What happens when there are more AI agents than humans browsing the web?
It’s 2025 and predicting the future of the internet is like staring into dense fog — if you ask me what happens in 6 months I have a pretty good idea, but if you ask me what the next 6 years will be like I have no idea. One thing is certain, the Internet as we know it is on the precipice of change as AI agents take over browsing, searching, and eventually purchasing decisions.
We get asked a lot of the same questions regarding AI:
- “How is AI going to impact my traffic?”
- “How are AI overviews impacting my click-through rates?”
- “How do we optimize for AI?”
These are important questions in the near term. But what if you imagine 3 years from now? What will marketing look like as AI starts to take precedence? For that matter, what will the future of the Internet look like?
The CEO of NVIDIA said 2025 is the year of AI agents. Google, Anthropic, ChatGPT, and others (Opera, Amazon, China!!) have all announced some form of AI Agents. But AI agents in early 2025 are hardly usable, let alone easier than Googling something yourself or booking your own Uber. But it will happen. When AI agents start to work well, humans will use them because humans prize convenience. If (or when) that becomes true, human traffic on the web will decrease while AI agent traffic increases.
As human traffic wanes, businesses must fundamentally rethink their digital presence as we transition from a human-browsed internet to an agent-mediated web. If your goal today is building traffic, you’re in trouble with the coming AI wave. But if you’re building an audience, a brand, or a community, you’ll be able to survive.
We’re starting to see early casualties from AI
Consider Chegg, once a $12 billion education platform that dominated the homework help space. By early 2025, its stock had plummeted to around $1 per share, leading the company to sue Google for scraping its content into AI search results.
“Chegg argues that by scraping websites and surfacing snippets directly in the search results page, Google is killing demand for original content and eroding the financial incentives for companies like Chegg to actually invest in producing the material that powers AI programs,” reported Gizmodo.
Chegg isn’t the only victim of AI. It’s the canary in the coal mine for traffic-dependent business models across the internet. When AI agents can extract, synthesize, and deliver information without sending users to your website, what happens to businesses built on pageviews and ad impressions?
As I explored in my earlier article on Google’s Generative Search, we’re heading toward a state of “content hyperinflation” where the perceived value of online content drops precipitously. When AI can generate content at scale, the oversupply causes each individual piece of content to become less valuable—like currency in an economy where the mint never stops printing money. In this environment, traffic becomes an increasingly unreliable currency.
If AI becomes the dominant user type on the web, it’s important to understand how they experience it.
AI agents won’t “see” websites the way we humans do, they will interact with them in fundamentally different ways. If we want to understand how we can “optimize” for AI, then it’s best to understand how they perceive and interact with the web. These are speculative, but potential ways AI might differ from human web users are:
- Parallel processing: We humans do one thing at a time, meaning we only look at one page at a time. AI can process information across multiple pages at once, comparing what they find simultaneously.
- Prioritize validation: Yes, humans do this too. However, as we force AI to make decisions for us, they will need to find information that is verifiable. That may mean that traditional marketing lingo and persuasiveness won’t carry as much water. Rather, they’ll be looking for facts, specs, pricing. Hard information, not the soft messaging that can persuade humans.
- API-first interaction: APIs are how machines communicate with each other. Current AI agents are browsing the web like a human. However, as we start to optimize for the AI experience, more and more websites and services may start to use APIs as a means to communicate with AIs. API connections will give AIs deeper, more efficient access to our data and services.
- Structure over aesthetics: I’m a sucker for a good looking website. The colors, fonts, layouts give me a good vibe of a brand and their products. Design may be used as an input for AIs into the trustworthiness of websites. However, it’s much more likely that AI will value information architecture. Much like how we make websites digestible for search engines, we’ll need to make sure AI can understand the structure and data our website is presenting.
While AI models are increasingly connecting to search indexes for fresh information (as Microsoft’s use of IndexNow demonstrates), how they process and prioritize that information differs fundamentally from how humans navigate websites.
New patterns of information consumption and commerce that could emerge
As AI agents become more prevalent, we could see shifts in how information and products flow through the internet. When the balance shifts from a mostly human web to a mostly AI agent web, there will be radically different information flows across the web.
Task completion & AI-to-AI communication
A huge shift will be away from search to task completion. Let’s say I want to buy a pair of shoes. In our current world, that may include 20 different Google searches and visiting 10 different websites and 30 different webpages. In the future, that could be one prompt for your AI agent of choice, and then the AI communicating with multiple vendors via API to research and buy a pair of shoes.
