a response.
Google says AI Overviews now have more than 2.5 billion monthly active users, while AI Mode has surpassed one billion monthly users.
The company has also introduced dedicated Search Console reporting and new controls allowing publishers to understand and manage how their content appears in generative AI Search features.
Google’s own announcement details both the scale of AI Search and the new publisher controls.
Now imagine that enormous legitimate human demand existing alongside industrial automated querying.
Rank trackers query Google.
Research bots query Google.
AI assistants may query Google.
Agents compare prices, products, companies, flights, regulations, people, financial information, and market conditions.
To Google’s infrastructure, all of these create computational demand.
But they do not necessarily create equal economic value.
The old problem was stopping abusive scrapers.
The emerging problem is much more complicated:
How does Google distinguish useful machine activity from extractive machine activity without crippling legitimate agents?
Google is not alone in confronting that problem.
Every serious publisher may soon face the same question.
The Old Web Bargain Is Starting to Break
For roughly three decades, publishing on the web relied on an imperfect but understandable economic bargain.
A publisher created information.
A search engine crawled it.
The search engine indexed it.
A person searched.
The person clicked.
The publisher received traffic.
Some of that traffic became revenue, subscriptions, customers, leads, influence, or audience growth.
There were always arguments over whether search engines captured too much economic value.
But there was still an exchange.
Content for discovery. Discovery for traffic.
AI can break the second half of that bargain.
An AI system can retrieve information from several websites, combine it, summarize it, and answer the user directly.
The publisher still paid to create the information.
The journalist still did the reporting.
The researcher still collected the data.
The photographer still took the image.
The editor still checked the story.
The developer still maintained the site.
The server still processed the machine request.
But the human reader may never arrive.
That is a fundamentally different economic model.
Who Exactly Is Visiting Your Website?
A request reaches your server.
What is it?
A human reader?
Googlebot?
An AI training crawler?
A retrieval system sourcing an answer?
A commercial agent comparing products?
A crypto research assistant examining protocol documentation?
A malicious scraper?
An autonomous system preparing to complete a transaction?
All of these may technically access the same page.
Economically, they are completely different visitors.
This distinction becomes particularly important in crypto, where automated systems already interact naturally with APIs, blockchain data, exchange information, smart contracts, wallets, price feeds, and transaction infrastructure.
The Crypto Encounter has already examined how AI agents could radically scale crypto fraud, precisely because machine automation allows one system to perform work that previously required many people.
Similar technology can obviously be used productively.
The question is not whether agents are inherently good or bad.
The question is what they are doing, whose interests they represent, and what economic value they create or extract.
The Battle Is Really About Who Captures the Value
The phrase “humans versus machines” makes for a powerful visual.
It also risks oversimplifying what is actually happening.
This is not a fight in which humans somehow defeat AI and return the Internet to 2019.
The real battle is over who captures value from information.
The traditional search model looked roughly like this:
Publisher → Google → Human → Publisher
The emerging AI model can increasingly look like:
Publisher → AI system → Human
The final return to the publisher can become optional.
That one missing step affects the economics of news, research, reviews, affiliate publishing, ecommerce, software discovery, travel, financial information, and countless other industries.
It also creates an important new distinction.
An agent that sends a qualified customer to a company could be enormously valuable.
An agent that consumes thousands of pages while sending no traffic, attribution, payment, or commercial opportunity may be extractive.
Both are machines.
They should not necessarily be treated the same way.
Google’s Position Contains an Obvious Irony
Google built one of the most powerful companies in history by crawling the open web.
That open web made Search possible.
Millions of publishers effectively allowed Google to discover, index, interpret, and organize their information.
In return, Google sent enormous amounts of traffic back.
Now Google itself increasingly operates in an environment where machines want systematic access to its outputs.
SEO tools want rankings.
Scrapers want information.
Agents want results.
AI systems want answers.
Google therefore finds itself confronting a version of the same question publishers are asking:
Who should be allowed to consume our infrastructure, under what conditions, and what should we receive in return?
The irony is obvious.
Google wants access to an open web while simultaneously retaining control over automated access to Google.
But publishers increasingly want exactly the same thing.
Publishers May Need Rules for Machines
The web has historically treated crawler access largely as a technical yes-or-no question.
Allow.
Block.
That model may soon be too primitive.
Publishers increasingly need to distinguish among different types of machine activity.
