Websites Didn't Die; the Old Traffic Path Did

In a discussion titled "the concept of the website is dying," the title is what draws the argument and the shift behind it is what is worth keeping: apps compete for the entrance, platforms reduce outbound links, users increasingly search and consume inside a platform, and the old path of walking from one page to another along hyperlinks keeps thinning.

Reducing that to "websites are dying" is inaccurate. The accurate version: the open web no longer monopolizes information discovery, and a search engine no longer guarantees that one information need becomes one website visit.

That distinction decides how SEO should be done for the next few years.

Separate four things first

When someone asks "platforms like Xiaohongshu get indexed too, so why call this privatization of information," at least four layers need separating:

1. Whether a search engine can discover a URL.
2. Whether the page content can be fully crawled and indexed.
3. Where a user searches first when the need arises.
4. Whether, having got an answer, the user still needs to open an independent website.

The first two are technical visibility. The last two are traffic allocation.

So even when a page on a platform is discoverable and indexed by Google or Bing, that does not imply platformization leaves independent sites unaffected. The user may open the platform directly and finish discovery inside its search, recommendation feed, or follow graph — the search never passed through Google at all.

The other case: the user still uses Google, and the results page already solves the problem. The old chain was query → blue link → website → answer. Increasingly it becomes query → summary or AI answer → check the source only if needed.

Being indexed is not being distributed, and being cited is not being clicked.

Privatization is not only "crawlers cannot see it"

"Privatization" is easily read as login walls, robots.txt, or in-app pages that cannot be crawled. That is one case at the technical layer.

Two other layers deserve more attention.

Distribution privatization. Content may be public while who sees it is decided mainly by platform recommendation, in-platform search, account relationships, and commercial rules. A search engine being able to fetch the page does not mean the search engine still holds the main entrance.

Relationship privatization. The classic structure of the open web is sites linking to each other, so users and crawlers can discover new sites by following links. Platform-shaped internet looks more like recommended content → stay on the platform → more recommended content. External links do not necessarily disappear, but they move from the default path to the exception.

The original post called apps a set of islands. Slightly absolute, and it catches the core: a platform does not need to make information unsearchable. It only needs to remove the user's reason to leave.

Which is why "can this platform's page be indexed by Google" is not a sufficient test of the open web's health. The question is who holds the power of discovery and the power of traffic allocation.

AI adds another layer: the search stays, the click may not

In-platform search changes where people search. AI search changes whether they need to enter a website afterwards.

Pew Research Center's analysis of US adult browsing in March 2025 found that when a Google results page carried an AI summary, roughly 8% of users clicked a traditional search result, versus roughly 15% when it did not. That is not a permanent click rate for every market and query, and it indicates a direction: the more complete the answer layer, the fewer external clicks are likely.

Meanwhile Google has not removed websites from the system. Its 2026 generative AI search guide states that AI Overviews and AI Mode still rest on core Search ranking, quality systems, and the search index, retrieving pages through RAG and query fan-out. Google's advice is not to invent a mysterious "AI SEO" but to keep doing sound SEO and to provide unique, valuable, non-commodity content.

Bing's move is more direct. In February 2026, Bing Webmaster Tools launched AI Performance, showing site owners their content's citations, cited pages, and grounding queries in AI answers. Microsoft defines the change as: visibility is no longer only rankings and clicks, but also whether content enters answers, citations, and grounding.

Google followed in June 2026 with generative AI performance reports in Search Console, pulling AI search visibility out as its own view.

Taken together, the signal is clear enough: SEO has not disappeared, and "ranking → click" is no longer the complete outcome chain.

Which sites are actually under pressure

The compression hits commodity informational content first.

If a page merely reorganizes public material into "what is X," "how to do Y," "top 10 Z," with no proprietary data, tool, experience, product, or judgment, both search engines and answer engines have strong incentive to compress it. A user does not necessarily need to open the original page to acquire common knowledge.

