1 From Being Seen to Being Understood: New Changes in Personal Digital Existence in the AI Era

In human society, “being known” has always been a scarce resource.

Before the advent of the Internet, a person's visibility often depended on whether they were within an existing communication system: it could be coverage by traditional media, the influence of their institution, or interpersonal communication in real society. For the vast majority of ordinary people, even if they possess unique abilities and experience, it is difficult to break through the communication boundaries of the real world.

The network era once allowed many people to see a new possibility: anyone could publish content through the network, allowing their thoughts, experiences, and works to enter a wider world, and forming their own digital existence through long-term accumulation.

But the Internet lowered the barrier to expression without fully solving the problem of individuals being recognized. In the search engine era, the basic unit of information organization is still the web page. The core task of search engines is to help users find relevant content from a large number of pages, rather than determining who the person behind these pages is.

Therefore, a person's digital visibility often depends on whether their content can continuously accumulate network influence through long-term dissemination, external citations, and community discussions. As these pieces of information continue to spread, what people eventually recognize is not just a few articles, but the author behind these articles.

But this path still presents a high barrier for most ordinary people: even if a person has rich experience and unique views, if there is a lack of sufficient communication scale, these contents can easily be scattered across different pages, platforms, and time nodes, making it difficult to break through original network boundaries or to be further associated into a stable personal identity.

In addition to relying on content dissemination to accumulate influence, there was another way to establish a personal identity in the traditional Internet era:Entering a knowledge system with public credit attributes. For example, open knowledge platforms like Wikipedia establish independent entries for some people with public influence through community rules and editing mechanisms. This method is equivalent to a credible knowledge system confirming a person's identity, experience, and achievements.

But this method also has a high barrier—biographical entries usually require sufficient independent and reliable sources, as well as verifiable public influence. In other words, a person often needs to first gain recognition from the public communication system to more easily enter this centralized knowledge system. However, for the vast majority of ordinary people, even with rich experience and long-term accumulation, it is difficult to establish sufficient public influence through traditional means.

But the development of AI is changing this situation. In the process of training and information acquisition, AI models encounter a large amount of public text from the Internet, including open-source communities, technical forums, blogs, documents, and other public materials. For individuals, as long as they leave accessible and citable information on the open Internet over the long term, these contents have the opportunity to become one of the information sources referenced by AI when understanding the world.

Compared to search engines, which rely more on page weight, traffic, and external links, AI's way of judging the value of information is more diverse. This allows some personal content, though limited in scale but possessing long-term accumulation and unique value, to also have the opportunity to be understood and associated.

On the other hand, the way humans interact with Internet information in the AI era is changing: in the search engine era, users needed to actively deconstruct questions, search for pages through keywords, and then read, filter, and organize information themselves; in the AI era, users are increasingly accustomed to directly stating their goals and letting AI help them complete information retrieval, summarization, and analysis.

Therefore, users“ questions have also begun to shift from looking for ”what content is relevant to my question“ in the past to judging ”who is the person behind these contents, and whether this person is trustworthy," for example: Who are the long-term practitioners in this field? Are there trustworthy experts? What are this person's views and experiences?

To answer these questions, what AI needs to process is no longer just a single web page, butinformation scattered across the Internet and its relationships: Only when a person's experiences, works, views, professional fields, and long-term behaviorscan be attributed to the same person, can AI gradually form an overall understanding of this person.

And on this basis, if these pieces of information can further reflect this person's knowledge domain, way of thinking, and long-term practice path, then what AI understands is no longer just “who this person is,” but can further form a complete characteristic cognitive profile of this person.

This means that in the AI era, the way individuals are recognized by the digital world is changing: a person no longer has to become a public figure in the traditional sense, have a large number of followers, or enter a centralized knowledge system to have the opportunity to form a clear public identity. As long as they continuously leave real experiences, unique views, practice records, and stable knowledge accumulation on the open Internet, these scattered digital traces can be re-associated and organized by AI, gradually forming a coherent digital persona about them.

2 In the Platform Internet Era, Why Is It Increasingly Difficult to Form a Personal Digital Persona?

Although the Internet has lowered the barrier to personal expression, the digital traces it leaves behind do not naturally converge into a complete understanding of an individual.

After entering the platform Internet era, this problem has further expanded. Although users leave more digital traces, this information increasingly exists around the platform's own ecology rather than forming a continuous accumulation around the individual: an article, a status update, or a video, even with high interaction and dissemination, often mainly exists within the ecology of the platform to which it belongs. Platforms can record these contents, but this information does not naturally converge into acomplete information system。

The first problem brought by platformization is that it is difficult for personal digital traces to form a unified association: A person may leave a large amount of information across different platforms: expressions on short video platforms, views on social platforms, and life records on image platforms may all show certain facets of the individual. However, because these contents are scattered in the data spaces of different platforms, they do not automatically establish connections with each other, making it difficult to jointly form a continuous personal image.

In addition to the difficulty of associating information, platform mechanisms also affect how users produce these digital traces: Platforms tend to recommend content that is easy to understand, easy to interact with, and can quickly obtain feedback,which in turn shapes the user's way of expression—complex experiences may be compressed into simple tips, long-term thinking may be broken down into short-sentence views, and complete experiences may also be packaged into fragments more suitable for dissemination. What is ultimately presented on the Internet is often just a facet suitable for dissemination, rather than an expression that fully displays personal experience, knowledge accumulation, and ways of thinking.

