Five years of blogging, and now I want AI to find me

A few days ago I did something fun: I opened DeepSeek and asked it, “Who is Lin Wenwei?”

It looked through 11 web pages and answered very earnestly. But the answers included a vice president at a multinational, an assistant professor of accounting, and the legal representative of a metals company. Not me.

I switched to ChatGPT and asked it to pretend to be someone who didn’t know me at all and search for “him” in the third person. It did find my GitHub and read a blog, but unfortunately it was my old blog repository from years ago, and it took that for who I am now. Later someone used an AI called WorkBuddy to search for “Lin Wenwei personal blog”. At first it searched with my former company’s name attached, and only found the right direction after going in circles.

I don’t think people have to get to know me through AI first. Talking face to face, or working on something together, shows far more sides of a person. I’m just curious: if someone who doesn’t know me at all asks an AI first, can they find me and see what I’m doing? It’s a fun question, I wanted to try, and I want my voice to be seen.

So these past few days I’ve been revising my blog while searching for myself in different AIs. This post is about why, how, and a few things I figured out along the way.

I wanted a personal brand early, yet my blog took five years

I’ve heard the term “personal brand” many times. In 2021 I took a Content Marketing course on Coursera and took lots of notes. One line has stuck with me: content has to earn attention first; nobody owes you their attention.¹ Later, in a communications course, the teacher said a personal brand is about letting an individual be discovered. In early 2025 I read Austin Kleon’s Show Your Work and thought of that course again. They say the same thing: put what you’re working on where people can see it.²

I learned it, but I did very little of it.

My blog started in 2021, built with Hexo and deployed on Azure. In 2023 I tried Blazor, and also Astro. In January 2025 I wrote a post called “I finally choose Astro to build this blog”, thinking I’d finally settled. By July I wanted to redesign it with the UX skills I’d learned, and even thought of turning the blog into a personal “curation”: hanging out on forums in primary school, playing with E Language in middle school, picking programming back up in university, laid out chapter by chapter. This March I asked AI to help me design it. We talked all evening, and in the end I wrote: “Not happy with it. Maybe even I don’t know what kind of blog I want?”

In five years I changed frameworks several times and redid the home page many times over, but wrote only a handful of posts.

Still, the blog wasn’t always useless. I especially love working on my blog before job hunting, maybe because I used to do front-end work. The closer an interview gets, the more I care about it.

In summer 2023 I was going to interview for a front-end internship and needed to show projects I’d done, so I spent a day building a blog by hand and added a Portfolio section with all my earlier work. On August 31 I wrote in my “three good things”: “Finished the project showcase on my blog.” I posted the portfolio to WeChat Moments and lots of friends liked it. A team lead at a company I’d applied to earlier saw it and reached out to me.

At the interview, the senior engineer interviewing me was quite interested in me and thought this kid was more interesting than the others. I got the offer, started right away, and stayed there for almost three years.

Now I’m back to the blog. This time is a bit different: before, I worked on the blog mostly to find a job; this time it’s more about expressing my own things and letting people get to know me, and it’s fine even if no job comes of it. But when it came to actually writing, I got stuck on “writing” once again.

A note from May 2025 still rings true: “This time I want to start updating my blog. But I haven’t figured out how to write… Why is it that the more I learn, the harder it is to take a step?”

The odd thing is that I’m not like that when I code. With code, I “write first and see”, get it running, then fix it. With words, I start worrying about structure, sit down and hold myself to perfectionism, and then can’t write anything. Design is a bit like that too: making a page that looks designed takes a lot of time. Doing design and copy at once, I’m not skilled enough yet, and I end up doing neither well.

A blog is a personal portal

So why have a blog at all, instead of just posting on WeChat or Xiaohongshu?

One very practical reason: web pages are easier to discover than apps. Content on WeChat Official Accounts and Xiaohongshu mostly stays locked inside their apps. AI agents can’t use apps; they rely on crawlers reading web pages, and if a platform isn’t open, they can’t read it.³ Web pages are different: as long as a crawler can open them and a search engine has indexed them, both people and AI can find them.

My current view: in an era with more and more independent individuals, a blog is a person’s own portal. It’s useful in at least five ways. The first four are for others, the last one is for me.

First, it’s where my content gathers. WeChat articles can be deleted for all sorts of reasons, and platforms can change their rules or shut down. What’s on my blog is mine, in my own place. I can post on other platforms, but everything can come back here.

Second, it shows a more rounded me. A résumé is more like a set of labels: which company, what role, what technologies. Through what I write, people can see how I think about problems, what I care about and what pitfalls I’ve hit, a fuller picture rather than a single point.

Third, it links to everywhere else. In the age of agents this is especially useful. I put links on my blog to GitHub, LinkedIn, and articles on Microsoft’s sites that mention me. People and AI can follow these links and piece scattered information together into the same person. The more links, the more paths, and the easier it is for people to discover what I’m doing.

Fourth, it leaves a way to reach me. Being found is wasted if people don’t know how to contact you. So you need a call to action telling people how to get in touch. I put my email on my GitHub profile, and the About page on my blog has a short “come chat with me”. Be generous about it.

