---
title: 'Obsidian AI SEO Increased My Traffic by 577%?!'
source: 'https://youtube.com/watch?v=SxdeynBLBQs'
video_id: 'SxdeynBLBQs'
date: 2026-11-17
duration_sec: 532
channel: 'Julian Goldie SEO'
---

# Obsidian AI SEO Increased My Traffic by 577%?!

> Source: [Obsidian AI SEO Increased My Traffic by 577%?!](https://youtube.com/watch?v=SxdeynBLBQs)

## Summary

This video demonstrates how using Obsidian as a knowledge base for AI agents dramatically improved SEO performance, resulting in a 577% increase in traffic. The creator explains the setup, the importance of unique data, and introduces a custom tool called Search Agent OS to identify content gaps and avoid keyword cannibalization.

### Key Points

- **Traffic Increase Claim** [00:02] — The creator claims a 577% increase in Google clicks, attributing it to using Obsidian as a 'brain' for AI agents, not a smarter AI model.
- **Common AI SEO Mistake** [00:43] — Most people use AI by prompting a chatbot to write an article, but this fails because the AI has no memory of the user's unique data, leading to generic content that doesn't rank.
- **Obsidian as a Vault** [01:09] — Obsidian is a free note-taking app that stores notes as plain text markdown files in a folder (vault), which AI agents can easily read, making it ideal for feeding AI with unique data.
- **PARA Organization System** [02:37] — The vault is organized using the PARA method (Projects, Areas, Resources, Archives) from Thiago Forte, with Resources being the most important for storing SEO briefs, style guides, and playbooks.
- **LLM Wiki Layer** [03:02] — A wiki layer is added with one note per person, tool, company, site, and idea, all linked together, following the 'LLM wiki' pattern discussed by Andre Karpathy, allowing the AI to follow links instead of guessing.
- **Real Brief Example** [03:15] — A real brief from the vault for the keyword 'best AI SEO tool' includes eight supporting keywords, search intent, word count, content angle, and a full H2 outline, giving the agent a structured starting point.
- **Traffic Numbers Breakdown** [03:45] — Over a 28-day window (June 5 vs Oct 24), main site clicks went from 2,776 to 18,790 (6.8x), impressions from 113,070 to over 1.1 million (10x), and average position improved from 10.8 to 8.
- **Three Key Differences** [04:27] — The difference is attributed to: 1) Memory (agent reads vault before writing), 2) One keyword per page (no stuffing), and 3) Stopping guessing (using data to decide what to write).
- **Search Agent OS Overview** [05:06] — A custom tool called Search Agent OS connects to Google Search Console (read-only) to show rankings, clicks, and opportunities, with features like animated rankings, impression heat maps, and searchable tables.
- **Opportunities View** [05:31] — The opportunities view checks existing pages and sitemaps to avoid suggesting keywords already covered, providing lists for content gaps, page upgrades, and already covered keywords.
- **The Full Loop** [06:12] — The workflow: Search Agent OS finds a gap, the agent reads the Obsidian vault, writes a unique page, publishes it, and the new data flows back into the system for continuous improvement.
- **Keyword Cannibalization Mistake** [06:52] — A major mistake was writing new articles for keywords that Google was already ranking a different page for, causing self-competition. The fix is to check which page Google ranks for a query before creating new content.
- **Pro Tips** [07:31] — Pro tip 1: Put case studies and playbooks in the vault as they are unique data. Pro tip 2: Link everything with notes to help the agent follow connections and understand the business better.

### Conclusion

The video concludes that the key to successful AI-driven SEO is feeding the AI with unique, proprietary data stored in a structured knowledge base like Obsidian, and using tools to avoid common pitfalls like keyword cannibalization. The bigger lesson is that the skill lies in knowing what to feed the AI, not just using it.