At the same time, the development of content forms has further reinforced this fragmentation: Short videos, images, and short texts all emphasize instant consumption, which, while lowering the barrier to expression, also makes personal experiences, knowledge accumulation, and thinking processes easier to be broken up into independent fragments. These contents are very suitable for instant dissemination, but are not naturally suitable for forming long-term, continuous personal records.

The deeper problem is that platforms organize information around content, not around individuals: Platforms focus on the dissemination and interaction of individual pieces of content, rather than an individual's experiences, what they focus on in the long term, and how their views change at different stages. Therefore, even if the information left behind by an individual grows, it often still exists as mutually independent content, making it difficult to further crystallize into a continuous understanding of this individual.

This also means that if we want these digital traces to truly form a meaningful long-term accumulation for an individual, we need a carrier that can continuously host content and connect these pieces of content with each other. It must not only record individual content, but also reflect long-term accumulation, thematic correlation, evolution of views, and practical paths.

Therefore, as AI gradually becomes the new entry point for information, a new question arises:If an individual hopes to let future intelligent systems truly understand them, in what form should this personal long-term accumulation exist on the Internet? This also becomes the starting point for rethinking the value of personal websites in the AI era.

3 Personal Blogs: A New Carrier of Personal Digital Assets in the AI Era

If the AI era requires personal long-term accumulation to be better understood and connected, then the next thing to consider is: what kind of information carrier can host this long-term accumulation.

From this perspective, personal blogs have regained a new value—they are not just content publishing platforms in the traditional sense, but an information space maintained by an individual over the long term.

The value of this information space lies not only in its ability to preserve content over the long term, but more importantly, in its ability to allow personal content to continuously accumulate, be fully expressed, and gradually form interconnected information structures. Specifically, this value is mainly reflected in the following three aspects:

• Personal blogs provide an open and stable information space: Unlike the platform-based Internet, personal blogs are usually built on independent domains. The content published by authors does not rely on the internal recommendation mechanisms of a platform, but has a long-term public address that can be accessed by search engines, cited by other websites, and continuously remain part of the open Internet. This openness allows individuals to continuously crystallize content around their long-term practices. Articles from different times and on different topics are no longer just scattered platform records, but can gradually form a continuous accumulation of information.

• Personal blogs can preserve a more complete process of expression: With a long-term, stable hosting space, blogs also provide a more complete narrative space. Compared to platform content that emphasizes instant dissemination, authors can record how problems arise, how solutions are designed, how practices are advanced, and how their own understanding changes with experience. The value of this long-term record lies in the fact that it shows not just the final answer, but an individual's way of thinking, practical methods, and cognitive evolution process when facing problems.

• Personal blogs can further organize this information: As content continues to accumulate, a mature personal blog is not a simple collection of articles, but can connect individual articles into a larger knowledge structure through categories, series, citations, and relationships: articles record specific practices, series present the process of technical evolution, themes reflect long-term focus areas, and these contents together constitute an information system centered around the author.

Therefore, the value of personal blogs in the AI era does not lie in replicating the content dissemination models of the traditional Internet era, nor in competing with various platforms for traffic.

Its more important significance lies in providing individuals with a long-term, open, and stable information hosting space, allowing past scattered experiences, practices, and thoughts to continuously accumulate around the same subject, and gradually form a more complete information structure.

This structure not only makes it easier for others to understand the author, but also provides a better foundation for future AI to identify, connect, and understand individuals from the public Internet.

4 How to Make Personal Websites Easier for AI to Understand?

Personal websites can become an important carrier of personal digital persona in the AI era, but this does not mean that traditional blog structures can directly adapt to the new information environment.

Over the past two decades, the development logic of personal websites has been mainly built on SEO (Search Engine Optimization). The focus of website construction is to make content discoverable by search engines, improving page visibility in search results through keyword matching, page quality, external links, etc. In this model, the core unit of a website is the “page.” For traditional blogs, this means organizing around individual articles, helping users locate information through categories, tags, and search functions. As long as an article can solve a specific problem, it can gain independent traffic.

But the arrival of the AI era has changed the way website information is understood. As generative AI gradually becomes the new entry point for information, websites also need to adapt to this new information environment, a direction often referred to as GEO (Generative Engine Optimization).

Unlike SEO, which focuses on “how to make pages discoverable by search engines,” GEO focuses more on “how to make the information on a website understood by AI.” For AI, the value of a website is not just about how many articles it contains, but whether there is a clear relationship between these contents and whether they can jointly express some stable information characteristics.

This also means that websites need to gradually shift from focusing on the visibility of individual pages in the past to building the overall content structure: The connections between articles, the evolution between series, the accumulation of long-term themes, and the author information behind the content can all become important bases for AI to understand a website.

To adapt to this change, personal websites need to reorganize their own information structures in three aspects.

First, shifting from page-centric to content relationships

Although traditional article categorization and page organization methods can still help users browse and retrieve information, they are no longer sufficient to fully express a website's information structure. This is because AI not only needs to know what articles are on the website, but also needs to understand why there are connections between these contents.

For example, if a website contains content related to network architecture, server deployment, artificial intelligence, and knowledge management at the same time, and these articles are simply arranged by different themes, AI may only identify several independent sets of information, without being able to determine why they appear together on the same website.

