1. Home
  2. Work
  3. SEO and AI search for an Amazon agency
Case study · SEO, AI search, GoHighLevel

Rebuilding an Amazon agency's site for buyers and AI search.

From July to October 2026 I ran marketing for a Los Angeles agency that grows brands on Amazon. In 12 weeks I set the search baseline, rebuilt the keyword map around buyers, shipped the technical SEO fixes, opened the site to AI search engines, and built the agency's first GoHighLevel lead magnet funnel.

  • ClientAmazon growth agency, Los Angeles (name withheld)
  • DatesJuly to October 2026
  • My roleDigital marketing manager, SEO lead
  • StackAstro, Sanity, Cloudflare, GoHighLevel, GA4, Search Console, Semrush
2days
To a full search baseline across four tools
174
Buyer keywords put into rank tracking
5
AI search crawlers confirmed open
1st
GoHighLevel funnel and workflow, built from a blank account

In short

The agency had just moved its site to Astro and Sanity. I measured where search stood, pointed the content plan at brand owners instead of Amazon shoppers, fixed what kept new posts out of Google's index, and set a day-one number for AI search traffic. When the engagement closed, the founder thanked me for three things by name: the website rebuild, the technical SEO, and the AI search baseline.

What was the starting point?

The agency had strong case studies and a new site, but its search traffic came almost entirely from people typing the brand name. The blog ranked for topics Amazon shoppers search, like promo credits and renewed products. The people who hire an Amazon agency are brand owners and manufacturers, and they search for different things.

How did you build the baseline?

On day two I pulled every source of search data the agency had and put it in one document, three days before it was due:

  • 490 tracked keywords from SE Ranking
  • Three months of Google Search Console clicks, impressions, and positions
  • 30 days of GA4 sessions by channel and page
  • A Semrush scan of the domain and four competitors

One finding shaped the strategy. Google showed an AI Overview on 443 of the 490 tracked searches. Informational posts would lose the click to that summary, so the plan moved toward commercial and comparison searches, written answer-first.

How did you choose the keywords?

I mined 119 buyer keywords in Semrush, such as "amazon marketing agency" and "amazon account management agency". None of them were in the agency's tracking. I sorted them into track, hold, and cut, mapped each cluster to the page that should rank for it, and presented the plan to the founder as a nine-slide deck.

After an ideal-customer call with the founder, I added a manufacturer and account-reactivation cluster and checked every new term for real search volume. In August I loaded 174 keywords into Semrush Position Tracking, so the team could watch the buyer terms each week.

What went into the website rebuild?

I worked on the new Astro and Sanity site from the first week. I reviewed the four paid search landing pages and gave the founder a seven-slide feedback deck with specific fixes for proof, pricing, and calls to action. I drafted the free Amazon audit offer, wrote copy for eight industry pages, and briefed and drafted blog posts aimed at buyer searches with hero images.

Which technical SEO fixes shipped?

In September, new blog posts were not getting into Google's index. I traced the cause and fixed it in Sanity and the site code:

  • Set a named author on 17 posts, so their BlogPosting schema points to a person instead of an organization
  • Took a duplicate legacy post out of the index
  • Added per-post lastmod dates to the sitemap, pulled from Sanity
  • Added 301 redirects for two legacy URLs in the Cloudflare worker
  • Requested indexing in Search Console for the priority pages

I also checked every sitemap URL for status codes, canonicals, noindex tags, and title and meta lengths.

How did you set up the AI search baseline?

AI search engines can only cite pages they are allowed to read. I tested the site with each crawler's user agent. The search and answer bots for ChatGPT, Claude, Perplexity, and Bing all got through. Only the bots that collect AI training data were blocked, which was the agency's choice.

Then I set the number to beat. GA4 already sorts visits from AI assistants into their own channel, so I recorded those sessions for the first 30 days as the starting figure for AI citation work, alongside the AI Overview count from the keyword set.

What did you build in GoHighLevel?

The agency's GoHighLevel account was a fresh install with no forms, funnels, or workflows. I built its first lead magnet: a casebook landing page with a form, and a delivery workflow that tags each new contact and sends the casebook by email. It gave the site a second way in for visitors who are not ready to book a call.

Results

In 12 weeks the agency went from scattered data to one baseline, a buyer keyword set under weekly tracking, a cleaner index, open access for AI search crawlers, and a working lead magnet funnel. Traffic and revenue figures belong to the client, so I do not publish them here. I am happy to walk through the method on a call.

Next step

Want results like this on your project?

Tell me what you are working on and where it is stuck. I reply within 24 hours with questions or a clear next step.