This also includes another example of what else the future web may hold — agent to agent transactions. Individuals will have their own AI agents who will then communicate with the organization’s AI agent to find the best products, prices, and terms.
Value-based content & reputation networks
For AI agents to be successful, they’ll need to rely on the information they find. That means they’ll be trying to verify both the value and reputation of a product or service. For content, that means AI won’t judge it based on how well it captures attention, but how verifiable and authoritative the information is.
New systems will also need to emerge to help AI agents identify trustworthy sources. Google invented Pagerank, which used links to derive authoritative and trustworthiness. No doubt, AI researchers are looking for an equivalent trust system to train their models to identify safe places to engage with on behalf of their human customers.
AI Agent Optimization: Potential Marketing Strategies for Beyond 2025
Many of us in marketing and SEO have spent years with Google’s algorithm at the center of our universe, crafting content for human readers, and building strategies to drive traffic. But what happens when the majority of your website visitors aren’t human at all?
API as a new potential front door to your business
Technical agencies will have a serious advantage in this new landscape. While traditional SEO focuses on content readability and keyword optimization, AI agent optimization might focus on API accessibility:
- Instead of monitoring search trends, you’ll need to track which questions AI agents commonly ask and what data points they consistently verify before making recommendations
- Rather than optimizing for featured snippets, you’ll be designing structured verification protocols that help AI agents confirm your claims are accurate and trustworthy
- The “top ranking” of tomorrow may not be about visibility but accessibility—ensuring your product specifications, pricing, and availability data can be instantly retrieved through standardized API calls
Microsoft is already building this exact type of structured API for Azure. WordPress agencies (like ours) will need to enhance their existing REST API implementations. The WordPress API already provides good endpoints for posts and pages, but we’ll need to extend it with custom endpoints for business-specific information, better structured data, and clearer metadata that helps AI understand content relevance.
From visual design to information architecture
The beautiful hero image on your homepage? AI doesn’t care. The clever headline copy? Lost on the LLMs. While these elements will remain important for human visitors, the new optimization will focus on structured information and verifiable data.
This means developing systems that verify your content’s accuracy, freshness, and originality in ways AI can programmatically check. Speed will be critical too – if your APIs don’t respond within milliseconds, AI agents might simply move on to faster sources.
Your audience becomes your lifeline
As we said before, if you’re building traffic, you’re in trouble with AI. If you’re building an audience, you’ll be able to survive.
Without relying on traffic-based metrics, your owned audience data becomes your most precious asset. This means:
- Email lists, SMS subscribers, mobile app users, and community members become gold
- Direct communication channels that don’t rely on search or social intermediaries gain tremendous value
- Community building becomes essential – providing human connection that AI can’t disintermediate
This isn’t new advice, but the stakes are higher. Those email lists you’ve been half-heartedly building? They’re about to become your business lifeline when AI agents start intercepting your traffic.
Trust signals in an AI world
Google used PageRank to determine site authority based on links. AI agents will need their own trust mechanisms, and smart marketers will get ahead of this:
- Develop explicit credentials and verification that signals your expertise to AI systems
- Make your information sources crystal clear and verifiable
- Use standardized markup to identify your content’s origin, date, and authority
Remember those “trust badges” websites used to display for security certificates? Imagine a new generation of machine-readable trust credentials that AI agents check before recommending your product or service. Can these be faked? I can imagine an entire future where AI is used to scam AI, but that’s a different article.
Walking in both worlds
The most successful marketers won’t abandon human-centered design and messaging – they’ll learn to balance both worlds:
- Design customer journeys that account for AI research followed by human decision-making
- Create content with both emotional resonance for humans and factual verification for AI
- Develop value propositions that work in both contexts – persuasive to people, verifiable to machines
The divide between “technical SEO” and “content marketing” will blur. Content without proper technical implementation will be invisible to AI agents, while technical excellence without human value will fail to convert when humans enter the journey.
Just as we’ve adapted from yellow pages to Google to social media, we’ll adapt to this next phase of the internet. The fundamentals remain – providing genuine value, building relationships, and solving problems for your audience – but the mechanics of how we accomplish this will undergo a dramatic shift.
All of this is speculation about what might develop over the next few years. The landscape of AI and marketing is evolving so rapidly that some of these predictions may materialize sooner than expected, while others might take different forms entirely. What’s certain is that change is coming, and adaptive strategies will be essential.
In our upcoming article, we’ll take a more tactical approach, exploring specific techniques and tools marketers can implement in 2025-26 to navigate this evolving terrain. The future may be foggy, but by preparing now, we can help our clients stay visible—to both humans and machines—in the post-human internet.