Therefore, an AI-oriented website structure needs to further clarify the connections between content, including citation relationships between articles, series relationships, thematic evolution relationships, and their connections with the author's long-term practical direction. In this way, the content on the website can gradually form an information whole with contextual relationships from a large number of independent pages.

Second, shifting from individual articles to long-term accumulation

In the SEO era, an article that can accurately solve a specific problem could gain independent traffic, so website content often revolves around specific needs.

But what the AI era needs to understand is not just what problem a certain article solved, but also the direction and characteristics reflected by these contents over the long term. An individual's professional ability, way of thinking, and practical experience are usually gradually revealed through long-term recording and continuous expression. Therefore, the value of a personal website is not just about recording “what problems I have solved,” but more about presenting “how I understand and explore these fields.”

Only when content from different times and on different topics can jointly demonstrate an individual's practical path and cognitive evolution does a website truly possess the ability to express an individual's long-term value.

Third, shifting from content collections to author entities

In addition to organizing the relationships between content, personal websites also need to express the author themselves more clearly. In the past, users usually got to know the author gradually while reading articles. A reader might enter a website because of a specific problem, and then learn about the author's background, experience, and characteristics through continuous reading. But in the AI era, AI also needs to determine whether these contents come from the same stable subject, and whether there is long-term consistency between these contents.

The “author entity” mentioned here is not simply adding a personal profile page, but transforming the author's identity from a signature behind articles into a clear subject in the website's information structure. Connections need to be formed between the author, articles, professional fields, and practical experiences to jointly answer: Who does this website belong to? What does this person focus on in the long term? Why can these contents form a continuous development path?

Only in this way can the content on the website form a continuous information structure around a clear subject, and gradually reveal this individual's long-term practices and cognitive characteristics.

Supplementary explanation on “Author Entity”

The “entity” mentioned here comes from the concept of Entity in knowledge engineering.

When AI organizes Internet information, it does not simply read isolated pages, but tries to identify different information subjects and establish relationships around these subjects. For example, Tang Wudi taking this as an example, AI needs to determine who this name corresponds to, and further associate their articles, technical fields, practical experiences, and other information.

These relationships can be simply understood as:

• Entity: represents a specific information subject, such as Tang Wudi this author;
• Relation: describes the connection between the entity and articles, fields, projects, and experiences;
• Evidence: sources of information used to support these relationships, such as long-term published practical articles and technical records.

Therefore, the value of a personal website in the AI era is not just about storing articles; more importantly, it provides a stable source of information for the author entity on the public internet, and gradually enriches the relationships and evidence surrounding this entity.

For more information about knowledge engineering, you can refer to another article: "Personal Knowledge Engineering (II): From Cognitive Structure to Knowledge Engineering—Why Does Knowledge Need to Be Structured?》。

5 From Content Recording to Knowledge Structure: Building a Personal Information System Understandable by AI

1 Attempts from Traditional Blogs to AI-Understandable Websites

After recognizing the changes in information organization faced by personal websites in the GEO era, I began to re-examine my own blog.

Over the past few years, I have always used my blog as a recording space for personal technical practices: from initially recording home infrastructure, network configuration, and the use of various tools, to later gradually delving into system architecture, AI, and knowledge management-related exploration, the content in the blog has continuously accumulated with the development of my personal technical path. This content records my long-term personal practice process and has also formed a certain scale.

However, from the perspective of the capabilities required for personal websites in the GEO era, there is still room for further optimization in the original blog structure: on one hand, as the number of articles continues to increase, the original connections between different topics are gradually hidden within a large number of independent pages; on the other hand, the technical evolution path formed by long-term accumulation, as well as the author's own information, have not been fully expressed.

Therefore, I began to rethink the blog structure according to the three directions proposed in Chapter 4 and tried to make corresponding adjustments:

• Let the content, which was originally centered on a single article, gradually form an information network with internal connections.

• Allow the articles accumulated over the years to more clearly present the personal practice path and cognitive evolution.

• Allow the website to not only display the content itself, but also more clearly express the author as a stable subject.

These adjustments are not simply to add blog functions, but to explore a new form of personal website: on the basis of retaining traditional blog recording capabilities, allowing long-term accumulated information to form a clearer structure and be more easily understood by humans and AI.

2 From Pages to Content Relationships: Building an Information Network Within the Blog

The first problem to be solved is how to make the blog structure, which is centered on a single article, gradually form connections between content. The blog's built-in categories, tags, and search functions do not help much, so I can only try to establish relationships between different levels by developing some new functions myself.

First, at the article level, I connected content with common themes, developmental relationships, or practical backgrounds by adding features such as “Series Article Cards” and “Related Articles” to specific article areas. Meanwhile, through the “Right Menu” feature, dynamic content recommendations are generated based on semantic association mechanisms, allowing the system to discover potential connections based on vector similarity between content, so that the association between articles no longer relies entirely on manual classification.

Secondly, I also tried to reorganize the content from the overall level of the website: through the “Blog Knowledge Map” and multiple “Specialized Knowledge Maps”, displaying the main areas covered by the blog, as well as the position of different thematic content in the overall structure; through the article series menu, organizing content with continuous evolutionary relationships by theme, so that scattered articles can form a clearer thematic context.

The goal of these adjustments is not to add more navigation entry points, but to gradually reveal the content relationships that were originally hidden within the blog.

