Connect, Don’t Query: A Manifesto for Rebuilding the Open Web as a Human Link Network

Full report

On this page

White paper on the viability, philosophy, and strategic rollout of a grassroots movement for decentralized discovery

The central argument of “Connect, Don’t Query” is stronger when stated precisely: the Web itself has not ceased to be decentralized, nor was the early Web literally a peer-to-peer protocol. HTTP has always been a client-server request/response protocol. What has changed is the dominant architecture of discovery. The Web was conceived as a “web of nodes” through which a person could follow links from one piece of information to another; contemporary discovery is increasingly mediated by centralized indexes, ranking systems, recommendation engines, and now generative answer layers.

That distinction matters. “Connect, Don’t Query” should not claim that search invented centralization or that search has no legitimate function. Search solved an essential scaling problem: once the Web became too large to traverse manually, indexes made billions of documents findable. Google itself began by exploiting the Web's existing hyperlink structure rather than replacing it. But the social consequence of making a centralized query box the default doorway to a decentralized publishing system was profound: the power to publish remained distributed while the power to be discovered became concentrated. Google’s current documentation still describes precisely this overlay: it crawls decentralized pages, stores information in a large index, and decides what to serve in response to a query.

Generative AI intensifies that asymmetry because the intermediary is evolving from a directory that points elsewhere into an answering system capable of satisfying the request without sending the user elsewhere. In observed Google usage, AI summaries have already been associated with markedly fewer outbound clicks, while AI companies have in some cases crawled publishers vastly more often than they send human referrals. Google, meanwhile, described its May 2026 changes as its largest Search-box upgrade in more than twenty-five years and reported more than one billion monthly AI Mode users.

The movement therefore has a plausible and urgent mission, but its strategically viable objective is not literally to “starve” search engines out of existence. Hyperlinks themselves help search engines discover and understand the Web; creating more links can simultaneously strengthen the open Web and improve centralized crawlers. The achievable objective is more interesting:

Rebuild enough direct, human-maintained discovery infrastructure that search stops being the mandatory intermediary between a reader and the independent Web.

Under that interpretation, “Connect, Don’t Query” is less an anti-search campaign than a program for disintermediation of discovery: publish at durable origins, link generously, subscribe directly, curate publicly, move audiences laterally between independent sites, and make the human-authored hyperlink graph socially important again.

The Architectural Shift

The founding metaphor of the Web was not a database to be interrogated but a graph to be traversed. Tim Berners-Lee and Robert Cailliau's early proposal described hypertext as information connected into a “web of nodes” in which a user could browse at will, explicitly contrasting this structure with a hierarchical tree or ordered list. The Web's transport protocol subsequently standardized HTTP as a stateless application-level protocol for distributed, collaborative hypertext information systems.

Calling that architecture “peer-to-peer” is therefore slightly misleading. A browser requests resources from servers; two ordinary users' machines do not have to exchange documents directly. The important decentralizing property lies elsewhere: there is no required central publication registry, universal content database, or single ranking authority in the basic Web architecture. Indeed, Google itself notes that there is no central registry of Web pages, which is why its crawler must discover URLs through links, sitemaps, and prior knowledge.

In graph-theoretic terms, the original discovery experience can be represented as:

     human recommendation
A ───────────────────────► B
│                          │
│                          │ contextual link
│                          ▼
└──────────────► D ◄────── C

Each site owns its node.
Each publisher chooses its outgoing edges.
Readers traverse the graph.
No universal ranking service is required.

A hyperlink in this model performs several functions simultaneously. It is a routing instruction for the reader, an editorial recommendation by the author, an attribution mechanism, and a new directed edge in the global graph. Early large-scale search systems recognized the informational value of exactly this structure: Brin and Page's 1998 description of Google said that the system made heavy use of hypertext structure and maintained both full-text and hyperlink information.

The breakthrough of search was therefore initially parasitic in the neutral ecological sense: search derived intelligence from connections that other people had already created. The Web's distributed editorial choices became input signals for a centralized retrieval system. That was extraordinarily useful because manual traversal does not scale to exhaustive discovery. It also created a new architecture:

Sites A ─┐
Sites B ─┼──► Crawler ─► Private index ─► Ranker ─► Query interface ─► User
Sites C ─┤
Sites D ─┘

The difference is not where documents live. They can remain on millions of independently administered origins. The difference is where discovery decisions are computed. In Google's documented pipeline, pages are crawled, analyzed into a large index, and then programmatically selected in response to user queries; Google also states that meeting its technical requirements does not guarantee that a page will be crawled, indexed, or served.

