AI at Work
Prefer another language? Prefere receber em português? ¿Prefieres recibirlo en español? Choose here / Escolha aqui / Elige aquí: language form

A lot has been said about how we see artificial intelligence. But I hear very few people talk about how artificial intelligence sees us.
I mean the searches people do today with AI tools. When someone asks an AI assistant a question, it doesn't give ten links. It gives three names, maybe five. If you are not one of them, for that question, you don't exist. So the question is simple: are you visible or invisible to AI?
In short
AI assistants answer with a few names, not a list of links. If they can't read about you, you're not one of those names.
Test it: ask an AI about yourself and about your work, while signed out.
Fix it: put clear answers on a page you own, tell Google and Bing about it, and don't depend only on LinkedIn.
What did Marina Mogilko (Silicon Valley Girl) teach me?
I have followed the Silicon Valley Girl channel for some time. It belongs to Marina Mogilko, who also runs a company and a newsletter on beehiiv. Interesting detail: her newsletter is how I discovered beehiiv, the platform I use for this newsletter today.
In August 2026, she published a video called "The Real AI Opportunity | The Ultimate Guide to Become Visible in AI". I saw it come out, but I didn't watch it right away. I saved it to my "watch later" list, and with everything I had going on, I only watched it on September 5, a month later.
I can tell you: it was worth saving. Many things I do today came from that video.
Her story is simple. AI search tools were ignoring her podcast. She asked an AI (Claude) why, and the answer surprised her: the transcripts of her episodes were hidden behind code, so the automated readers that AI tools send to websites, called crawlers, "just saw an empty shell". A crawler is like a library assistant who skims every page to build a catalog. If the page looks empty, nothing goes into the catalog.
So her team did a few things, in plain words:
Gave every episode its own simple page with the full text, so AI can read it.
Fixed the public facts about her. One public database still described her as a "vlogger", what she was 10 years ago. They updated it and translated it into 11 languages.
Told Google about every page. Google found 91 pages it had never indexed, and about 700 new clicks a month came from that.
Measured. They tracked 48 questions against 11 podcast brands. At the start, she showed up in about 7 of the 48. Now she is the 5th most visible of the 11.
One more number from her stayed with me: 82% of what AI cites is "earned media", what other people write about you, not what you write about yourself. Your own page is required, but it's not enough on its own.
Was I invisible to AI?
Right after the video, I thought: am I invisible? I did some of the tests she shows. To be honest, I'm not sure it was a surprise, but yes: I was practically invisible to AI.
That was a big alert for me. The job market today is fast and very competitive. Being invisible is bad for any professional. It is bad for a small business. And it can be bad even for medium and large companies.
But I'll be honest with you. Seeing the result was one thing. Knowing how to reverse it was another. Her video gave me some very good ideas, and I decided to follow them.
Why did I wait five weeks to start?
I registered my personal website, alexrosa.pro, on October 14. That's a little more than a month after I watched the video.
Why the wait? I needed time to decide what to do. And I had doubts. The question always comes: "Do I really need this?" I also thought the work would be too heavy for me, with my full-time job.
Most of all, I didn't want to do it on impulse. I didn't want to build anything without a plan. I wanted a good architecture first.
That planning didn't only start my website. It started the two newsletters I have today, and it's how I'm reshaping my YouTube channels.
What removed the doubt about the workload was AI itself. I started to use my Claude subscription (from Anthropic) much more, and I got very familiar with the tool. I'm not saying this to brag, but today I use it very well. It increased my productivity, and my visibility too.
Then I did what she calls fixing the plumbing:
October 17: my site was verified in Google Search Console, three days after I registered the domain. Search Console is Google's free tool to tell Google your site exists and which pages it has. Think of it as giving the post office your new address.
November 4: I did the same with Bing Webmaster Tools, Microsoft's version for Bing. Why so late? I didn't know it existed. I'm a database professional, learning on the fly.
What does AI say about me now?
On November 9, I ran a few tests on my wife's phone, without signing in to any account.
First, my name. I asked Gemini: "What can you tell me about Alex Rosa that works for Microsoft?" Most of the answer was right: my role, my team, my 30+ years in tech, and both newsletters.

Two things were wrong, and they teach something:
"Data science and career mentor on ADPList." Years ago, I volunteered on ADPList for a few days. The profile stayed, and AI still uses it. An old profile you forgot can speak for you.
"Trilingual, including Spanish." I'm not. The AI filled a gap with a guess.
Then a harder test: a question without my name. "Who offers free SQL Server mentorship?"
Gemini in its fast mode (Flash-Lite) and Copilot in Auto mode: big communities and user groups. Not me.
Copilot in Think Deeper mode, signed out: my free mentorship program was the first answer, with the right details, taken from my own page. That was five days after I added my site to Bing.

And one funny moment. Gemini Pro, on my wife's phone with her account, answered: "The most accessible and high-level SQL Server mentorship you have is actually right in your own home." It knew me because it knows her, not because it found me on the web.