For example, an article about a certain technical practice is no longer just an independent problem solution, but can be connected to related backgrounds, content of the same series, and larger thematic directions. In this way, a single article provides specific information, while the relationships formed between articles jointly present a more complete information context.

Through these methods, the content in the blog gradually transforms from a large number of independent pages into an information network with contextual connections. This not only makes it easier for readers to understand the connections between content, but also provides a clearer information foundation for AI to understand the overall structure of the website.

For related technical details, please refer to: "Building a “Lightweight Knowledge Index“ for My Blog“》

3 From a Single Article to Long-Term Accumulation: Recording the Personal Technical Evolution Path

Unlike content relationships, long-term accumulation is not achieved through a specific function, but is gradually formed during the process of continuous recording and practice.

Fortunately, this is not something I need to deliberately fuss over for my blog. In fact, since its inception, my blog has not been a collection of content built around a single field, but has continuously expanded with personal technical practices—from early home infrastructure, network architecture, and server deployment, to later Cloudflare, system architecture, AI, and knowledge management, recording the process of technical exploration directions continuously evolving with practice.

On the surface, this content involves multiple technical topics, but they are not completely independent: every change in technical direction comes from the accumulation of problems in previous practices, and the problems encountered in the previous stage often become the starting point for subsequent technical exploration—early exploration of infrastructure promoted subsequent thinking on system design, website architecture, and information organization methods, and these experiences further became the foundation for exploring knowledge management and personal digital expression in the AI era.

From this perspective, the value of my blog is not just about recording “what problems I have solved”, but more about presenting the process of “how I understand and explore these problems”. When content from different stages can jointly reflect personal technical experience and ways of thinking, long-term accumulation itself becomes an important value that distinguishes personal websites from ordinary content platforms.

4 From Content Collections to Author Entities: Letting the Website Express “Who Is Creating These Contents”

4.1 From Author Information to Author Entities

In traditional blogs, the author is usually just a piece of information attached to the article. For example, “Author: tangwudi” displayed at the bottom of an article is enough for human readers. Readers can judge in combination with the website context: this is the maintainer of the blog, this person has been publishing technical articles for a long time, and this name may also have corresponding accounts on other platforms.

However, for machines, “tangwudi” is first and foremost just a string—it cannot judge solely by this name: Is this a real person? A user account? An organization name? Or a brand name? Similarly, it cannot automatically confirm whether “tangwudi” in the blog also corresponds to the corresponding account on GitHub or other platforms.

This is precisely the difference in how humans and machines understand information: humans can establish identity judgments through context, while AI requires more explicit information descriptions. Therefore, in an AI-understandable website structure, the author cannot just be a name in an article, but needs to become an explicitly described information object, that is, an author entity.

The “entity” mentioned here is not simply adding a personal profile page, nor is it adding more text descriptions to the author, but establishing a stable identity identifier for the author through a structured approach, and associating information such as articles, professional fields, and practical experiences on the website with this identity.

For example, my blog website can, through structured data, describe my online alias Tang Wudi as a Person type of entity:

{
  "@type": "Person",
  "@id": "https://blog.tangwudi.com/#person",
  "name": "tangwudi"
}

Among them,@type is used to tell the machine: this object belongs to the Person type, which represents a real person;name is used to describe that this person's alias is Tang Wudi; while @id is used to provide a unique identifier for this entity: ”https://blog.tangwudi.com/#person“.

The string herehttps://blog.tangwudi.com/#personis not a webpage address that requires access, but is more like a unique ID in a database. It indicates that in the information structure of this website, there is a Person entity representing the author tangwudi. Later, if other content on this website references this identifier, the machine will be able to recognize that they point to the same subject. For example, an article contains:

{
  "author": {
    "@id": "https://blog.tangwudi.com/#person"
  }
}

Another article also contains:

{
  "author": {
    "@id": "https://blog.tangwudi.com/#person"
  }
}

Then for AI, what it sees is: both articles are associated with the same Person entity. This means that the author's identity behind the articles is no longer just repeatedly appearing text, but has become a stable information node.

It should be noted that,@id The specific string that follows does not have a fixed format requirement, as long as it can uniquely identify this entity and remain consistent within the website. In fact, website maintainers can define different forms of @id, such as using custom numbers or other identifiers with clear meanings.

The reason why this is used here https://blog.tangwudi.com/#personis mainly because WordPress's SEO plugin Rank Math generates such an entity identifier by default based on the website domain name. Since this identifier already meets the requirements, I did not redesign it in the actual modification, but directly continued to use the default form generated by Rank Math.

In addition, the author entity can also, through sameAs and other methods, establish corresponding relationships with identities on other platforms. Taking the association of my GitHub account as an example:

"sameAs": [
  "https://github.com/tangwudi1979"
]

What it expresses is not “these are my other links”, but rather that the [something] in the blog Tang Wudi and the [something] in GitHub tangwudi1979 point to the same real-world subject. In this way, AI can attribute information distributed across different platforms to the same author entity.

Overall, for personal websites, the significance of the author entity lies in: allowing the content on the website to be clearly attributed to a stable individual subject—in plain terms, the author entity can improve the recognizability of personal identity on the Internet.

Taking myself as an example, the blog can provide machines with clearer identity clues, allowing Tang Wudi this name to more easily establish stable associations with my blog, GitHub account, and long-term accumulated technical content. As these associations continue to accumulate, the AI's understanding of Tang Wudi will no longer be just an ordinary string, but will gradually form a clearer identity recognition.