That produces what might be called discovery-layer centralization. Publication remains relatively decentralized, while visibility is increasingly allocated through an intermediary's index, models, ranking signals, interface design, commercial incentives, and distribution agreements. The distinction between decentralized storage and decentralized attention is fundamental: a million independent websites do not constitute a meaningfully decentralized media system when most readers approach them through one or two ranking surfaces.

This concentration is no longer merely a critical-theory characterization. In the U.S. search antitrust litigation, the District Court concluded in August 2024 that Google was a monopolist that had acted to maintain its monopoly; the September 2025 remedies restricted certain exclusive distribution arrangements and required access to specified search-index and interaction data for competitors. The Justice Department said Google had accounted for approximately 90% of U.S. search queries for years. Separately, in April 2025 a federal court held that Google had monopolized open-web digital-advertising markets, with the court finding harm to publishing customers, competition, and consumers of open-Web information.

For the independent creator, the structural transformation can be summarized this way:

The movement's diagnosis is therefore substantially correct but should be phrased carefully. Search did not destroy the hyperlink graph; it became a dominant interface for consuming it. And because search crawlers themselves discover pages through hyperlinks, the open graph survives underneath the centralized index.

That distinction changes the political strategy. A successful open-Web movement does not need to invent a replacement Internet protocol. It needs to change the traffic pattern on protocols that already work.

The most useful abstraction is:

\[ \text{Open-web resilience} \propto \frac{\text{direct links + direct subscriptions + independent referral paths}} {\text{dependence on mandatory discovery intermediaries}} \]

This is not an empirically established equation; it is a design heuristic derived from the architectural distinction above. Its point is that decentralization should be measured not merely by the number of independent servers but by the number of independent paths through which one node can lead a human to another.

There is an important warning from network science as well: decentralization does not automatically produce equality. Barabási and Albert's classic network model identified growth and preferential attachment to already well-connected nodes as mechanisms capable of producing highly unequal degree distributions in networks including the Web. In plain language, a human-curated hyperlink network can still generate super-hubs, incumbency advantages, cliques, and obscure peripheries.

“Connect, Don’t Query” must therefore defend distributed authority, not romanticize a frictionless egalitarian past that never existed. Its goal should be a Web with many overlapping paths, many curators, many hubs, and low switching costs—not a Web without hubs at all.

The AI Threat Multiplier

Traditional search created an intermediary but normally preserved a crucial economic and cultural bargain:

crawl → index → show result → send click

That exchange was never perfectly balanced, but the destination remained structurally important. Search results functioned as pointers. A publisher created the document; a search engine helped a user locate it; a click transferred the user's attention to the publisher's origin. Google's conventional Search architecture still formally describes crawling, indexing, and serving relevant information in response to queries.

Generative interfaces introduce another possibility:

crawl → retrieve → synthesize → answer → perhaps cite → perhaps no click

The word perhaps is decisive.

Pew Research Center's study of browsing behavior from 900 U.S. adults examined 68,879 unique Google searches, including 12,593 that produced AI summaries. In those observed sessions, users clicked a traditional search result in 8% of visits when an AI summary appeared, versus 15% when one did not. Only 1% of visits with an AI summary resulted in a click on a source cited in that summary. Users also ended their browsing sessions after 26% of pages containing an AI summary, versus 16% for pages with conventional results only.

Those findings do not prove that every AI answer causes traffic loss; query type and user intent matter, and Pew's observational design should not be interpreted as a randomized causal experiment. They do, however, demonstrate the central risk empirically: a synthesized answer can satisfy enough of the user's informational demand that the source ceases to be the destination.

The threat becomes more visible when crawler consumption is compared with referral output. Cloudflare calculated that in June 2025 Google crawled roughly 14 HTML pages for every referral attributable to Google. It estimated ratios of approximately 1,700 crawls per referral for OpenAI and 73,000:1 for Anthropic, while explicitly cautioning that missing Referer headers from native apps can affect such measurements. Cloudflare characterized the pattern as a breakdown of the older “crawl in exchange for traffic” relationship.

The numbers should therefore be treated as telemetry from a major infrastructure provider, not universal Internet-wide constants. But the architectural point does not depend on the exact ratios. An AI system can derive marginal utility from large amounts of publisher content while requiring comparatively few users to visit each source.