The lesson: when you test, sign out, or use a private window. Otherwise the AI answers about the people it already knows.
Am I at the top of every result? Not yet, not 100%. But in the searches I did, my name now appears among the first answers. And this was one evening of tests, not a study. I'll keep testing.
And remember Marina's 82%. My site fixed only the part I control: what I say about myself. The bigger part is what other people write about me. That part is still open for me.
Can AI get your product wrong, even from your own documentation?
This was the most interesting result, and it happened inside Microsoft's own world.
I asked Copilot (Think Deeper): "Is Azure SQL Hyperscale a good fit for a small database?" Hyperscale is a type of Azure database service. I wrote about it in October: it's not only for big databases.
Copilot answered: "Usually not by default." Its advice was to start with another option, General Purpose. And it cited Microsoft's official documentation.

But that isn't exactly what we recommend today. Microsoft's documentation calls Hyperscale "the recommended and default service tier for all new and modernizing" workloads of this kind, and "a great option for any workload, not just for large databases".
Here's the important part: the AI didn't make a mistake. It based its answer on the text it found in our documentation. The way some of that text is written, and where other text sits on the page, doesn't support the way Microsoft positions the product now.
I shared it with the product team. They saw the same answers, and the wording is being reviewed.
So here's the heads-up. Look at your own pages, your blog, whatever you have published. If you have a business, of any size, look at your website. What message do you want to pass? Is that message being passed on correctly, to people and to AI?
Is LinkedIn enough for AI to find you?
I've had my LinkedIn profile for many years. I thought it gave me good visibility. But LinkedIn is a more closed platform, and AI tools don't really get all the data that is there.
I asked my AI tool to explain why, in plain words. Here is what we found. (When I say "we", I mean me and my AI tool. It sounds strange, I know. I got so used to working this way that I talk in the plural.)
Much of LinkedIn only shows after you sign in. Crawlers, the automated readers AI tools use, don't sign in.
LinkedIn limits which automated readers may read its pages. Every website has a public file of rules for them, called robots.txt. It works like a sign at a building's entrance that says which rooms visitors may enter. When my AI tool tried to read LinkedIn's crawling rules for this post, it was refused.
Some basic public profile information can still appear in Google and Bing. But your posts, your articles and the details of your experience mostly stay behind the door.
So keep your LinkedIn. It's still important for your network. But don't depend only on it. Put your real answers on a page you own, where anyone, and any AI, can read them.
Find what’s slowing your GTM.
Most GTM problems aren’t one big failure. They’re one or two weak spots quietly dragging down the rest.
The free GTM Audit scores your go-to-market across four dimensions, then sequences a 90-day roadmap for your stage.
Drop your URL. Get your score in about 10 minutes.
How can you check if AI sees you, or a business you know?
Open your favorite AI tool. Sign out, or use a private window. If you can, try a fast mode and a deep mode, because my results changed between them. Then ask:
"What can you tell me about [your name], [your job] at [your company or city]?"
"Who offers [what you do] in [your city, or online]?"
"What do you know about [business name] in [city]? What do they sell?"
"What is the best [type of product or service] for [a need your customers have]?"
"Is [your product] a good fit for [a question your customers often ask]?" Then compare the answer with what your own website says.
For each answer, check three things: Is it right? Is an old profile speaking for you? Is the AI guessing?
You can do this for a friend too. Maybe you know someone with a small business who has no idea this matters. You can help them be more visible, and that can help their business.
And if you want, ask about me too. I'd like to know what you find.
On Friday: a step-by-step guide to being visible to AI. Marina's steps and every step I took, in order, so you can follow them too.
Your turn
Open a private window right now. Ask your favorite AI: "Who offers [what you do] in [your city, or online]?" Then reply and tell me: did your name show up? I read every reply.
See you next time.
Glossary
Term | What it means |
|---|---|
AI search | Asking an AI assistant (Gemini, Copilot, ChatGPT, Claude) a question instead of typing it into a search engine. The AI reads web pages and answers with a few names or a summary. |
Bing Webmaster Tools | Microsoft's free tool to tell the Bing search engine your site exists and which pages it has. Copilot uses Bing to search the web. |
Crawler | An automated program that visits web pages and reads them, for a search engine or an AI tool. Like a library assistant who skims every page to build a catalog. It doesn't sign in to any account. |
Earned media | What other people write or say about you: articles, lists, forums. The opposite of owned media, your own website and posts. |
Google Search Console | Google's free tool to tell Google your site exists and which pages it has. Like giving the post office your new address. |
Hyperscale | A service tier (a type of plan) of Azure SQL Database, Microsoft's cloud database service. |
Private window | A browser window that doesn't use your account, history or cookies. Good for testing what a stranger would see. |
robots.txt | A public file on every website with rules for crawlers: which pages they may read and which they may not. Like a sign at a building's entrance that says which rooms visitors may enter. |
Signed out | Using a tool without logging in to your account, so its answer doesn't use what it already knows about you. |
Silicon Valley Girl | The YouTube channel, podcast and brand of Marina Mogilko. |
Prefer to listen? Two AI hosts debate this post, one for it and one skeptic, on our podcast, We The People, in Tech or Not: The Debate: listen here
How I write: every post starts from my own notes and experience. I use AI tools to help draft and edit, and I review every word before it goes out.
Disclaimer: The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of Microsoft. The author is a Microsoft employee. All technical details and product behaviors described are based on publicly available Microsoft documentation as of November 2026 and on the author's own tests, and are intended to provide a high-level overview for educational and planning purposes.