In layman's terms, it means continuously supplementing contextual information about this name, making it easier for AI to “know” the specific subject behind this name, thereby reducing confusion with other accounts of the same name, nicknames, or other meanings.

This is also the core meaning of personal websites moving from a “collection of content” to an “author entity”.

4.2 Giving Personal Websites a Clear Author Identity Through JSON-LD

In the previous section, we talked about the structured data used to represent the author entity:

{
  "@type": "Person",
  "@id": "https://blog.tangwudi.com/#person",
  "name": "tangwudi",
  "sameAs": [
    "https://github.com/tangwudi1979"
  ]
}

This type of structured data is usually represented in JSON-LD (JavaScript Object Notation for Linked Data) format—it is not content displayed to users on a webpage, but rather a machine-understandable description provided for search engines and AI systems to read, used to explain what the entities in the webpage are and what relationships exist between them.

Therefore, to give a personal website a basic author entity, it essentially involves adding JSON-LD structured data describing this person into the webpage.

For static websites, you can usually add the following code directly to the <head> section of the HTML template:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Person",
  "@id": "https://blog.tangwudi.com/#person",
  "name": "tangwudi"
}
</script>

In this way, when search engines or AI crawlers visit the webpage, they can read the author entity information within it.

However, for dynamic websites, such as WordPress, since webpage content is usually dynamically generated by themes, plugins, and databases, it is not suitable to manually modify every page. A more reasonable way is to use existing SEO plugins to generate the basic structure and then adjust it according to actual needs.

Taking my blog as an example, the SEO plugin used is Rank Math (for a related introduction, you can refer to:Rank Math SEO Setup and Optimization: From Installation to Introduction of Common Functional Modules)

At the website level, a default subject entity will be generated:

{
  "@type": ["Person","Organization"],
  "@id": "https://blog.tangwudi.com/#person",
  "name": "tangwudi"
}

Meanwhile, in articles, Rank Math will generate author and publisher associations respectively:

{
  "author": {
    "@id": "https://blog.tangwudi.com/me/tangwudi/#author"
  },
  "publisher": {
    "@id": "https://blog.tangwudi.com/#person"
  }
}

In other words, there are actually two different identity references in the same website:author pointing to the WordPress author identity;publisher pointing to the website subject identity.

Although they both use the name Tang Wudi, because of the different @id , they still belong to two different entities in Schema semantics. This is also a common problem encountered by dynamic websites when building personal digital identities: plugins can generate structured data, but they do not know how the website operator wants to define their long-term identity relationships.

Therefore, instead of continuing to use the default author subject generated by Rank Math, I used the following provided by Rank Math via PHP:

rank_math/json_ld

filter to perform secondary processing on the final output JSON-LD, mainly making three adjustments:

First, clarify the website subject entity as Person

Rank Math default:

{
  "@type": ["Person","Organization"],
  "@id": "https://blog.tangwudi.com/#person"
}

Modified to:

{
  "@type": "Person",
  "@id": "https://blog.tangwudi.com/#person",
  "name": "tangwudi"
}

The reason is simple: this website is essentially a personal technical blog, and there is no corporate subject that needs to be identified, so there is no need to keep the Organization type.

In this way,https://blog.tangwudi.com/#person becomes the core entity representing the author's identity uniquely across the entire website.

Second, unify the article author to point to the core Person entity

By default in Rank Math, articles may reference the entity corresponding to the author archive page:

{
  "author": {
    "@id": "https://blog.tangwudi.com/me/tangwudi/#author"
  }
}

This association points to the author archive page, rather than the personal identity entity that is hoped to be accumulated over the long term. Therefore, modify the author reference in the article to:

{
  "author": {
    "@id": "https://blog.tangwudi.com/#person",
    "name": "tangwudi"
  }
}

In this way, all articles on the website will be explicitly associated with the same author entity.

Third, add identity attributes to the Person entity

The basic Person entity can only state that “this person exists,” but it cannot explain this person's fields and associations. Therefore, I further added:

{
  "description":
  "A technology practitioner exploring infrastructure, AI, and knowledge engineering through systematic thinking and engineering practice.",

  "sameAs": [
    "https://github.com/tangwudi1979"
  ],

  "knowsAbout": [
    "Infrastructure Engineering",
    "Web Architecture",
    "Artificial Intelligence",
    "Knowledge Engineering"
  ]
}

Where:description Used to describe who this entity is;sameAs Used to declare that accounts on other platforms on the internet belong to the same person;knowsAbout Used to describe the fields of knowledge that the entity has long focused on and practiced.

Ultimately, through these adjustments, the structured data on the webpage is no longer: "This article was published by an author named tangwudi." Instead, it becomes: "There exists a Person entity named tangwudi on the internet, who has a clear website identity, GitHub association, and continuously generates content in the fields of infrastructure, AI, and knowledge engineering."

However, it is important to note that establishing the basic author entity is only the first step. JSON-LD solves the problem of letting machines know that “a clear author identity exists on this website.” But it cannot fully describe: which fields this author has long focused on; what technical practical experiences they have; and why these articles can form a continuous path of development.

In other words:The author entity solves the problem of “identity declaration,” not “personal knowledge expression.”