That changes the media economy.

Under classic Web browsing:

\[ \text{value of source} \rightarrow \text{reason to visit source} \]

Under generative mediation:

\[ \text{value of source} \rightarrow \text{input into intermediary} \rightarrow \text{value of intermediary's answer} \]

The intermediate step can detach informational value creation from audience acquisition. This is an inference from the observed reduction in outbound clicking and the crawler/referral imbalance, rather than a claim that every generative product behaves identically.

That is why AI is a multiplier rather than merely another search feature. Traditional search centralized selection. Generative search can centralize selection, interpretation, summarization, and presentation.

Google's own trajectory illustrates how rapidly this interface model is expanding. On May 19, 2026, Google described a new AI-powered Search box as its biggest Search-box upgrade in more than twenty-five years. It said AI Mode had surpassed one billion monthly users, that queries in the mode had more than doubled each quarter since launch, and that Search was adding conversational follow-ups and agents capable of continually scanning blogs, news sites, social posts, and other Web information to produce synthesized updates.

For the independent Web, that creates four distinct pressures.

The referral problem is the most immediate. A useful article can contribute to an answer while receiving no visit. Pew's measured click gap shows why zero-click behavior is not simply a theoretical concern.

The attribution problem follows. Even where citations exist, citation is not equivalent to audience transfer. A footnote-sized source indicator does not reproduce the social encounter of visiting a creator's site, seeing related work, subscribing to a feed, joining a community, or financially supporting the author. Pew's 1% observed click rate on AI-summary citations makes this distinction concrete.

The context problem arises because synthesis separates statements from the surrounding artifact. A Web page is not merely a bag of propositions; it may contain design, argument structure, caveats, links, authorial voice, comments, archives, related essays, and a community context. Generative retrieval can compress that artifact into answer-relevant fragments. The capacity of AI Search to synthesize across Web sources is explicitly part of the product direction Google described in 2026.

The dependency problem is recursive. If independent sites receive less direct discovery, creators have stronger incentives to format material for whichever systems still deliver attention. That can move editorial labor yet another step from creating a destination worth visiting toward making information extractable, rankable, citeable, and model-readable. This is an inference from the documented crawler/index/answer architecture and referral evidence, not evidence that all creators have already made that transition.

This is the strongest intellectual foundation for “Connect, Don’t Query.”

A query says:

Find or produce the answer for me.

A connection says:

This source is worth establishing a durable relationship with.

The former delegates discovery to a computational intermediary. The latter makes the reader part of the topology.

Yet the movement should resist declaring AI inherently illegitimate. Search and language models solve real retrieval problems that manual curation cannot: obscure factual lookup, cross-domain synthesis, unfamiliar terminology, accessibility assistance, exhaustive retrieval, and navigating enormous corpora. The movement's coherent target is therefore not computation but dependency.

Its objection should be:

No discovery intermediary should become so convenient that society stops maintaining the independent pathways from which the intermediary itself derives value.

That is the AI-era version of the tragedy of the commons: everyone can rationally prefer the easiest answer interface while collectively allowing the source ecosystem underneath it to become poorer.

Psychological Reprogramming

The most consequential element of “Connect, Don’t Query” is not its technical proposal. RSS already exists. Hyperlinks already exist. Personal domains already exist. Webmention already exists. WebSub already exists. The missing layer is habit. Atom is standardized specifically for syndicating Web content directly to user agents; WebSub defines a publisher-hub-subscriber update flow; Webmention lets one URL notify another that it has been linked. The primitives for a less intermediary-dependent social Web are not hypothetical.

The cultural problem is that users have been trained to interpret navigation as a retrieval problem.

When a person wants information, the habitual sequence is increasingly:

desire → query box → ranked/synthesized answer → consumption

“Connect, Don’t Query” proposes:

encounter → follow → evaluate → save/subscribe → return → link onward

The difference is sociologically significant because the second loop creates durable structure.

A query is usually ephemeral. Its output answers the present need.

A bookmark is memory.

A feed subscription is relationship.

A blogroll is public recommendation.

A hyperlink embedded in a published page is infrastructure.

A Webmention can turn that hyperlink into an interoperable notification between independently operated sites.

This is the core psychological reframe: browsing is not merely consumption; navigation choices can constitute maintenance work on a commons.

