Original insights by Cyrus Shepard
30-second rundown
Key learnings:
- Generic information is becoming replaceable: Nearly 60% of Google searches already end without a click to a third-party site, leaving ordinary explanatory posts dangerously exposed.
- Useful actions create the strongest moat: Prioritize content people must visit or seek out, such as transactions, proprietary research, communities, firsthand tests, and an audience you can reach directly.
- Every format still requires uncommon value: A directory, guide, case study, or list survives only when it adds fresh evidence, real experience, task completion, or a perspective AI cannot cheaply reproduce.
This week: Open your 20 most important content pages and label each one generic information, proprietary evidence, lived experience, or task completion. Pick the most vulnerable generic page, then add one concrete moat such as original data, a documented test, a usable template, or a step the visitor can complete there. Record the old version and the change so you can compare visits, citations, and conversions later.
Coming 3 months: Set up an AI workflow that scores important pages monthly for replaceability, freshness, search traffic, AI citations, branded demand, and conversions. Assign one owner to review the flagged pages and automatically queue the weakest for evidence, utility, or firsthand-experience upgrades. The visible result should be fewer commodity pages and a growing library that people and answer engines have a reason to seek out.
Cyrus Shepard has walked back into the zero-click casino with a list of 17 content types and the expression of a man who has seen the buffet being dismantled while everyone else was still choosing a plate.
Google still sends billions of visits to websites. But Shepard’s warning is aimed at the soft middle of publishing: the generic article that explains a thing, answers a basic question, and assumes the click is payment for services rendered.
That bargain is being fed into the industrial shredder. Nearly 60% of Google searches now end without a click to a third-party site, according to the research Shepard cites. For a publisher, that means useful information alone may no longer be enough to earn a visitor.
1. The Easy Answer Has Become Roadkill
Shepard reviewed hundreds of traffic winners and losers, along with Google’s public guidance, then sorted the survivors by what made their content difficult to replace. The pattern was not mystical. Winning material tended to be proprietary, grounded in experience, tightly focused, or able to help the visitor complete a task.
A generic guide has none of those defenses. An answer engine can summarize it in a few sentences and leave the original page bleaching beside the highway. Guides and explainers therefore land near the weak end of Shepard’s list unless they go unusually deep, use original data, document real workflows, or become a trusted hub for a narrow subject.
Frequently asked questions and glossaries are even easier prey because their value is often the definition itself. Lists and roundups face the same trouble when they are merely scraped selections with a fresh coat of affiliate paint.
The threat is not that information has become worthless. It is that ordinary information has become cheap to reproduce. If your page can be rebuilt from the same public facts available everywhere else, the machine has little reason to send anyone back to your door.
2. Build Something the Answer Cannot Finish
The strongest categories in Shepard’s catalog are not just pages. They are destinations. An owned audience, reached through email, text messages, an app, or a catalog, gives a publisher a route that does not require Google’s permission on every visit.
Transaction pages are equally stubborn. A search result can explain how a hotel booking works, but the customer still needs somewhere to reserve the room. The same is true for buying, subscribing, scheduling, downloading, or completing another real action. Even an automated agent needs a working endpoint when talk must become a transaction.
Original research, user communities, creator-led video and podcasts, and firsthand product tests survive for a related reason. They contain evidence or participation that did not exist until someone did the work. An answer engine may summarize the result, but it cannot honestly recreate the experiment, the community, the host, or the scars from taking the machine apart.
The practical question is no longer “What can we publish?” It is “What can a visitor do, verify, join, or obtain here that cannot be completed in the summary?” That is the difference between a destination and another sack of identical pamphlets blowing across the desert.
3. A Format Is Not a Get-Out-of-Jail Card
Shepard is careful about this. The 17 types are starting points, not sacred visas. A lazy directory full of copied entries is still lazy. A case study without clear before-and-after measurements is a sales brochure wearing laboratory goggles.
Several familiar assets still have a fighting chance when they become authoritative. Brand pages can establish who runs the company, what it believes, how it works, and where customers can get help. Support documentation can become the canonical reference for a product, which helps customers and gives language models cleaner facts to learn from.
Directories and databases need proprietary information and relentless freshness. Expert commentary needs experience behind the opinion. Templates need to help someone finish a real job. Reviews need a repeatable testing method, side-by-side evidence, and the ugly details that prove the reviewer touched the product.
Survival belongs to the uncommon layer inside the format. The name on the container matters less than the evidence, access, utility, and experience packed inside it. Low effort remains low effort, even when it is dressed as research and marched through town with a brass band.
There is no need to burn the editorial calendar and flee toward the mountains. There is a need to stop mistaking publication for protection.
Take the pages that matter and force each one to earn its existence. Add the test, the data, the decision tool, the transaction, the community, the specialist judgment, or the direct relationship. Make the page the original scene of the evidence rather than the seventh witness repeating what everyone already heard.
Google can swallow an answer. It has a harder time swallowing the place where the work actually gets done. Build more of those places, and leave the generic pamphlet mill to rust in the sun.