Search indexing may be welcome.
AI retrieval with attribution may be welcome.
Commercial recommendation agents may be welcome.
Training may require separate permission.
Bulk extraction may require licensing.
Premium datasets may require payment.
Autonomous transactions may require identity and authorization.
This already matters to crypto publishers because synthetic information can create serious trust problems. TCE has documented how an entire crypto team can be fabricated using AI, making machine-readable evidence and independent verification increasingly important.
The Internet therefore does not simply need better AI.
It needs better rules governing the relationship between machines and information owners.
The Web May Be Developing a Machine-Access Economy
Cloudflare is already experimenting with one possible response.
Instead of forcing publishers to choose between unrestricted crawler access and complete blocking, its emerging model attempts to let content owners establish terms around machine access and usage.
The broader idea is simple:
“Yes, you can use this, but not necessarily for free.”
That is potentially significant.
For three decades, most websites monetized humans.
Humans saw advertisements.
Humans bought subscriptions.
Humans purchased products.
Humans generated leads.
Machines generally did none of those things.
Now machines can potentially become economic participants themselves.
That shift connects directly with another technology TCE has been tracking closely: programmable payments.
AI Agents Could Become Crypto’s Most Important Non-Human Users
Crypto may be particularly well suited to an agentic Internet because much of its infrastructure is already designed for software.
Blockchains are machine-readable.
Wallets are programmable.
Smart contracts can execute automatically.
Stablecoins settle digitally.
APIs expose markets and financial information programmatically.
An AI agent does not naturally want to enter credit-card information into a checkout form.
Software wants software-native payment rails.
This is why protocols such as x402 are attracting attention.
TCE previously examined how security vulnerabilities in x402 implementations exposed the risks of allowing AI agents to spend crypto at machine speed.
The security concerns are real.
But so is the underlying opportunity.
An agent capable of purchasing information, compute, API access, research, software tools, or digital services without requiring a human to complete every transaction represents an entirely new commercial actor.
Stablecoins could become part of that infrastructure because they allow digital value to move through programmable systems relatively easily.
That does not mean stablecoins automatically become the universal currency of AI agents.
TCE’s broader coverage has repeatedly shown that digital payments remain dependent on surrounding infrastructure. Crypto payments become considerably more complicated when refunds, freezes, compliance, and redemption enter the picture.
Likewise, the question of whether stablecoins can genuinely replace conventional bank transfers depends on much more than transaction speed.
The agent economy will inherit many of those complications.
Crypto Could Become One of the First Agent-Native Industries
Consider how much crypto information already exists in machine-readable form.
Token prices.
Blockchain transactions.
Exchange order books.
Smart-contract activity.
Protocol documentation.
Treasury balances.
Governance votes.
Token unlocks.
Stablecoin supply.
Wallet movements.
Developer documentation.
Regulatory announcements.
On-chain analytics.
An investor today might manually visit several websites to understand a development.
An agent can increasingly perform that research across many sources and return one synthesized response.
Imagine asking:
What materially changed in Ethereum staking during the past week, which primary sources confirm it, what does it mean for validators and liquid-staking providers, and what should I watch next?
An agent could potentially inspect protocol documentation, blockchain data, regulatory notices, news reporting, market data, and company disclosures before returning one answer.
That is extraordinarily convenient.
It also creates a dangerous new problem:
Which sources did the agent trust?
Source Quality Becomes More Important When Humans Stop Visiting the Sources
Crypto already has a severe information-quality problem.
Fake projects.
Sponsored promotion presented as analysis.
Outdated token information.
Impersonation.
Manipulated social sentiment.
Fraudulent websites.
AI-generated identities.
Deepfakes.
Fabricated partnerships.
If a human reads several websites, that person at least has an opportunity to inspect the sources and compare credibility.
When an AI agent compresses those sources into one response, much of that source-selection process disappears from view.
The quality of retrieval therefore becomes just as important as the quality of generation.
TCE has already documented how AI scam bots can operate continuously and personalize crypto fraud at scale.
The same automation that makes legitimate research faster can make misinformation dramatically more scalable.
That is why the agentic Internet will need transparency around provenance, sourcing, authorization, and trust.
The Same Machines Creating Risk Can Also Create Opportunity
It would be easy to frame the rise of agents entirely as a threat.
That would be a mistake.
Agents could become enormously valuable distribution channels.