Every time a publisher creates an ordinary link from site A to independent site B, a new machine-readable edge becomes available to human readers and crawlers. Google explicitly says links from known pages are one way it discovers new URLs, demonstrating that links remain part of the Web's operational discovery infrastructure rather than decorative typography.

That provides the movement with a powerful concept: structural participation.

Conventional digital activism often asks people to express themselves inside a platform—post, react, reshare, follow. Structural participation asks them to alter the network itself:

Own a node.

Create an edge.

Maintain a route.

Subscribe to a source.

Send a reader outward.

The language aligns with psychological mechanisms identified by Self-Determination Theory, provided the connection is treated as an interpretive application rather than proof that hyperlinking itself improves well-being. The theory identifies autonomy, competence, and relatedness as core psychological needs: autonomy concerns volition, competence concerns effectiveness and mastery, and relatedness concerns connection and significance to others.

“Connect” can symbolically activate all three:

This should not be oversold as a clinical effect. The useful point is motivational design: a movement that makes users feel that a small action creates visible infrastructure is more likely to produce a sense of agency than one whose message is merely “technology companies are bad.” Self-Determination Theory associates need-supportive contexts with higher-quality motivation and engagement, while the IndieWeb community's own principle-design discussions similarly emphasize actionable, positive, memorable guidance.

That insight reveals both the brilliance and the problem in the slogan.

“Connect” is a positive instruction. “Don’t Query” is a prohibition.

The IndieWeb community's documented methodology explicitly argues that principles should state what to do, favor positive direction over negative inaction, and use short, memorable phrases. “Connect, Don’t Query” satisfies the memorability requirement and gives the affirmative verb first, but its negative half can become psychologically counterproductive if interpreted literally.

The movement should therefore teach a behavioral interpretation:

Before you query again, ask whether you could create a connection that makes the next query unnecessary.

Examples make the psychology concrete.

You discover an excellent independent analyst through search. The query was useful. The movement's intervention occurs after discovery: do not repeatedly return to Google to rediscover the analyst. Subscribe to the analyst's feed.

You read a useful essay and later recommend its argument. Do not tell people “Google the title.” Publish or send the direct URL.

You maintain a personal site. Do not make visitors restart discovery at a search engine after reading you. Give them contextual links to adjacent independent work.

You follow thirty writers. Do not keep their names in your head and search for them periodically. Put their feeds into a reader.

You know five outstanding sites in your specialty. Do not keep that topology private. Publish a directory or blogroll.

A blogroll is explicitly described by the IndieWeb community as a list of sites one reads, follows, or recommends, and it can expose feeds or an OPML file so visitors can import subscriptions.

That converts taste into public infrastructure.

The essential cultural shift is from:

“The Internet is a service I ask for things.”

to:

“The Web is an environment whose paths I help maintain.”

That is a much more radical proposition than installing a new browser extension. It restores responsibility to the user without pretending that every user must become a systems administrator.

Strategic Viability of the Slogan

As a battle cry, “Connect, Don’t Query” is unusually strong because its two words operate simultaneously at technical, social, and moral levels.

“Connect” is both a network verb and a human verb. It can mean connecting nodes by hyperlink, establishing a feed subscription, connecting one reader to another author, or reconnecting independent sites into communities. This semantic compression is strategically valuable because the movement is fundamentally trying to make social behavior and network architecture feel like the same act.

“Query,” by contrast, is deliberately colder. For technically literate audiences it evokes databases, SQL, search boxes, API calls, and request/response systems. A query treats the information universe as a corpus awaiting extraction. A connection implies continuing relation.

The opposition therefore contains an entire media theory:

\[ \textbf{Query:}\quad \text{user} \rightarrow \text{intermediary} \rightarrow \text{answer} \]

\[ \textbf{Connect:}\quad \text{person} \leftrightarrow \text{person/site} \leftrightarrow \text{person/site} \leftrightarrow \dots \]

That makes the phrase especially well suited to software engineers, bloggers, researchers, digital minimalists, RSS users, and IndieWeb communities because it is simultaneously interpretable as a technical instruction and a cultural critique. The IndieWeb's own movement-design discussions argue that actionable, short, catchy phrases are particularly useful because they are easier to remember and can make principles more accessible beyond programmers.

Its syntax is also efficient. Two monosyllabic imperative verbs are placed in direct antithesis. The comma functions almost like a rewrite operator:

QUERY  →  CONNECT

It does not require the listener to memorize a policy platform. It asks for substitution.