Fifth, it helps me think things through. I have lots of notes, but they’re scattered: a line today, a paragraph tomorrow. A blog post is different. I have to string scattered ideas together and work out what to say first, what comes next, and what I’m really trying to say. Written down, it becomes a system. Writing is really a way of getting to know myself.

In the end, everyone is building a personal brand. Creators on Xiaohongshu use the platform to be themselves and be seen. Otherwise, how would anyone get to know you? I used to think it was a bit of a pose. Now I think learning to present yourself, being a bit more generous about it, is nothing bad.

A blog is a kind of publishing too

On the afternoon of October 31, at a bookstore, I heard Cai Xin talk about her experience using a personal agent. She has worked in publishing for twenty years, and that day she talked a lot about how publishing will change.

She shared one idea: when writing was first invented, farmers probably thought, why would I learn to read? Later, literacy became something everyone needs. When the internet arrived, everyone could have their own media, and some people still thought, who writes anything these days?⁴

She also said that anyone with something worth spreading and sharing is a writer, “not just famous authors”; future publishing won’t only mean helping an author put out a book, but also personal websites, personal agents, and helping authors connect with others; “words are only a starting point”.⁵

That day I wrote just one line in my notes: “Really need to speak up and let myself be known.”

Afterwards I thought: then running my own blog counts as a kind of publishing, doesn’t it? Publishing, at bottom, is passing on information. And even if it doesn’t count as publishing, it’s marketing, letting my voice be heard. I’m happy with either description.

I want to write mainly to share my experience, help others, and let people see a more rounded me. I’ve written quite a few things before: Git basics, setting up WSL, inclusive design, book reviews, a look back at graduating from the student ambassador program. Looking back, they weren’t perfect, but people found them through search and read them, so they weren’t written for nothing.

What Cai Xin helped me let go of was mainly my worry about language. Whether my writing reads beautifully is something I can learn slowly; it just takes time. I can’t stay silent just because I haven’t figured everything out.

How I write now

So I made a trade-off: set design aside and write the content first. Write the content for machines first, so it gets indexed and found; a beautiful version for people can come later.

These past few days, this is how I’ve been writing:

1. Say whatever comes to mind first. I mostly use voice input now. Whatever I want to write, I just talk it out at my computer. It’s fine if it’s messy.
2. Have my personal agent find material. I’ve built a knowledge base holding years of notes, conversations with AI, book highlights and work history. I give it a topic and it digs up the related things: what I said on which day, what I did when, even details I’d forgotten.
3. Have AI draft a first version. With the material in place, AI can write a first draft quickly. Like iterating on a product, get an “ugly” version first.
4. Revise by reading and talking it through with AI. As I read, new ideas often come up, and only then do I know how the piece should really go. Then I keep telling the AI: the structure’s wrong, this isn’t what I meant, that should go in too. Sometimes I change a few sentences; sometimes I throw out the whole thing and start over.
5. After publishing, search for myself in AI. See if I can be found and what comes up, then go back and revise the blog.

This approach isn’t new. In June 2025 I wrote: “Making every post high quality is a bit hard… better to let AI write it, as a first draft or a final one; at least I’d have a sample to revise, like writing front-end code.”

It didn’t work back then because AI didn’t know me, and what it wrote was empty. Now it’s different: my knowledge base holds my experience, so AI’s first drafts have real events and real details, and I have something to work with. Yesterday’s two posts, one on how I code with AI and one on the AI features I’ve built, were both written this way.

So was this one. It went through many versions:

• The first version opened with getting stuck writing a Zhihu answer and talked about perfectionism. Reading it, I felt it was off: the core of this piece isn’t fear of writing, it’s wanting my work to last and be found.
• The second version was rewritten along those lines, but I felt it was too shallow and too short, not as good as the two posts from the day before.
• So I had AI go through all my notes from these years on blogging, writing and being found, and we restructured it together. While talking, I also remembered personal branding, the blog as a portal, blogging as publishing, and added them one by one.
• When I read “I’m job hunting again”, I said that’s not right, I’m not blogging to find a job. When I read the part about choosing a framework, I added: web pages are easier to discover than apps, and the framework has to be crawlable. The story of building a blog by hand for an interview in 2023 also only came back to me while we were talking.

From the first version to now, it took only a dozen or so rounds of back-and-forth with AI before the structure settled. If I’d had to write it myself in front of a blank document, I’d probably still be agonising over the first paragraph.

Getting AI to find me: what I’ve tried

Here’s what I’ve tried these past few days, in order of how useful I found it. I plan to write up the technical details in a separate post.

1. First, test “me as AI sees me”

That’s what those searches at the start were. Use different AIs, ask as a stranger would, and see what they find, what they miss and what they get wrong.

This step turned up several problems: my old blog repository was still public, and AI took it for the current me (I’ve since made it private); AI knew very little about what I’ve done and could only guess from my company’s website; and too many people share my name.

