I run a grooming brand on Shopify. It's just me — no SEO team, no marketing hire, no agency. A few months ago I started duct-taping AI agents together to handle my SEO. Some of it works surprisingly well. Some of it was a waste of
time. Here's all of it.
Why I Built This
I sell men's grooming supplies. Think shaving creams, shaving soaps, deodorants, skincare, post shave products. I have a laptop and stubborness.
The thing that pushed me over the edge: I realized more and more people are discovering products through AI. Not just Google anymore — ChatGPT, Perplexity, Google's AI Overviews. Someone asks "what's the best tallow shaving soap?" and an AI just... picks brands to recommend. I had no idea if my brand was one of them. That bugged me enough to do something about it.
What I Stitched Together
Claude Code is the brain. It's Anthropic's CLI tool — basically a terminal where Claude can use tools, read files, call APIs. Everything else plugs in through MCP servers, which are connectors that let Claude talk to external services.
Here's what's plugged in:
a. DataForSEO — keyword data, rank tracking, backlink checks, and the thing I actually care about most: their AI Optimization API that tracks LLM brand mentions
b. GA4 or Google Analytics 4 — traffic, conversions, the usual
c. GSC or Google Search Console — impressions, clicks, indexing
d. Airtable — I track content and campaigns here
On top of that I built custom "skills" — basically reusable recipes that combine prompts with tool calls. One skill generates SEO pages from live SERP data. Another runs competitive analysis. They're not magic. They're templates that save me from doing the same thing over and over.
Scheduled jobs run on GitHub Actions. First of every month: pull rank snapshots, check LLM mentions, flag anything weird in analytics. I wake up to a report instead of spending a morning pulling data.
The Part Nobody Else Is Doing
Here's where it gets interesting. I've been tracking which AI models mention my brand. Monthly. Like rank tracking, but
for AI answers.
Three months of data and the picture is wild:
Google AI Overviews loves me. 21 mentions, 758K impressions. For a small brand, that's surprisingly strong. When someone searches a grooming question and Google shows an AI answer, my brand comes up regularly.
ChatGPT pretends I don't exist. Same queries, same niche — near zero mentions.
And here's the kicker: my biggest competitor? Opposite pattern. They dominate ChatGPT but barely show up in Google AIO.
Same products. Same keywords. Completely different results depending on which AI is answering.
I don't fully understand why yet. But I know it means "optimizing for AI" isn't one thing. Google AIO and ChatGPT are different games with different rules. Most brands don't even know which game they're playing.
What Actually Saves Me Time
The daily/weekly snapshots. Before this, I'd manually check rankings when I remembered to, which was never consistently. Now it just happens. I can actually see trends instead of guessing.
Programmatic content generation. I built a skill that pulls SERP data for a keyword, looks at what's ranking, and generates a long-form page. I structure the content in ~500-word chunks that each stand on their own — my theory is that LLMs are more likely to cite a self-contained chunk than pull from a wall of text. The pages still need editing, but starting from a researched draft instead of a blank page saves hours.
GA4 anomaly checks. I used to notice traffic drops a week late. Now I catch them the next morning.
What Doesn't Work (Or I Screwed Up)
I built content tools before measurement tools. Classic mistake. I was cranking out pages before I had any way to track if they were working. Should've set up tracking first, then created content. Measure, then optimize. I did it backwards.
Early programmatic pages were too generic. The first batch read like they were written by an AI because, well, they were.
The pages that actually rank have real opinions, specific product knowledge, and the kind of detail that only comes from someone who's used these products for years. The AI gives me structure. I give it soul. That split works. Full autopilot doesn't.
I should've started LLM tracking a year ago. Three months of data tells a story. Twelve months would tell a much better one. If you're reading this and thinking about doing something similar — start now. Future you will be grateful.
What It Costs
a. Claude Code: $50-100/month depending on how much I'm building
b. DataForSEO: $30-50/month in API credits
c. GitHub Actions: free tier
d. GA4 / GSC APIs: free
Call it $150/month. That's less than one hour of an SEO consultant's time, and the system runs 24/7.
I'm not saying it replaces expertise. I still make all the strategic calls — what to write about, which pages to prioritize, when to fix technical issues vs. create new content. But it handles the tedious stuff so I can focus on the decisions that actually matter.
What's Next
I want to figure out what makes an LLM cite one brand over another. Is it structured data? Freshness? Sentence structure? Brand mentions on Reddit? I don't know yet. But I have the tracking in place now, so I can start testing.
I'm going to run experiments such as change content structures, add schema markup, test different formats — and track whether LLM citation rates change. If it works, I'll share the data here.
I'm Sri. I run whollykaw, a grooming brand I built from nothing. I'm documenting this whole AI + SEO experiment in public. If you're doing anything similar, or you think I'm completely wrong about something, I'd love to hear from you.