The slogan also benefits from being pro-social rather than nostalgic. “Bring Back the Old Web” would define the project as restoration, inviting accusations of nostalgia for an era that was smaller, less accessible, and hardly free of concentration. “Connect, Don’t Query” makes no claim that 1997 was ideal. It argues that one architectural primitive—direct connection—should regain importance now.

Its principal weakness is exactly the same feature that gives it force: “Don’t Query” is false when read literally.

Queries are indispensable for finding something whose location you do not know. Search became useful precisely because the Web is too large for exhaustive manual traversal; Google's original search work explicitly addressed the difficulty of scaling information retrieval across an uncontrolled hypertext collection.

A campaign that tells ordinary users they must browse thousands of hand-curated pages instead of searching for a tax form, medical paper, API reference, obscure historical detail, or unfamiliar concept will lose immediately.

Worse, eliminating search would not automatically decentralize attention. Human hyperlink networks naturally create hubs, and preferential attachment can increase the prominence of already well-connected nodes. A pure curation culture can also reproduce social homogeneity, clique formation, status hierarchies, language barriers, and blind spots.

The slogan therefore requires an official doctrine that distinguishes lookup from belonging:

Query for the unknown. Connect to what proves valuable.

Or, as a secondary explanatory line:

Search when necessary. Subscribe, link, and return at the source.

This preserves the sharper primary slogan while preventing ideological overreach.

The other strategic weakness is the movement's proposed rhetoric of “starving search monopolies.” That objective is emotionally legible but technically confused. Search engines discover pages partly through links, and Google's founding architecture explicitly exploited hyperlink structure; increasing the number and quality of open hyperlinks can therefore make the corpus easier for a search engine to crawl and interpret.

The movement's actual leverage is not deprivation of crawlers. It is reduction of behavioral dependence.

That changes the success criterion from:

Did Google crawl less?

to:

How much discovery happened without requiring Google?

That is strategically much stronger because it measures something participants can directly alter.

A useful viability scorecard is therefore:

The legal and market context makes the critique of gatekeeping particularly legible. U.S. courts have found unlawful monopolization in Google's search and open-Web ad-tech businesses, while remedies in the search case reached distribution arrangements and specified data access. The movement does not need to establish that concentrated gatekeeping exists from scratch; institutions have already documented severe concentration.

What it does need to demonstrate is that a grassroots intervention can matter.

The strongest theory of change is not top-down substitution:

Google dies
      ↓
open web returns

It is bottom-up redundancy:

more direct links
      ↓
more human discovery paths
      ↓
more feed subscriptions / bookmarks / repeat visits
      ↓
more creator-reader relationships independent of ranking
      ↓
less existential dependence on any single discovery surface
      ↓
greater resilience of independent publishing

This model is compatible with continued use of search. That is an advantage, not a compromise.

A resilient decentralized system does not need to destroy centralized services. It needs to make them non-mandatory.

Actionable Implementation

The movement will fail if it remains a discourse about the Web. Its unit of success must be a changed routing behavior.

The practical doctrine can be expressed in one sentence:

Own your origin, subscribe at the source, link at the point of reference, expose your curation, and send people outward.

Each component is implementable with existing Web standards. The IndieWeb community explicitly prioritizes an identity and content presence controlled by the user, while Atom, RSS, WebSub, and Webmention provide interoperable mechanisms for syndication, update delivery, and cross-site notification.

The individual behavior stack

The minimum viable participant does not need to run servers or write software. The movement should establish a ladder in which every level is useful.

Atom's standards-track specification identifies direct syndication to user agents as a primary use case. RSS supports feed items with durable item links and even mechanisms for update notification, while WebSub formalizes a publisher/topic/hub/subscriber model.

For publishers, feed autodiscovery should be treated as basic infrastructure:

<link
  rel="alternate"
  type="application/atom+xml"
  title="Site feed"
  href="/feed.xml">

Atom defines application/atom+xml, feed documents, entries, and links; the purpose is to make regularly updated Web content available for syndication directly to user agents.

The movement should treat RSS/Atom subscription as the functional replacement for the “follow” button. The conceptual difference is ownership: the publisher exposes a standard feed endpoint, while the reader chooses an independent client and maintains the subscription list. No universal recommendation service is required by the protocol.

