SEO Strategy & AI Workflow — 2026-04-05

Overview

Weekly call between Mark Hope and Ben San Fratello reviewing the current SEO state of ^[1] and demonstrating a repeatable AI-driven workflow for generating keyword-optimized service page content. Two critical gaps were identified: absence from the Google map pack and thin, misaligned service page content. The session also introduced Claude Projects as a persistent knowledge base for client SEO work.

Attendees: Mark Hope, Ben San Fratello
Recording: Fathom

Key Findings

Site Health (Ahrefs)

Content Misalignment

Traffic is being driven almost entirely by informational blog posts, not core service pages:

• Top traffic pages: Tyvek article, LVL beam article, copper gutters article
• Near-zero traffic pages: Roofing, Siding, Decks, Windows

The service pages exist but carry minimal copy (~200 words on the roofing page) and lack local keyword targeting. The site is not optimized for its primary business.

Local SEO Failure

• GMB map pack: Exterior Renovations does not appear for searches like "roofing Madison" — a critical gap for high-intent local leads
• Moz Local: GMB profile is not connected; several social/directory networks (Facebook, BBB, etc.) are unlinked
• Comparison: Asymmetric's own Moz Local profile shows GMB, LinkedIn, Facebook, Instagram, and Yellow Pages all connected — this is the target state

Key Decisions

1. Local SEO is the top priority. GMB access must be obtained and connected in Moz Local before other work proceeds.
2. Service pages need full rewrites targeting local keywords (e.g., "Roofing Contractor in Madison, Wisconsin") with proper H1/H2/H3 hierarchy, expanded word count, and FAQ sections.
3. Claude Projects will be the standard workflow for generating SEO content briefs — one project per client, loaded with GSC/Ahrefs/SpyFu exports as PDFs.
4. Developer handoff process: AI-generated briefs go to the developer as a draft page for client approval before going live.

Action Items

• [ ] Ben — Install Awesome Screenshot Chrome extension for full-page captures
• [ ] Ben — Obtain GMB access from Exterior Renovations client
• [ ] Ben — Connect GMB profile and other social networks in Moz Local
• [ ] Ben — Export GSC queries (3-month) and Ahrefs organic keywords; convert to PDF; upload to Claude project
• [ ] Ben — Use Claude project to generate SEO content briefs for all core service pages: Roofing, Siding, Decks, Windows (each H1 should include "Madison")
• [ ] Ben — Send AI-generated draft pages to client for approval; publish on sign-off
• [ ] Mark — Share Claude "Exterior Renovations" project link with Ben (completed during call)

AI Workflow Demonstrated

Claude Projects for Client SEO

Mark walked through building a Claude Project as a persistent, client-specific knowledge base. The workflow:

1. Create a Claude Project named for the client (e.g., "Exterior Renovations")
2. Write project instructions describing the client, their relationship with Asymmetric, and the engagement goals
3. Upload data exports as PDFs — Claude handles PDFs better than spreadsheets:
   - Google Search Console → export → Google Sheets → download as PDF
   - Ahrefs Organic Keywords export → PDF
   - Ahrefs Backlinks export → PDF
   - SpyFu domain report → PDF
   - Screenshots of Ahrefs overview, competitor pages, etc.
4. Ask questions or request deliverables — the project knowledge base informs every response without needing to re-upload or re-explain context
5. Generate a content brief — prompt: "Write an SEO-optimized service page for Roofing in Madison, Wisconsin based on this data"
6. Review output — the demo produced an 856-word roofing page (vs. 200 words currently) with proper heading hierarchy, local keyword density, schema markup notes, and an FAQ section
7. Send to developer as a draft page for client review before publishing

Why PDFs? Mark noted that Claude handles PDF uploads more reliably than raw spreadsheet formats, even though it seems counterintuitive.

Team access: Projects can be set to team visibility so all Asymmetric staff can access and contribute.

See also: ^[2] for the generalized workflow.

AI Chat Management Best Practices

A secondary topic covered managing long AI conversations to avoid losing work.

Problem: Context windows have hard limits. When a chat fills up, the tool stops accepting input — and if you haven't summarized recently, starting a new chat means losing all prior context.

Solution — the Summarize habit:

• Type summarize this chat every ~30 minutes during a working session
• The summary becomes a searchable anchor point (use Ctrl+F to find it later)
• When approaching the context limit (you can ask: "how much context window do I have left?"), copy the latest summary and paste it into a new chat to continue seamlessly
• Summaries can also be pasted across tools (e.g., from Claude to ChatGPT when token limits are hit)

Token limit reality: Even paid Claude accounts have hourly/weekly limits on premium models. Having a second tool (ChatGPT) as a fallback is practical — paste the Claude summary in and continue.

See also: ^[3]

Content Brief Output — Roofing Page (Example)

The demo brief generated for the Roofing service page included:

The brief was generated by uploading the GSC export to the Claude project and prompting: "Make me a landing page to optimize for these keywords."

Related

• ^[4]
• ^[2]
• ^[3]
• ^[5]
• ^[6]