A few kinds of content are much harder for the answer layer to absorb:

• Original testing, experiments, measurements, and first-hand data.
• Calculators, lookups, configurators, and other interactive tools.
• Databases, directories, and continuously maintained references.
• Product documentation, API docs, changelogs, and explicit sources of fact.
• Specialist B2B content, complex comparisons, and high-stakes decisions.
• Community discussion, real cases, and user experience that cannot be mass-replicated.

Those pages will not win a click every time, and they stand a better chance of being the source a search engine or AI system has to cite.

This is the next step along the line of GEO is not SEO for ChatGPT: if the answer engine reads the page on the user's behalf, a site's competitiveness is no longer "writing like an answer." It is having something the engine cannot generate on its own.

Content strategy is moving from writing answers to manufacturing sources.

Do not swap "build a website" for "platforms only"

The other easy extreme: since entrances are fragmenting, abandon the independent site and put everything on YouTube, Reddit, TikTok, LinkedIn, or elsewhere.

That narrows the problem just as badly.

Platforms are good at distribution. An independent site is good at being a stable source of fact and an asset center. They do different jobs.

For a long-running site aimed at overseas markets, the sane structure:

Layer	Primary role
Google / Bing	Capture explicit search demand; provide web and AI search visibility
Community / video / social	Discovery, discussion, recommendation, brand diffusion
AI answer engines	Aggregate information, cite sources, assist decisions
Independent website	Canonical source, product, documentation, conversion, first-party assets
Email / direct	A continuing relationship not entirely dependent on algorithmic distribution

SEO used to be treatable as a standalone channel: pick keywords, win rankings, harvest clicks. The better framing now puts the website at the center of an information supply chain and lets different entrances consume and distribute from it.

What I expect over three to five years

I would not take "the death of the website" as the base case. A website is still the most stable addressable unit of content on the open internet, and an important source for search indexes, AI grounding, product transactions, and official brand information.

But part of the traditional SEO business model will keep shrinking: producing volumes of substitutable informational articles, intercepting search traffic, monetizing through ads or affiliate links. Those pages are easy to replace with in-platform content and easy to compress into an AI answer.

What is worth accumulating is not article count but four kinds of asset: proprietary information, verifiable evidence, executable tools, and a stable brand entity. Technical SEO still makes those assets discoverable, indexable, and understandable; content and product answer a different question — why should a search engine, an AI, or a user come back to you rather than pick any comparable page?

So my read is neither "SEO is dead" nor "nothing changed."

What actually changed: SEO is expanding from competing for a position in a results page to competing for source status across the whole discovery chain.

Rankings still matter. Clicks still matter. They are simply no longer the whole of it.

What is worth doing now

Planning a new overseas site today, I would prioritize five things:

1. Define what proprietary information the site can keep producing, before listing 500 keywords.
2. Keep important facts, data, tests, product information, and documentation on URLs you control, and distribute summaries and discussion to platforms.
3. Keep doing crawling, indexing, internal linking, canonicals, structure, and update mechanics well — AI's arrival is not a reason to abandon basic SEO.
4. Add brand search, cited pages, AI visibility, direct traffic, and final conversion to the metrics alongside rankings and organic clicks.
5. Stay restrained about "GEO tricks." Build non-commodity content and evidence first; format-level optimization comes after.

If those five hold, platformization and AI search do not strip a website of value. They change how it gets found and used.

Boundaries

This discusses structural change for informational sites and overseas SEO, not a uniform prediction for every site, country, and query. Mainland China's super-app ecosystem, the overseas open web, Google/Bing search, and different AI products still differ significantly, and platform crawling rules and AI presentation will keep changing.

Pew's click data comes from a US adult sample in a specific period. It supports the statement that an observable difference exists between AI summaries and click behavior; it cannot be extrapolated into a fixed global SEO click rate. The official Google and Bing material confirms that search, AI indexing, and citation tooling are converging; it does not prove that all users will stop visiting websites.

One thing is reasonably certain: judging a website's value can no longer stop at "can search engines index it." The follow-up questions are where users discover it, why systems cite it, and what still has to happen on that site and nowhere else.