Google Rich Results Test verification

After completing the Schema adjustments, you can verify the final output using Google's Rich Results Test tool. From the results, Google is already able to correctly parse the relationship between the article and the author entity, while also displaying the identity attribute information I added for the author entity:

4.3 Enhancing Author Entities: Letting AI Understand “Who This Person Is”

After completing the basic author entity, the website is already able to clearly tell machines that this content belongs to the same author. But for AI, simply knowing “who this person is” is not enough.

For example, the basic author entity can state: the author of this website is Tang Wudi; this author is associated with a certain GitHub account; the articles on the website are all published by this author. However, it cannot further explain: which fields this author has long focused on; what practical experience they possess; why this content appears on the same website over the long term; and what professional direction and cognitive characteristics are reflected behind this content.

Therefore, beyond the basic author entity, it is also necessary to further supplement information that can describe the author's characteristics, allowing the author entity to gradually expand from a simple identity identifier into a more complete information node. Based on this idea, I tried to add some information entry points oriented towards the future AI environment, mainly by adding author-profile.json 和 llms.txt these two files, to provide more context for machines to understand the author's identity and the website's content.

Use author-profile.json Describing Author Information

To supplement the information of the author entity, I created the author-profile.json file, which is used to centrally describe the author's basic information, technical fields, long-term practical directions, and content characteristics.

Unlike the Person entity in JSON-LD,author-profile.json is not used to declare that “this person exists,” but rather to further describe: who this person is, and in what aspects this person's long-term accumulation is reflected.

For example, a simple author profile description could look like this:

{
  "name": "tangwudi",
  "description": "个人技术博客作者，长期记录家庭基础设施、网络架构、AI 和知识管理等方向的实践。",
  "fields": [
    "HomeLab",
    "Network Architecture",
    "Cloudflare",
    "Artificial Intelligence",
    "Knowledge Engineering"
  ],
  "focus": [
    "技术实践记录",
    "系统架构探索",
    "个人知识管理"
  ]
}

The focus here is not the fields themselves, but rather explicitly expressing information that was originally hidden in human text through a structured approach.

This way, when the AI understands the website content, the information it obtains is no longer just: an author named Tang Wudi published some articles. It can also further understand: this is a technical author who has long practiced and continuously documented experiences around infrastructure, network architecture, AI, and knowledge management.

During actual deployment, the file can be placed directly in a publicly accessible location on the website, for example:

https://blog.tangwudi.com/author-profile.json

These files can serve as machine-oriented information portals, providing more structured context for search systems, AI Agents, or other automation tools that may visit the website in the future.

In this way, a more complete information relationship is formed among the author entity, the website content, and the author profile description.

Use llms.txt Providing AI-oriented information portals

In addition to describing the author themselves, you can also use site instruction files oriented towards AI systemsllms.txt, to provide more direct website context. For example:

https://blog.tangwudi.com/llms.txt

The following is the content of my blog's llms.txt:

# tangwudi blog

This website is the personal knowledge base of tangwudi, a technology practitioner exploring how complex technical systems are built, operated, understood, and transformed into structured knowledge.

## Author

Author profile:

[https://blog.tangwudi.com/author-profile.json](https://blog.tangwudi.com/author-profile.json)

tangwudi focuses on:

- infrastructure engineering
- web architecture
- artificial intelligence
- knowledge engineering

The author's methodology includes:

- first principles thinking
- system evolution analysis
- engineering practice
- knowledge structuring

## Core Topics

### Infrastructure Engineering

Exploration of:

- home data centers
- self-hosted services
- networking
- virtualization
- Linux and container technologies

### Web Architecture

Research and practice around:

- WordPress architecture
- Cloudflare ecosystem
- CDN and security optimization
- reliable personal websites

### Artificial Intelligence

Exploration of:

- large language models
- embeddings
- semantic search
- AI-assisted workflows

### Knowledge Engineering

Research into:

- knowledge organization
- semantic indexing
- structured knowledge representation
- AI-readable knowledge systems

## Knowledge Map

The [Blog Knowledge Map](https://blog.tangwudi.com/roadmap) provides a structured overview of the site's core knowledge domains, topics, and related articles. It can be used as the primary navigation entry for understanding the structure and relationships of the site's content.

## Content Philosophy

The website focuses on understanding technologies through their underlying principles, historical evolution, practical implementation, and architectural trade-offs.

The goal is not only to document solutions, but to transform engineering experience into reusable knowledge structures.

Compared to ordinary navigation on web pages,llms.txt what is provided is a website summary oriented towards AI understanding. It does not replace the website content, but rather helps AI quickly understand: what this website is; what the author focuses on; and which content is representative.

However, it should be noted thatcurrently llms.txt it has not yet become a de facto standard, and the level of support for it varies among different AI systemsThis article merely regards it as an attempt at information expression for the future AI environment, rather than a mature, universal specification.

Through the previously added author information, structured data, and these additional enhancement methods, the author entity is no longer just a name and an identity identifier, but gradually becomes an information node containing identity, experience, professional direction, and long-term accumulation.

From a deeper perspective, the goal of these efforts is not simply to add an author introduction, but to continuously strengthen Tang Wudi the association between this entity and related content in the open internet. Through continuously accumulated practical articles, technical records, and structured information, it helps AI more accurately distinguish information from different sources and gradually establish a stable understanding of the subject represented by that name—in plain English, the hope is that in the future, when mentioning Tang Wudi, AI can prioritize me as an individual subject who engages in long-term recording and continuous practice.

