English concept: Searchless World · Zero-Click
In the history of the internet, “search” was once the entrance of all traffic and attention. Enterprises put their budget on keywords and blue links; creators also disciplined their titles and paragraphs with SEO. Now the situation is reversed: AI Overviews, conversational assistants, and all kinds of “zero-click (Zero-Click)” experience are quietly taking away the ritual of “find the material → click the link.” Several studies and platform data from 2024–2025 sketch the same curve — when a search-results page directly gives a usable answer, users click off-site links less often (Pew Research in the United States observed that, in situations where an AI summary appeared, the share of users clicking a traditional link was only 8%); at the same time, the overall zero-click share approached 60% in Europe and in the United States, and for queries that met AI Overviews, the median of zero-click could rise to about 80%. Google, from the other side, has claimed that for the types of query into which AI Overviews were introduced, frequency of use and satisfaction rose, and drove growth in overall search use. Even if the conclusions each have a position, the meaning of the trend is already clear enough: move from “writing for a click” to “writing to be cited.”
This is not a dramatized announcement that “search is dead.” It is a change of house in the position of power. The entrance moves from “the address bar” to “the semantic layer”; content is broken from a complete page into knowledge atoms that can be recalled; the platform moves from “listing options” to “answering directly.” For enterprises and content producers, what truly needs to be rewritten is not the title, but the whole way of stacking content assets so that they “can be understood and trusted by AI.”
NO° 01|The disappearance of active search: how AI turns “finding” into “giving”
In the past, the journey of information always began with a conscious action: I think of a need, enter a keyword, scan ten links, then make an assessment of trust and a comparison. An AI interface compresses this path into the operating system and into the inner layer of an application: enter “family health check” in the calendar, and the sidebar of the calendar “fills in” a choice of hospitals and a record of past visits; a mail title says “Q3 budget meeting,” and the inbox sidebar automatically spreads out a curve of historical spending and the related attachments; a social conversation asks “what is worth buying for Double 11,” and an assistant instantly stitches your consumption preferences, a summary of reviews, and a price forecast, and assembles three shopping lists of different budgets. Search is no longer an independent action. It is “invisibly embedded” by the system in every situation.
This distribution, driven by “the filling-in of intent,” recasts the economics of content. What a brand has to contest is no longer first place on the SERP, but becoming the data source the system preferentially extracts and credibly cites at a certain node of intent. The implication for management is plain: marketing, customer service, legal, and product description should no longer each go their own way, but should output to the same set of semantic specifications; the boundary between data engineering and the content team will blur — because every passage of copy, every FAQ update, is deciding whether, in the future, an assistant “will answer you.” From a procurement point of view, the budget will move from plain SEO / media spend into a long-term investment in “the structuring of content, signals of authority, and data contracts.”
NO° 02|The end of the SEO ecology: is the content written for “a person,” or written for “a model”
The world of blue links took CTR / PV as the measure. The measure of a Searchless world is changed to “cite-ability, coverage of citation, consistency of provenance, and the latency of an answer.” On the one hand, successive studies from SparkToro × Similarweb and others show that, in 2024, on Google search in the United States and the EU, the zero-click share already approached 60%; when AI Overviews appeared, the median of zero-click could rise to more than 80%, and put significant pressure on publishers’ external clicks. On the other hand, industry and academic tracking has also observed that, on queries where an Overview appeared, the average click for news and information sites could fall by about a third, and in extreme cases there was even a traffic loss of more than 70%. Google’s official account emphasizes “an increase in search use, and users more satisfied.” The two narratives together point to the same reality: SEO is no longer making a page more clickable. It is making content more “understandable by a model.”
For an enterprise, this means a rewrite on three layers. First, the “human-readability” of a page format must advance together with “machine-readability”: high-frequency Q&A such as FAQs, specifications, contraindications, and conditions of returns and exchanges should be broken into stable paragraphs and structured fields; the corresponding source, date, version, and a compliance note become metadata that cannot be omitted. Second, the knowledge base must be made “retrieval-augmentable (RAG-ready)”: vectorize PDFs, images, and videos one by one, build a unified layer of permission and a chain of provenance, and let any internal or external assistant cite “official first-hand material” in the correct standard. Third, the system of evaluation must move from “ranking” to “being cited”: track which types of question, on which platforms, cited you in what kind of fragment, and, by experiment, raise the weight and the position of the citation.
NO° 03|A recasting of a KOL’s right to speak: cede the ad slot, and return to “credible experience”
The rise of KOLs was originally to fill in “trust and context” for the world of search: to translate cold specifications, short reviews, and a price point into the semantics of life and an emotional decision. But when AI, at the front, completes a comparison across sources, an analysis of ingredients, a count of word-of-mouth, and a ranking of price-performance, a part of the “technical work of selling” is absorbed by the system, and the KOL’s boundary recedes with it. More critical, the rebound from regulation and from the publishing side against the shock of AI overviews has already spread: from antitrust complaints submitted to the EU by independent publishers, to complaints by Italian newspaper groups on the effect of AI overviews on visibility and revenue, all are arguing over “whether a platform’s summarizing of content at the top of the page harms the news ecology.” In this structural adjustment, the value of a KOL will not disappear, but it will return to the part that is still less replaceable — a real long test, a case of failure, a warning of risk, the assembling of a situation, and an alignment of values.