A small specialist publication might struggle to outrank multinational websites for a broad Google keyword.
But an intelligent agent searching for the most authoritative source on a narrow question could discover that smaller publication because it contains original information unavailable elsewhere.
A startup may gain customers because an AI purchasing agent determines that its product solves the user’s exact problem.
A crypto protocol with exceptionally clear technical documentation may become easier for agents to understand than a larger competitor whose documentation is fragmented or ambiguous.
A specialist analyst may gain influence because AI systems repeatedly retrieve and cite original research.
In that environment, visibility becomes less about occupying one numbered Google position and more about becoming the best source for a particular problem.
That could weaken some established advantages while creating entirely new ones.
SEO May Have to Serve Two Audiences
For years, SEO professionals essentially asked one question:
How do we make this website easier for humans to discover through search engines?
Another question now sits beside it:
How do we make information discoverable, understandable, trustworthy, and useful to machines acting on behalf of humans?
The two audiences overlap, but they are not identical.
Humans respond to narrative, emotion, design, trust, clarity, personality, usability, reputation, and experience.
Machines care strongly about structure, factual consistency, entities, accessibility, relationships, dates, provenance, and retrievability.
The answer is not to build two websites.
It is to create strong human content that machines can understand accurately.
This means clear authorship.
Stable URLs.
Logical site architecture.
Accurate structured data.
Useful headings.
Reliable dates.
Consistent entity information.
Original reporting.
Proper sourcing.
Expert analysis.
Unique datasets.
First-hand experience.
Google itself continues to emphasize unique, useful, non-commodity information rather than tricks designed exclusively for AI systems.
That may produce one of the great ironies of generative AI.
The better machines become at recreating generic information, the more valuable genuinely original human knowledge becomes.
AI Is Already Forcing Other Industries to Rethink Control
This struggle over control is not limited to search.
AI companies, governments, businesses, and infrastructure providers are increasingly arguing about who controls the rules under which AI systems operate.
The Crypto Encounter recently examined this from another direction when Anthropic’s dispute with the Pentagon raised a broader question about who controls AI safeguards.
The context was completely different.
The underlying question was remarkably similar.
Who gets to set the boundaries once intelligent software becomes deeply embedded in critical systems?
Search is now confronting its own version of that argument.
The SEO Industry Needs to Stop Treating Every Estimate as Ground Truth
The decline of easy SERP observation could have one healthy consequence.
It may force businesses to distinguish estimates from measurements.
Third-party SEO platforms remain extremely useful for research, competitor discovery, keyword exploration, link intelligence, and market analysis.
But first-party data should become the operating truth.
Google Search Console tells a website what Google actually showed.
Analytics tells the site what users actually did.
Server logs show what actually requested resources.
Conversions show what generated business value.
Revenue shows what ultimately mattered commercially.
A rank tracker remains useful context.
It should not become an oracle.
This change could push SEO toward a more mature model:
less obsession with estimated positions and more attention to actual information demand, visibility, traffic quality, engagement, conversions, and business outcomes.
Search Traffic May No Longer Be the Only Visibility That Matters
Businesses may soon need to measure something new.
Agent visibility.
Was our content retrieved?
Was it cited?
Was it used in an answer?
Did the user subsequently visit?
Did an agent purchase something?
Did it recommend our product?
Did it misrepresent us?
Did it consume thousands of pages while providing nothing in return?
The web still lacks universally accepted measurements for these questions.
That will change.
Google’s growing AI Search reporting is one early sign.
Infrastructure providers measuring agent traffic are another.
Once businesses can clearly measure the ratio between machine extraction and economic return, policies around AI access will become much more sophisticated.
The Payments Layer Could Become One of Crypto’s Biggest Opportunities
If AI agents eventually need to purchase data, APIs, software, research, computing resources, or digital services autonomously, they need payment infrastructure suitable for software.
This is where digital assets become particularly interesting.
Banking is already trying to combine AI-driven payments, tokenization, stablecoins, instant settlement, cloud infrastructure, and legacy systems simultaneously. TCE recently examined how banks are attempting to modernize payment infrastructure while old systems remain deeply embedded.
The same tension will exist in agent commerce.
Software-native transactions sound simple until authorization, fraud controls, reversals, identity, compliance, settlement, and dispute resolution enter the picture.
Crypto infrastructure could solve parts of that problem.
It could also introduce new risks at machine speed.