6 When AI Begins to Understand a Personal Website

1 Author Profiles Formed by AI Based on Public Information

The previous chapters discussed the capabilities that personal websites need to have in the AI era, and how I made the content, relationships, and author identity in my blog easier for machines to understand through structured transformation.

However, the final effect of these transformations cannot be judged solely by the technical implementation itself. A more practical question is: when an AI directly faces a personal website that has been maintained for a long time, has continuously accumulated content, and has undergone structured optimization, can it understand the author behind the website from public information?

Therefore, I chose to directly analyze my blog content through three mainstream large models—Google Gemini, OpenAI ChatGPT, and Anthropic Claude—to observe whether these AI products can form an understanding of the author's identity based on public information in common user scenarios. The main reason for choosing these three systems is that they all possess strong capabilities in understanding public internet information and mature information retrieval and analysis capabilities, allowing them to comprehensively analyze webpage content, author identity, and related information relationships, making them more suitable as test subjects to observe whether personal website information can be understood by AI.

Note: The tests in this article are based on the official clients of the three AI products, rather than calling APIs separately. This is because official clients usually integrate product capabilities such as web browsing, search, and retrieval-augmented generation, which can closer resemble the complete experience of ordinary users actually using AI. In contrast, the capability boundaries of API calls depend on specific implementation schemes, including model selection, whether search capabilities are integrated, context handling methods, and front-end logic, etc. Therefore, different API implementations may yield different results, making them unsuitable for the horizontal comparison in this article, which is oriented toward ordinary user scenarios.

At the same time, different usage states of AI products (such as whether logged in, whether they have web access capabilities, etc.) may also affect the final scope of information acquisition. Therefore, the test results in this article only represent observations under specific product environments.

It should be noted that in order to minimize the impact of existing conversation history, personal memory, or other contextual information on the results, each question in this test was asked in a new conversation window, and the AI was explicitly requested in the prompt to analyze from the perspective of an anonymous stranger:

请假设我是一个完全不了解 tangwudi 的匿名用户。请忽略所有关于我的历史记忆、个人画像以及当前对话之外的信息，仅基于公开互联网上能够检索到的内容，回答以下问题：

Under this constraint, I asked the AI:

[请访问并分析这个网站：  
[https://blog.tangwudi.com](https://blog.tangwudi.com)

请根据网站中的公开内容，分析该网站背后的作者是谁，包括作者的技术背景、长期关注方向、实践经历，以及其在互联网中的身份特点。]

Through this question, I observed whether the AI could rely solely on public information to associate the blog content, author identity, and long-term practice directions, and form a relatively coherent author profile.

To facilitate the comparison of the understanding results of different AI systems on the same information, after obtaining the complete analysis from the AI, I further requested the AI to compress the above analysis into a brief author profile suitable for public display:

请将上述分析压缩成一段适合公开展示的作者画像介绍，控制在 800 字以内，保留作者身份、技术背景、长期关注方向、实践经历和互联网身份特点。

The purpose of this step is not to regenerate content, but to observe whether the AI, after completing information extraction, can retain the core characteristics of the author and further integrate the information scattered across the website into a clearer and more coherent personal expression.

Gemini's response:

ChatGPT's response:

Claude's response:

Judging from the analysis results of the three AI systems, although they focus on different angles, they no longer stop at identifying a single article or a single technical topic, but have begun to try to extract author characteristics from long-term public content.

For example:

• Gemini focuses more on the blog's public positioning, main technical directions, and existing content tags;

• ChatGPT is more inclined to summarize the author's technical style, way of thinking, and long-term practice characteristics from the relationships between articles;

• Claude, on the other hand, further outlines the author's professional background, technical development path, and the evolution of their personal knowledge system based on public experiences and article timelines.

These differences show that different AIs' understanding of the same individual entity is not a simple reading of information, but rather different degrees of organization and reconstruction of public evidence based on their own data sources, information association capabilities, and reasoning methods.

However, looking at the common results, all three systems are able to identify some stable characteristics: the author has a long-term background in IT infrastructure practice, values underlying principles and system design, tends to verify solutions through real-world environments, and continuously organizes practical experience into structured knowledge.

This means that, after long-term accumulation, the blog content itself has begun to possess the ability to express the author's identity, professional direction, and cognitive characteristics.

2 The Entity Association Status of the Name “tangwudi”

In addition to analyzing the blog author's profile, I also made a more direct observation: when AI is faced with Tang Wudi this name, whether it can stably associate it with the same individual subject.

To avoid factors such as historical conversations and user profiles affecting the test results, the test still adopts the perspective of an anonymous stranger:

请假设我是一个完全不认识 tangwudi 的匿名陌生人。请忽略所有关于我的历史记忆、个人画像和后台上下文，仅依据公共互联网上能够公开检索到的信息，回答这个问题：“tangwudi 是谁？”

Gemini's response:

It should be noted that the entity understanding formed by different AI systems on the same public information is not completely consistent. For example, Gemini's response focuses more on the blog's public positioning, main technical directions, and existing content tags; ChatGPT further combines the relationships between articles to summarize the author's technical style, practical methods, and long-term thinking characteristics; Claude is more inclined to outline the author's development path from public experiences, timelines, and the website ecosystem.