## Transcript

Obsidian AI SEO increased my traffic by 577 % What if the next article you write is the one holding your site back? Same site, same niche, no new staff. Our Google clicks still jumped 577 % and it wasn't a smarter AI model. It was a free
notes app plus the agent OS that reads it. Stick around because later I'll show competing against our own pages. Most people make it every week and never people make it every week and never notice that 577 % is real and I'll walk
numbers in a few minutes. The short version, we gave our AI agents a brain Obsidian. I'll show you how we set it up for SEO and the dashboard inside agent next. Here's how most people do AI SEO right
now. Open a chatbot, type write me an article about something, copy, paste, publish, hope, and it doesn't rank. Why? Because the AI has no memory. It doesn't know your results, your clients, your stories, your voice. So it writes the
has seen that article a million times and at this point it could just write it agent reads your brain first, your case studies, your keyword briefs, your writes something only you could have written. That's the whole game. Unique
data wins. So what actually is Obsidian? It's a free note taking app, but under the hood it's just a folder of plain text notes on your computer, markdown files. That's it. Obsidian calls that
folder a vault and that's exactly why AI agents love it. Claude, Code, Hermes, Codex, pretty much any agent that can open files can read those notes. No database, no complicated setup. Your notes stay as simple files you own. Now
the whole idea here is simple. AI writes better SEO content when it reads your stuff first. If you want to learn how to actually apply AI to your SEO work like this, grab my free AI SEO course. It's a free course on using AI for SEO. It's
for business owners, marketers, and anyone tired of AI spitting out generic articles. Tools like Obsidian and AI agents only help if you know how to point them at the right job. That's the real skill. The link to the free AI SEO
course is the description. And if you'd rather have experts look at your site, this ties straight in. Content built from your own data is exactly what helps you show up, not just on Google, but inside AI answers, too. So book a free
AI brand visibility strategy session. You'll get a custom visibility plan. don't across Google AI overviews, ChatGPT, Perplexity, and Gemini. You'll get a competitor gap analysis, and you'll get direct answers to your
questions from a strategist. It's 100% free with no obligation. The link to book your strategy session is in the description. Okay, back to the vault. Here's how ours is organized. We use a simple system called PARA. That stands
for projects, areas, resources, and archives. It comes from productivity writer Thiago Forte. Projects are things with a finish line like our multi-site SEO operation with a checklist for every keyword run. Areas are the ongoing parts
of the work like Goldie Agency, the SEO agency side of the business, and resources is the most important one. That's where our SEO briefs, writing style, conversion playbooks, and memories live. Everything we've learned
captured automatically. Then on top of that, there's a wiki layer. One note per person, tool, company, site, and idea all linked together. This is the pattern AI researcher Andre Karpathy talks about. He calls it an LLM wiki. You let
you let it follow links instead of guessing. So what does that look like in real life? Here's a real brief from our vault for the keyword best AI SEO tool eight supporting keywords, the search intent written out in plain English, a
word count, the content angle, and the full H2 outline. H2s are just the main agent doesn't start from a blank page. It starts from this. And because the and our proof, the agent pulls in stuff Google can't find anywhere else. Our own
numbers, our own stories, our own case studies. Now let's talk numbers. Every one of these is a 28-day window from Google Search Console, June 5th versus October 24th. Our main site first, clicks went from 2,776
to 18,790. That's about 6.8 times more. Impressions, which is how often you showed up in search, went from 113,070 to over 1.1 million, nearly 10 times more. Average position went from 10.8 to
8, so on average we moved onto page one. Second site, Julian Goldie AI brand new, zero clicks, zero impressions. Now it's at 1,311 clicks and 89,749 impressions. And our third site went
impressions. And our third site went from 1,152 clicks to 2,230, nearly doubled. So what actually made the difference? Three things. Number one, memory. Before writing, every article starts with the agent reading the vault,
not a prompt from scratch. That's why the content has real data in it. Number two, one keyword, one page. Every keyword we target gets its own dedicated page. We don't stuff 10 keywords into one article. Number three, stop guessing
I promised you at the start. I'll break it down in a second because first you need to see the tool we built to fix it. It's called Search Agent OS and it sits You connect your own Google Search Console and the connection is read only.
So it can look at your data, but it can't change anything on your site. Then you pick your site and a time range, 7, 28 or 90 days. The overview shows you everything at a glance, animated rankings, daily clicks, which keywords
are on page one, page two and page three, an impression heat map, your best weekdays, your top pages. Then you've got searchable tables for rankings, any keyword and send it straight into the Content Studio. And here's my
favorite part. The opportunities view won't suggest a keyword you already have a page for. Before it suggests anything it checks two things. First, the pages Search Console already shows for that keyword. Second, your site's sitemaps,
so brand new pages count too. Then you get three lists, content gaps, which are keywords with no page yet. So these are pages you need to create. Page upgrades could be optimized better and pushed higher and already covered, which you
leave alone because they're doing fine. And if it ever gets one wrong, you press mark as built and it won't suggest that one again. Now here's the full loop inside agent OS, search agent OS finds a gap from your real search console data.
Your agent reads your Obsidian vault, the brief, the style, the proof. It writes a page with data nobody else has. You publish it and get it indexed. Then the new numbers flow back into search agent OS and it goes again. Agent OS can
even turn selected Obsidian notes straight into SEO content. Pick the notes, hit create. That's how our main site went from 2,776 clicks to 18,790. It wasn't one magic article, it's a loop
that keeps improving. And the same idea applies to GEO and AEO, which is just showing up inside AI search engines and AI answers, not only the classic blue links. Okay, the mistake. This one really set us back. Early on, we'd look
keyword wasn't there, so we'd write a brand new article for it. Turns out Google was already ranking a different page of ours for that search. So now we keyword. We were competing against ourselves. That's called keyword
cannibalization and it can drag both pages down. The fix is simple. Don't trust your own list. Check the page Google actually ranks for that search. console. Open the performance report, filter by the search query, then click
pages Google is already showing. If one's there, upgrade that page instead of writing a new one or let search agent OS do that check for you automatically, which is exactly why we built it. Now two pro tips before you go. Pro tip one,
It's the most important part of the vault. Put your case studies, your playbooks in there. That's the stuff no other site has, so it's the stuff that makes your content different. Pro tip two, link everything. Give every
own note and connect them. Link your briefs to your style notes and your proof. That way your agent follows links to the right information instead of your agent suddenly knows a lot more about you. So here's the bigger lesson
hard part, knowing what to feed it is. If you want to get better at that, grab my free AI SEO course. It's a free course on using AI for SEO, and it's a great next step after what you just saw. Start small, set up a simple vault,
your agent and see how the output changes when it has your own data. The more you practice, the faster it clicks. The link to the free AI SEO course is in the description. And if you want expert eyes on your own site, book your free
strategy session. You just saw how easy it is to miss what's really happening in your search data, like two pages fighting over one keyword. A strategist get a custom plan for your brand. You'll get competitor insights so you know
who's showing up where you're not on Google and inside AI answers, and you'll walk away with clear recommendations on what to do next. It's 100% free with no obligation. The link to book your strategy session is in the description.