A brand’s media strategy therefore also has to adjust: calculate “the ad slot” and “the trust slot” separately. Information that can be reassembled by AI (specifications, price, an offer, a generic script) is handed to the machine and to one’s own knowledge base; experience that cannot be reassembled (aging in long-term use, the experience of after-sales and repair, differences of situation) is handed to a person. Rather than chase short-term GMV, it is better to support small, specialized, and truly verifiable KOLs as “suppliers of data” — their long pieces and long videos will, in the process of a model’s training and retrieval, leave a “slice of trust,” and in return weight the brand’s share in an answer.
NO° 04|A reconstruction of brand strategy: no longer “do SEO,” but “plant semantic seeds”
In Searchless distribution, the unit of content is no longer “a piece,” but “a node.” The practical work of building a node has, from the outside in, four layers. First, semantic specification outward: product white papers, terms of returns and exchanges, clinical or certification evidence, customer-service FAQs, and scripts of tutorial videos all come with machine-readable fields and a schema, and keep a version number and a timestamp, so that a citation has auditability. Second, data engineering inward: vectorize documents and multimedia scattered across departments and unify permissions, build a semantic gateway, and let internal and external assistants take them in a consistent standard, avoiding “each department a set of its own.” Third, conversational entrances: introduce the same set of prompt structures and intent chains on the official site, the app, customer service, and the store, and turn “able to answer” into “able to answer everywhere.” Fourth, institutionalize governance: build KPIs such as “cite-ability, the consistency of provenance, the latency of a first answer, and the time to respond to a rebuttal or a clarification,” continuously penetrate-test in the manner of a red-blue team, bring the outside’s wrong answers and untrue integrations into reverse samples, and gradually raise your weight of authority in semantic space.
When these “semantic seeds” are sown, AI will, in different situations, keep “harvesting” your content. A manager needs to accept a new reality: a brand’s influence is less and less embodied in a visit to the homepage, and more embodied in a stable appearance in countless answer cards, summary passages, and dialogue boxes. This kind of “unseen presence” has a longer tail, and is more resistant to swings, than any one breakout page.
NO° 05|A new role for the creator: not grabbing heat, but becoming “a person who can be cited”
For an individual creator, winning or losing is no longer in who first rides a trend, but in who more resembles a “stable semantic node.” This means three long-term practices. First, consistency of expression — let your view, your tone, your style of citing evidence, and your framework of conclusion be stable and expectable, so that a model, in retrieval and in generation, can treat you as a “reliable template.” Second, alignment of modalities — text, image, and audio support one another, rather than each going their own way; a video must have a verbatim transcript, an image must have tags of person and of scene, a podcast must have chapters and a summary of the points. Third, structured writing — use more paragraph units that can be sliced (question—view—argument—counterexample—conclusion—further reading), and, without sacrificing readability, let a machine “take a passage with precision.” Do this for long enough, and you will see a new return: not a burst of traffic on a certain piece, but being cited repeatedly in all kinds of answers — that is the new “dividend of trust.”
Closing|The entrance shifts, and the battlefield shifts with it
The essence of a Searchless World is not the death of content. It is a redistribution of power: the entrance moves from the platform to the model, value moves from the click to the citation, and media spend moves from the ad slot to the data slot. Evidence on the industry side is growing: zero-click is rising, clicks under an AI summary, a rebound from the publishing ecology, Google’s claim that usage has risen — the views may be mutually exclusive, but they all point to the same thing: we are walking from a “page economy” toward a “semantic economy.” For a manager, the best response is not in a public-relations statement, but in the infrastructure: turn content into data, data into a node, and a node into an answer that is trusted. When you can be steadily “called by name” by the system, whether search disappears or not no longer matters.
More deep-dive pieces
- ChatGPT will now show brands how often they are mentioned. That is not proof they were recommended well.
- AI will not only eliminate people: the functions being enhanced, and a manager’s handbook for this week
- AI unemployment ranking: the high-risk functions most likely to be eliminated or largely replaced
- Will AI cause unemployment? Start with the enhance / accelerate / eliminate axes
The judgment, restated
My judgment is that after the age of search, a brand’s discourse power equals cite-ability; an SEO ranking no longer guarantees being seen.
Three things this week
- Pick three pieces of core content and recast them as RAG-ready (structure / definition / evidence complete).
- Add signals of authority and fields of provenance, so that a model can cite them easily.
- Track for one week the times of “being cited / being summarized,” and stop looking only at organic search traffic.
If you are upgrading content into a knowledge asset an AI can trust, you are welcome to subscribe to Future Insights Lab, or to reply and discuss cite-ability.