2. Make a page that says clearly who I am

I added an About page to my blog that says in Chinese and English who I am, what I’ve done and where to find me. The page also has structured data telling search engines and AI that accounts like GitHub and LinkedIn belong to the same person. Google has documentation specifically on structured data for profile pages.⁶

3. Link everything to everything

GitHub is the easiest place to be found, LinkedIn next. So I put my blog link on my GitHub profile, and the blog links back to GitHub and LinkedIn. It’s worked reasonably well: AI often finds GitHub first and follows the links to the rest.

4. Get indexed by search engines first

This was my biggest pitfall. Many AIs’ web search still relies on search engine indexes behind the scenes. My site had just gone live and wasn’t indexed yet, so AI couldn’t find it.

Different AIs differ a lot too. Using Claude, I found that if a page isn’t in its search results, it can’t open it, so the blog has to be indexed first. DeepSeek’s web search relies on Chinese search engines; I’d only just submitted my blog to Baidu, so it probably won’t find me until Baidu indexes it.

So I submitted the site to Google, Bing and Baidu. Submitting doesn’t guarantee immediate indexing; new sites often wait days to weeks.

5. Don’t fight for ranking: use “name + identity”

With the shared-name problem, my first thought was how to rank higher. Later I found there’s no need. AI search splits a question into several searches and adds qualifiers itself, like “Lin Wenwei Microsoft Student Ambassador”. So the job isn’t to win the three characters of my name, but to make my name always appear alongside a few fixed identities: product manager, manufacturing software, AI agents, Microsoft Student Ambassador. Once indexed, searches with these words can find me.

6. Put keywords in the page title

People searching for my blog used “name + blog”. My post titles don’t contain those words. At first I wanted to change the titles, but thought it would look ugly. Then I learned that search engines read the page’s title and description, not necessarily the big heading you see on the page. So the page title is written as “post title · my name (nickname)”, while the heading on the page stays as it is. The word “blog” goes in the site name and on the post list page; once or twice is enough. Don’t stuff it.

7. Choose a crawlable framework

For pages to be discoverable, crawlers have to be able to read them. Many AI crawlers don’t run JavaScript and only look at the HTML the server returns directly; if a page needs JavaScript in the browser to render, they may see a blank page.⁷ So pick a framework that’s crawler-friendly.

My current blog is written in Next.js. It can render pages on the server too, but you have to keep track, piece by piece, of what runs on the server and what runs in the browser, which is a bit of a hassle, and I don’t use the back end at all. A few days ago I read Nolan Lawson’s comparison of front-end frameworks: server-rendered multi-page sites come out cheaper overall.⁸ I’m considering switching back to Astro, which I’ve used before. It outputs static HTML by default, loads faster, and may make SEO and multiple languages easier. This one isn’t settled yet.

Being mentioned by others matters most

After all this, searching for myself again, I found that what best shows who I am isn’t what I’ve written, but what others have written.

A few days ago I searched my name on Google and came across event pages on Microsoft’s sites from my student ambassador days, and a graduate interview about me.⁹ My first reaction was “Microsoft is great”, and then I corrected myself: no, it’s Liang Di who’s great.

At Microsoft she mainly ran the MVP program, the Most Valuable Professionals, and later also took on the student ambassadors in China, which is when I met her. She introduced me to many online and offline events and to a group of MVPs working at the front line of tech. At Global Azure in Shanghai in May 2023, she brought me on stage for a roundtable; in August that year, at Microsoft’s AI Skills Challenge, I co-ran a low-code session with an MVP. After going to events in Shanghai with her, I often went to Microsoft Reactor and other places for events, and she’d introduce MVPs to me. Whatever articles we wrote or events we ran, she thought of us and put our names on them. I’m truly grateful to her.

Those pages are still there today, and they’ve become the most reliable trail for people to find me. Later Microsoft reorganised and laid people off, and she left Microsoft. Thinking about it still makes me sad.

She once told me I needed to learn to present myself. I didn’t pay much attention then. After I started working, I was always the head-down type, rarely talking about what I’d done, thinking people would notice quietly. That’s steady, but a lot of the effort goes unseen.

Now I think presenting yourself isn’t necessarily a pose. Laying out what you’ve done and linking to evidence others have written is enough.

What I haven’t let go of

Reading this far, it may sound like I’ve worked it all out. I haven’t.

Every morning when I wake up, there’s a voice in my head: what will people say about me? If I post this, will someone think I’m showing off, or that it’s shallow? That voice is loud, and I can’t make it go away yet.

I’ve hesitated too: is spending this much time on being found putting the cart before the horse? Instead of marketing, shouldn’t I go learn more? Later I thought the two don’t conflict. I’ve spent years on B2B software and do have things to write about; I just never wrote them down. Writing them down is a way of organising them too.

So I’ll write first.

There’s an often-quoted story: a teacher split students into two groups, one graded only on the quantity of work they handed in, the other only on quality. By the end of term, the best work all came from the quantity group.¹⁰

Rather than polish one piece, better to write 100 bad ones first. Write enough and it gradually gets better. This post is probably one of them.