This difference does not mean that a certain result is definitely correct, but rather reflects the differences in information sources, data updates, content association capabilities, and reasoning methods among current AI systems—the information left by the same person on the internet needs to go through the information processing procedures of different AI systems themselves to form the final entity understanding result.

From the current test results,Tang Wudi this name has already been able to establish a relatively clear association with blog.tangwudi.com and related public technical content in multiple mainstream large models. Although different AIs have different focuses, overall they have been able to identify from public information an individual subject who has long maintained a personal technical blog, continuously recorded IT practices, and explored directions such as home data centers, network architecture, and AI and knowledge systems.

However, this association has not yet reached the level where all AI systems can stably and accurately identify it. For example, in tests of some models (such as Qwen, DeepSeek), when directly asking “Tang Wudi who is”, the recognition results are not stable, but after adding qualifying conditions such as “blogger”, it is easier to get accurate association results.

This may be related to differences in the information sources and information processing methods that different AI systems rely on. Some models also mentioned in their answers that their understanding of public information mainly relies on Chinese internet content, so for Chinese personal websites and public information in the local internet ecosystem, their recognition effect may differ from other large models facing the global internet.

Of course, I also have a slightly “whimsical” long-term goal: I hope that one day in the future, when directly asking “Who is tangwudi” in different AI systems, they will all be able to stably associate it with me as an individual subject based on public information, and form a relatively consistent and complete understanding.

If this can really be achieved, perhaps it can be considered a true completion of the transition from a username to an individual entity with a stable identity on the open internet that can be continuously understood by AI.

The testing method in this section is actually still not rigorous enough. Although a new conversation was used during testing and it was explicitly requested to ignore historical memory, personalization factors at the account and device levels may still affect how AI obtains and processes information, such as region, language preferences, and past usage behavior—these factors are not necessarily completely excluded by the instruction to “ignore historical memory.”

Therefore, if we want to strictly verify Tang Wudi For the public internet association with this blog, ideally, a third-party device and account completely unrelated to me should be used, keeping the default settings, so that a real stranger can directly ask mainstream AI: “Who is tangwudi?” Only then will the results be closer to a test result free from the interference of personal history.

Therefore, what the experiments in this section can prove is that after these structural modifications, AI is indeed already able to Tang Wudi associate this name with this blog; however, based on this set of experiments alone, it is still impossible to determine how much of this association comes from structured data and how much might be influenced by personalized factors. To further separate these two factors, more rigorous controlled experiments are needed, which is currently impossible to complete due to personal limitations and can only be left for future verification.

7 From Digital Persona to Digital Twin: The Future Evolution of Personal Knowledge Assets

In the traditional internet era, a person's digital presence was mainly reflected in accounts, articles, works, and public materials. This content can prove “what this person left behind,” but it is difficult to fully present “how this person thinks.” In the AI era, acquiring knowledge, organizing materials, and generating standard answers are becoming easier and easier; what is truly difficult to replicate is instead the way of judgment, accumulation of experience, and problem-solving paths formed by an individual in real practice.

For personal websites, the modifications mentioned earlier in the article first solve a practical problem: how to make the information accumulated by an individual over the long term form a more complete and continuous expression in the AI era. But from a longer-term perspective, the significance of these developments is not just to optimize a website, but to gradually build a kind of personaldigital asset. And the truly valuable part of this digital asset is not just what content a person leaves behind, but the experience, methods, and ways of thinking gradually formed behind this content.

When these long-term accumulations gradually form connections and can be understood by AI, they begin to transform from information records into a part of a personal digital persona. And as AI's understanding of the individual further develops, this digital persona may also become an important foundation for the future personal digital twin.

Supplementary notes on “Personal Digital Twin”

“Digital Twin” was first mainly applied in fields such as industrial manufacturing and urban management, referring to mapping real-world objects through digital models and continuously updating them based on real data.

When extending this concept to the personal field, the focus is not on simply replicating a person's digital image, but on using long-term accumulated digital information to continuously form a dynamic mapping of this person.

When this mapping is complete enough, AI may be able to further simulate this person's way of thinking and judgment logic, thereby showing analysis and decision-making tendencies similar to this person when facing specific problems.

For example: Why did this person ultimately choose a certain plan? Why did they give up several other seemingly better choices? How do past experiences affect today's judgment? These contents are what truly belong to personal cognitive ability.

And this ability cannot be obtained through short-term collection, nor can it simply rely on a single model training session to generate. It requires real records, continuous accumulation, and constantly transforming personal experiences into understandable information.

Personal blogs happen to be able to preserve such long-term records: An article records a specific problem and its solution process. Over long-term accumulation, these records will gradually present a person's practical experience, choice methods, problem-solving approaches, and cognitive changes.When this information can be continuously understood and associated by AI, they are no longer just historical records left by individuals, but may become a new personal advantage.

In the future, there may even emerge a new group of digitally advantaged people: they are not necessarily public figures in the traditional sense, nor do they necessarily have a large number of fans or social influence, but because they have continuously left real, continuous, and valuable information records on the open internet, AI can associate relevant information with a stable personal subject, such as directly knowing “who tangwudi is”. Of course, this association also has a very realistic premise:The more unique the name, the lower the difficulty of disambiguation usually is— “tangwudi” is definitely much easier than “Zhang San” or “Li Si”~~.

In this sense, these people can be called “digital asset accumulators” in the AI era.And the long-term value of personal websites lies precisely in providing an open, stable, and continuously traceable carrier for this accumulation.