A blogroll should similarly be reframed from retro ornament into human-readable routing infrastructure. The IndieWeb community describes blogrolls as lists of sites a person reads, follows, or recommends and notes that they can expose feed links or OPML so another reader can import subscriptions.

A high-quality modern blogroll should therefore not be fifty unexplained domain names. It should look more like:

Distributed systems
  → Alice — unusually clear writing on storage engines
  → Bob — long-running notes on network protocols

Independent media
  → Site C — local investigative reporting
  → Site D — excellent media criticism

Small-web discovery
  → Directory E — curated personal sites

The annotation is crucial. Search engines rank; humans should recommend. A sentence explaining why this link deserves traversal supplies the context that algorithmic ranking usually obscures.

Webrings can provide a complementary mechanism. A contemporary webring need not recreate 1990s visual conventions. It can simply be a transparently governed membership set whose pages expose navigation such as:

← Previous member   Random member   Next member →

The design objective is not exhaustive search. It is serendipitous lateral motion across peer sites. That directly instantiates Berners-Lee's original idea of users browsing from information node to information node.

Webmention can add a missing social layer. The W3C Recommendation defines it as a mechanism by which one URL notifies another that it has been mentioned; its introductory example explicitly describes one independent blogger responding to another on a separate site and notifying the original. That makes it particularly important to “Connect, Don’t Query”: it allows linking to become not merely outbound navigation but cross-site conversation.

Hyperlinking hygiene as a social norm

The movement should establish a simple editorial ethic:

Link to the thing itself.

When discussing an essay, link the essay.

When referring to a research paper, link the paper or authoritative record.

When crediting a creator, link the creator's own site where one exists.

When mentioning a project, link its project page.

When quoting a statistic, link the underlying study rather than a search-result page about the statistic.

This is not just citation etiquette. It minimizes intermediary hops and preserves the reader's ability to verify provenance directly. The architecture of the Web was designed around links that move users from one information object to another.

A related norm should be:

Never tell a reader to “just Google it” when you already possess the URL.

That phrase outsources a connection you could have created.

Another norm should be:

Every independent page should have somewhere useful to go next.

The open Web weakens when every page becomes a cul-de-sac whose implicit exit route is the browser's search box.

Search as bootstrap, not habitat

The movement needs an explicit search doctrine to avoid fundamentalism.

Use query systems when:

• the destination is unknown;
• breadth or exhaustive retrieval matters;
• the question crosses unfamiliar domains;
• a local human-curated graph has insufficient coverage.

Then, once a valuable source has been discovered, convert retrieval into relationship:

QUERY
  ↓
DISCOVER A GOOD SOURCE
  ↓
VISIT ORIGINAL
  ↓
BOOKMARK / SUBSCRIBE
  ↓
LINK TO IT
  ↓
RETURN DIRECTLY

Search becomes a bootstrap protocol for establishing future connections, rather than the permanent front door through which every recurring relationship must pass.

That is strategically much more credible than abstinence.

The movement rollout

The rollout should begin with people who already understand that the Web is a graph: software engineers, independent bloggers, academics, open-source maintainers, digital archivists, newsletter writers with personal sites, and digital-minimalism communities. The IndieWeb's existing principles and protocols demonstrate that there is already a community vocabulary around owning domains, controlling content, and building interoperable Web practices; the new movement's contribution would be to make discovery itself the focal political issue.

The first wave should be visible practice rather than persuasion. This follows the IndieWeb's own emphasis on actionable principles and “show before tell.” Participants should add feed buttons, “sites I read” pages, direct-source links, outbound recommendations, and webring navigation before producing lengthy denunciations of search.

A recognizable badge can function as a protocol hint rather than merely branding:

Connect, Don’t Query

This site links outward. RSS available. Human-curated Web.

The second wave should standardize portable curation. Public blogrolls should expose ordinary HTML first, because that is universally browsable, while optionally supplying OPML for bulk feed import; the IndieWeb community already documents OPML as a method for making a blogroll's feeds readily subscribable.

The third wave should build federations of curators rather than one master directory. A centralized “official Connect, Don’t Query search engine” would reproduce the architecture being criticized. Instead, many topical directories should cross-link: programming, local journalism, woodworking, experimental music, history, personal sites, climate science, literature, neighborhood organizations, and so on.

The movement should deliberately cultivate overlapping curatorial centers because network science gives reason to expect preferential attachment to otherwise concentrate links around already prominent hubs. A healthy movement would therefore ask major participating blogs not merely to link to one another but to introduce new peripheral nodes into the graph.

