The Premise
In classical economics, search theory models how individuals or businesses expend time and effort to find the best possible option, whether it’s a product, partner, or piece of information.
But much of that effort is being eliminated in today’s AI-powered landscape.
AI engines, large language models, and synthesized search results now act as zero-friction intermediaries, delivering answers instantly, often without any need to browse or click.
This dramatic reduction in effort has sparked a new dynamic:
Some companies are losing traffic not because they did something wrong—but because AI realized they were never truly necessary.
This is what is referred to as a Delphic Cost:
The hidden price organizations pay when they’re quietly removed from the discovery process is not due to poor content but because the system has found a more efficient answer.
On a personal note: I’ve been studying how digital technologies reduce friction in international trade since my original business school thesis in 1994. Back then, I predicted the internet would transform global commerce by directly connecting buyers and sellers. Today, with AI, we are witnessing another leap forward, making connections even easier and eliminating inefficiencies. Check out my What is old is new again post.
What Are Delphic Costs?
Delphic Costs are an emergent phenomenon—a real-world consequence that arises when classical search theory (which values optimal matches and friction reduction) intersects with modern AI systems that remove user effort entirely. Modern researchers build on this by applying it to web search behavior, extending the idea of economic search into cognitive, interactional, and time-based costs that users incur while seeking information.
This is the essence of the framework Andrei Broder and Preston McAfee introduced in their 2023 paper, “Delphic Costs and Benefits in Web Search.” In it, they describe Delphic Costs as the hidden price organizations pay when they’re quietly removed from the discovery process, not due to low quality, but because the system found a more efficient way to deliver answers.
Their framework signals a critical shift—from merely routing users to resolving their intent, which aligns perfectly with the rising trend we’re now observing: the publisher fallout from AI systems becoming the dominant intermediary.
They argue that user satisfaction with search engines depends not on rankings alone, but on the experience of Delphic costs and benefits on the path to completing a task. This framing supports my emphasis on friction reduction as a strategic imperative.
Key insights from Broder and McAfee’s 2023 paper that reinforce this concept:
- Searchers carry hidden non‑monetary costs—time, cognitive load, and interaction effort.
- Delphic utility matters more than traditional precision metrics: user satisfaction is tied more to reduced friction than to exact match rankings.
- Reducing Delphic Costs has shaped search evolution: the move from IR to rich UI and now to LLMs was driven by minimizing user effort.
- Integrated actions (e.g., direct bookings from results) show how AI removes multi-step friction, eliminating the need to visit the original source.
Brands pay a Delphic Cost when:
- They ranked due to early visibility, not lasting authority
- They captured traffic for content they didn’t truly enrich
- They’re quietly replaced in AI answers by someone (or something) better
Named after the Oracle of Delphi, the once-revered source of truth, these costs reflect what happens when the web’s “truth-givers” are algorithmically re-evaluated and deemed unnecessary.
Revisiting Broder’s Search Intent Taxonomy
If you are a student of SEO, the name Andrei Broder may ring a bell. In 2002, Andrei Broder, while at IBM, introduced a landmark taxonomy of web search intent, classifying queries into three core types:
- Navigational – finding a specific site or brand
- Informational – learning about a topic or concept
- Transactional – completing an action (e.g., buying or signing up)
For over twenty years, this framework guided website structure, content creation, and SEO execution. It shaped the SERP, user expectations, and strategy.
The irony? The very framework that helped us map user intent is now being used against us.
AI doesn’t ignore Broder’s taxonomy—it accelerates it. It fulfills intent more efficiently. Especially, informational intent is often satisfied before the user even considers clicking.
So perhaps Broder wasn’t just right. He was even more right than we realized.
Mapping the Connection: Search Theory → Delphic Consequences
| Search Concept | AI-Era Behavior | Delphic Cost Trigger |
|---|---|---|
| Search friction creates value for intermediaries | AI removes friction via synthesized answers | Brands built on visibility are bypassed by instant responses |
| Broder’s “Informational” queries create discovery funnels | AI short-circuits discovery by delivering resolved intent in place | If you’re not the final answer, you’re not surfaced |
| Visibility ≠ Authority | AI prioritizes trust, clarity, and structure | Content built for SEO, not substance, is demoted or excluded |
| Intermediaries win by reducing effort | AI becomes the dominant intermediary | Traditional publishers, blogs, and product sites are disintermediated |
The Hollow Content Trap: Adjacent but Not Authoritative
In an attempt to “capture” the funnel, many B2B and e-commerce companies flood the web with shallow adjacent content, including glossaries, topic pages, and basic definitions, hoping to rank for broader informational queries. The faulty yet straightforward logic is that they must want my ultra-wiz-bang cloud computing ecosystem if they are interested in cloud computing. Or, more commonly, how can we super-tune a webpage to get a featured snippet and drive them to a page suffed with dozens of non-relevant ads and autoplay videos to fund our generally failing tech publication?
But most of it is regurgitated, redundant, and devoid of insight.
Examples:
- B2B: Cybersecurity firms are defining “What is a firewall?” with no context or link to their unique capabilities.
- E-commerce: Retailers listing vague specs like “600 watts” or “ABS plastic,” offering no guidance on suitability, fit, or decision factors.
AI can summarize that kind of content faster, better, and from more trusted sources.
You’re not being cited because you’re not needed.
The system is trained to skip friction. If your content doesn’t clarify, differentiate, or reduce user effort, it’s ignored. These costs reflect what happens when the web’s “truth-givers” are algorithmically re-evaluated and deemed unnecessary.
Erosion of the Pageview and Attention Economy
For publishers and content marketers, this signals the collapse of a familiar economy. The old web rewarded those who could capture attention through SEO, content velocity, or emotional hooks.
Delphic Costs flip that model.
Today’s AI engines reward those who reduce friction and resolve intent instantly. The winners are answer engines, not publishing platforms. Zero-click answers replace multi-page funnels. Structured data replaces scrolling. Authority replaces engagement hacks.
You’re not losing traffic—you’re being removed as an unnecessary step.
Strategic Imperative: Make Friction Reduction Your Business Model
This isn’t just a content or SEO issue. It’s a business strategy issue.
- If your onboarding is complex, AI will summarize you and route around you.
- AI will compare others and omit you if your product selection lacks clarity.
- If your answers require effort, AI will rewrite them—and take credit.
In the age of AI-mediated discovery, friction reduction is not a UX best practice—it’s a survival strategy.
Every function—marketing, product, operations—must now ask:
- Are we the fastest path to clarity?
- Are we indispensable to the resolution?
- Are we still part of the user’s intent fulfillment or just a remnant of their search journey?
From Intent to Resolution — and the Cost of Being Skipped
Broder helped us understand what users wanted. AI shows us what happens when someone else gives it to them faster.
Delphic Costs are not algorithmic punishment—they’re algorithmic clarity.
They emerge when you’re no longer needed to fulfill the user’s intent. And they remind us that in this new paradigm, visibility isn’t enough.
To survive, you must be indispensable to the resolution.
