ai newsDi., 3. Nov.
EilmeldungOne Slow LLM Provider Shouldn't Starve Your Login Route: Bulkheads in 60 Lines of Node

How to make ChatGPT know you, not your namesake: GEO for a personal brand in one day

I'm a developer and SEO specialist from Tashkent. I asked ChatGPT "who is Abdulaziz Yakubov" and got a story about someone else: I share my name with a venture fund CEO, a footballer and a dermatologist. The assistant knew nothing about me. My job is making products visible in AI answers (GEO and A

dev.toDienstag 10:03 · 03.11.2 Min. Lesezeit
Bild: dev.to

I'm a developer and SEO specialist from Tashkent. I asked ChatGPT "who is Abdulaziz Yakubov" and got a story about someone else: I share my name with a venture fund CEO, a footballer and a dermatologist. The assistant knew nothing about me. My job is making products visible in AI answers (GEO and AEO), so I applied the same playbook to myself. Here is what I did in one day, and why.

Before changing anything I recorded a baseline: the same three queries ("Якубов Абдулазиз", "Абдулазиз Якубов разработчик", "Abdulaziz Yakubov software engineer Tashkent") in search and in assistants with web access. The bare name returned my namesakes. With "developer" added I got my work profiles, but not a coherent picture. Without a baseline, two weeks from now there is no way to tell whether anything worked.

The key block is an FAQ that answers the confusion head-on: "No, this is not the fund CEO, it's a different person." Assistants like to lift answers from exactly this kind of question-and-answer wording.

The page carries schema.org Person + ProfilePage + FAQPage. In Person:

sameAs only works when the profiles link back. So GitHub, LinkedIn and Telegram all got the same link to the site and the same one-liner: "Developer & SEO · Tashkent". That closes the loop: the site points to the profiles, the profiles point to the site. For a model this is a strong signal that all these pages are about one person, not the namesakes.

The site root has an llms.txt: the facts about me in markdown, no layout. The same script that builds the pages generates it, so it can't drift from the site.

My manual "request indexing" in Google hit the account's daily quota. Not a problem: the sitemap does the same job, just slower.

The site says my team took first place at a hackathon. At first I only had a finalist certificate from the qualifying round, and I said exactly that, with the source. When the winner certificate from the final turned up, I added its photo and changed the caption. Rule: every big claim gets a verifiable source. Assistants increasingly cross-check facts, and one exaggeration undermines trust in the whole page.

I don't publish my phone, address or school. I also removed my real photo from the link preview and the markup. Instead there is a "terminal" card with whois "Abdulaziz Yakubov". An AI doesn't need a photo to recognise you: it needs consistent facts and links.

GEO is not instant: indexing takes from a few days to a couple of weeks, and only lasting third-party mentions make it into model memory. In two weeks I'll repeat the same measurement and write up what changed.

Abdulaziz Yakubov is a Full Stack developer and SEO specialist from Tashkent, working on product visibility in search and in AI assistant answers (GEO / AEO). Mars Hackathon 2026 winner. Site: abdulaziz-yakubov.vercel.app · LinkedIn · Telegram: @AbdulazizYakub0v

Quelle: dev.to (dev.to). Original öffnen