Buckle up, we’re going on a journey (but don’t worry, it’ll all make sense soon). I promise this isn’t just filler—everything will become crystal clear.
Chapter 1: On the Road to Delphi!
Picture me in ancient Greece, rocking sandals (I own it). I’ve got a deep, existential question, something crucial, and since we’re in 500 BC, there’s no Google, no Wikipedia. Not even a dusty old encyclopedia at the local library.
Back then, if you had a question, you did what everyone else did: you grabbed your staff, your courage, and set off on a pilgrimage to Mount Parnassus in Delphi. Because that’s where the Oracle was—the Pythia, the ultimate source of knowledge.
So, I put on my tunic (pretty flattering, by the way), and off I go. First step: crossing Greece.
Now, Delphi isn’t exactly next door. The roads aren’t paved, the inns smell like goat (and still cost money), and I have to dodge a few bandits along the way. It takes me days and days, a bunch of money, and a lot of effort to get there. Clearly, accessing information in 500 BC is expensive and anything but fast.
Finally, I arrive—filthy, exhausted.
Of course, there’s a line. The Oracle isn’t a chatbot available 24/7. I wait under the blazing sun, trying not to die of dehydration, and when it’s finally my turn, I have to go through an elaborate ritual.
Laurel leaves are thrown into a fire, a goat is sacrificed (RIP), and only then can I ask my question. I have to phrase it perfectly to make sure it’s understood—otherwise, the answer might be completely off. I get one shot. No do-overs.
So, with all this pressure, after all this hassle… I forget my question.
Panicked, I blurt out, “What’s the recipe for pancakes?”
Awkward silence. The Pythia stares at me. The priests stare at me. Even the sacrificed goat seems disappointed.
And the Oracle, in her deep voice, replies: “Mix milk and flour, and the rest will follow.”
Not exactly precise. Not exactly reliable.
So, after weeks of travel, I head home with a useless recipe in my bag.
Useless? Not for everyone. It was a similar parable that inspired Andrei Broder and Preston McAfee from Google to come up with the concept of Delphic costs—a framework for understanding the costs associated with searching for information, detailed in their aptly named paper, Delphic Costs and Benefits in Web Search: A Utilitarian and Historical Analysis (available in PDF).
During my journey, I paid a high price in terms of access cost (travel, lodging, food), cognitive cost (phrasing my question correctly, interpreting the answer), and interactivity cost (I literally sacrificed a goat and made a fool of myself with my question!). When I calculate the overall cost of this expedition versus what I got in return, the ROI isn’t looking great.
Clearly, I won’t be making a habit of trekking to Delphi for every random question that pops into my head!
Chapter 2: Welcome to the 21st Century (But It’s the Same Thing, Really)
Fast forward to 2018. Nothing has really changed.
Sure, I don’t have to walk across Greece to get an answer, but I still have to go through a whole obstacle course to get a precise one.
Let’s take a concrete example: I have a question: “Why do cats meow?”. I go to Google (much closer than Delphi). I type my query (gotta phrase it properly, but at least no one’s judging me).
I check the results. One site says it’s a form of communication (true? You decide). Another mentions something about cat stress. A third explains that some breeds meow more than others.
I open multiple tabs (too many different answers). I spend a few minutes reading. I cross-check the info, and eventually, I get my answer.
Total time spent: a few minutes. Better than days of walking, sure, but Delphic costs are still there.
I invested time, had to check multiple sources, separate fact from fiction, click, scroll, use electricity, and rely on highly developed infrastructure.
Bottom line: finding information still takes effort. Less than before. But I’m kinda lazy…
Chapter 3: Generative Search Engines—A Better Pythia (Almost)
And then, at the end of 2022, generative search engines (GSEs) like ChatGPT, Gemini, and friends arrived. Their promise? To eliminate all those steps and give you the answer directly.
How does it work? I type (or say), “Why do cats meow?” and in the background, the GSE searches the web. But instead of just handing me a list of pages that might answer my question, it goes the extra mile: it reads the content and constructs an answer.
“Cats meow to communicate with humans. They use different vocalizations depending on their needs, such as hunger, boredom, or attention-seeking.”
Done. A clean, well-written, ready-to-use response.
Why is this amazing? Because it destroys Delphic costs.
I didn’t waste time—got my answer instantly.
I didn’t have to think—no sources to cross-check, the GSE did it for me.
It understood my question perfectly—no goat sacrifices, no embarrassment.
Sound familiar? It’s like the Oracle of Delphi, but instant. Except, where the Oracle gave cryptic, sometimes useless answers, the GSE synthesizes a clear, understandable response.
Chapter 4: But Hold Up, It’s Not Magic
If it were perfect, I’d have already switched completely to GSEs. But I still use Google the old-fashioned way sometimes. Because even though Delphic costs have plummeted, they’re still pricey.
First, generative search engines aren’t encyclopedias. They know a lot, they generate convincing text, but… sometimes they make stuff up.
I ask for a historical date? Usually, it nails it. Sometimes, it doesn’t.
I want a scientific quote? It’ll give me one, even if it’s completely made up.
Second, the “single response” format creates an illusion of truth. I want to believe this polished, well-dressed answer. But that’s dangerous—what if it’s wrong? Then again, the Pythia was cryptic too, and people just interpreted what suited them.
Finally, running a GSE is insanely expensive. For now, Google, OpenAI, Microsoft, etc., are covering a big chunk of the technical costs to attract users. But with OpenAI launching $200/month subscriptions, can we really say Delphic costs are decreasing?
Conclusion: The Inevitable Shift to Simplicity
Despite its flaws, generative search is the future.
Why? Because simplicity always wins. Even if it hallucinates, even if it’s not perfect, I’d rather have a quick, imperfect answer than a long, precise but tedious process. Except for critical things. But hey, I can lower my standards if it means more nap time…
And anyway, we’re promised that GSEs will keep improving, becoming more accurate and reliable. So, who knows?
I may not walk across Greece anymore for answers, but I still want knowledge as smoothly as possible.
And GSEs, with their instant responses, mark the next step in this quest for simplicity.
So, ready to say goodbye to SERPs and follow Oracle 4.0?