Real-world payment adoption is already testing where stablecoins fit into broader financial infrastructure. Ripple’s expansion into African payments, for example, shows how stablecoins and blockchain settlement are moving beyond purely speculative crypto markets.
Agent commerce could accelerate that transition considerably.
What Companies Should Do Now
Companies do not need to wait for the entire industry to settle on new terminology.
The first step is understanding what is already happening.
Know who accesses your website.
Understand which automated systems consume the most resources.
Separate traditional search crawlers from AI systems where technically possible.
Monitor how generative search surfaces your content.
Understand whether AI referrals behave differently from conventional search visitors.
Protect proprietary information where uncontrolled extraction creates risk.
At the same time, do not block legitimate machine discovery simply because the visitor is not human.
An agent that brings business could eventually be as valuable as a Google search visitor.
The strategic question is not:
How do we stop machines?
It is:
Which machines create value, which machines extract it, and what terms should govern both?
What Publishers Need to Understand
Publishers face an especially difficult transition because their core product is information.
If an AI system can consume that information while bypassing the pageview, advertising impression, subscription prompt, affiliate link, or direct relationship with the reader, the economic consequences can become serious.
Blocking everything is not necessarily the answer.
Allowing everything is not necessarily sustainable either.
The industry is moving toward a negotiated middle ground in which machine access may increasingly involve identity, attribution, permission, measurement, and payment.
That would represent one of the biggest changes in the economics of the open web since search advertising itself.
What Investors Should Watch Next
- Google’s treatment of automated Search access: Watch whether rank tracking and industrial SERP collection become materially harder, more expensive, or more selective.
- Generative Search reporting: More granular first-party reporting should reveal whether AI visibility produces meaningful human traffic or increasingly replaces it.
- Agent traffic economics: Watch whether publishers begin publicly measuring how much machine consumption results in referrals, transactions, licensing revenue, or other measurable value.
- Machine payments: Monitor x402, stablecoin settlement, HTTP-native payment systems, and other mechanisms that could allow autonomous software to pay for information and services.
- Publisher access policies: Watch whether major websites increasingly separate search indexing, AI training, retrieval, summarization, and commercial agent use rather than treating every crawler identically.
- Crypto agent security: Authorization and payment vulnerabilities could become more important as machines gain the ability to move assets autonomously.
These are observable developments, not predictions of guaranteed outcomes.
The agentic economy remains early, and its commercial rules are still being negotiated.
The Internet Has Already Changed Forever
Some technological transitions happen so gradually that people barely notice them until the old world has already disappeared.
The migration from desktop to mobile was like that.
Social media was like that.
Cloud computing was like that.
The agentic Internet may prove to be another.
Machine traffic is already enormous.
Generative search already serves billions of people.
Agents can increasingly retrieve information, reason across sources, compare options, use software, and perform tasks.
Infrastructure companies are experimenting with ways for websites to charge machines.
Publishers are demanding greater control.
Search engines are attempting to distinguish valuable automation from abusive extraction.
Payment systems are becoming programmable.
None of this resembles the Internet of five years ago.
The transformation is not approaching.
It is already underway.
The First Human-Machine Battle May Already Have Started
So are we witnessing the first genuine battle between humans and machines?
Not in the way science fiction imagined it.
No nuclear ruins.
No killer robots.
No resistance fighters hiding underground.
But something historic is happening.
Humans built the Internet for themselves.
Then humans built machines sophisticated enough to use it too.
Those machines are now competing for access to the same information, infrastructure, search visibility, computational resources, attention, and economic value that were once consumed primarily by people.
Google’s resistance to uncontrolled automated querying is one front.
Publisher compensation is another.
AI retrieval is another.
Search visibility is another.
Agent payments may become another.
The fight is not about eliminating the machines.
Machines are already part of the Internet.
The battle is over the terms of coexistence.
Who gets access?
Who creates the information?
Who extracts it?
Who receives attribution?
Who sends traffic?
Who pays?
Who controls the rules?
And ultimately:
Who captures the lion’s share of value from the next version of the Internet?
For 25 years, humans fought other humans for Google’s most valuable search positions.
Now AI agents have entered the battlefield.
And if current traffic trends are any indication, they have arrived in extraordinary numbers.
The battle is on.
Not for humanity’s survival.
For something considerably more immediate:
the lion’s share of the next Internet.