The fourth wave should target tools. Feed readers, static-site generators, blogging systems, browser extensions, and CMS themes can make connection cheaper by exposing actions such as:

Subscribe to this site
Add to my directory
Copy canonical link
Recommend on my site
Send Webmention
Explore this site's blogroll
Random site in this webring

This matters because ideology loses against friction. The dominant query box succeeds partly because it reduces discovery to a single, universally understood gesture. A grassroots alternative needs comparable ergonomic simplicity.

Metrics that measure the right victory

The movement should refuse vanity metrics based primarily on follower counts. Its metrics should reveal topological change.

A participating publisher could measure:

Direct Discovery Share

\[ DDS = \frac{\text{direct + feed + independent-site referral sessions}} {\text{all identifiable human sessions}} \]

The exact analytics methodology will vary, but the conceptual target is clear: increase the fraction of readership not dependent on general search or algorithmic social recommendation.

Outbound Domain Diversity measures how many distinct independent origins a site deliberately links to over a period. This discourages a nominally decentralized movement from collapsing into a small clique.

Feed Relationship Count measures direct subscribers where technically available, rather than only social followers.

Human Referral Share measures inbound visits from other independent sites.

Connection Conversion Rate asks what proportion of valuable discoveries become durable relationships—for example, bookmarks, feed subscriptions, direct revisits, or published links.

Peripheral Introduction Rate could track how often a curator sends readers to sites substantially smaller or less connected than itself. That metric directly counteracts the preferential-attachment tendency documented in network research.

The objective is not to prove that querying has vanished. It is to demonstrate that the Web contains more independent routes than it did before the intervention.

Strategic guardrails

The campaign must also avoid several failure modes.

It should not become anti-search absolutism, because search remains indispensable for large-scale retrieval and because Google's original success addressed a genuine scaling problem in open hypertext.

It should not become anti-AI essentialism. The problem is an asymmetric discovery economy, not machine learning as a category. The empirical case is strongest when focused on measurable referral behavior and source/intermediary relationships.

It should not equate decentralization with fairness. Human networks can concentrate attention through preferential attachment just as machine-ranked systems can.

It should not require everyone to become a webmaster. Atom, RSS, WebSub, and Webmention are valuable precisely because software can hide much of the plumbing behind ordinary subscription and publishing interfaces.

It should not encourage link spam. A hyperlink matters because it embodies judgment. Automating thousands of meaningless outbound links would recreate the logic of optimization rather than human curation.

And it should not confuse crawler starvation with human connection. Search engines themselves use links for discovery, so adding hyperlinks may improve both decentralized navigation and centralized indexing. The movement wins when people use those links directly, not when crawlers somehow fail to see them.

The final doctrine is therefore more sophisticated than the slogan's literal wording:

Query is a request. Connection is infrastructure.

Query retrieves value from the network. Connection adds value to the network.

Query asks an intermediary what matters. Connection records your own judgment about what matters.

Query solves the present information problem. Connection reduces future dependence on the intermediary.

The open Web does not require the abolition of search. It requires the restoration of paths that do not begin with search.

The urgency is real. Contemporary search is moving further toward an answer-and-agent architecture: Google reported in May 2026 that AI Mode had passed one billion monthly users and was expanding Search into conversational and continuously monitoring agents. Observational evidence already shows substantially lower outbound clicking when AI summaries appear, and infrastructure telemetry has documented very large crawl-to-referral asymmetries for some AI systems.

But the underlying Web primitives remain intact. HTTP still supports independently operated resources. Links still join those resources. Atom and RSS still permit direct syndication. WebSub can distribute updates. Webmention can connect conversations across sites. Blogrolls can expose human judgment. Personal domains can provide durable origins.

That makes the movement technically conservative and culturally radical. It is not asking the Internet to adopt an untested replacement architecture. It is asking people to use the architecture they already possess differently.

Its most defensible manifesto is therefore:

The independent Web will not survive merely because anyone is allowed to publish. It survives when independent publications are connected strongly enough for people to find, revisit, recommend, and sustain them without asking a gatekeeper for permission every time.

A Web page without outward connections is an island. A reader who depends on a query box for every voyage has delegated navigation. A publisher who links creates a road. A curator who maintains a directory creates a neighborhood. A subscriber creates continuity.

Search when you must. Connect whenever you can.

Connect, Don’t Query.