The SEO Francisco AI Search And SEO Guide For Claude
Download the free SEO Francisco guide for using Claude in AI search and SEO workflows: 15 skills, prompt packs, scheduled audits, and release-day QA gates.
TL;DR: I built a free SEO Francisco AI Search and SEO guide for Claude. It includes 15 reusable skills, a prompt pack, a skill pack, scheduled audit cadences, and the release-day QA gates I would use before trusting Claude with real SEO work.
Download it here: PDF guide, prompt pack, and skill pack.
Claude AI Search And SEO Guide
Watch the AI search and SEO guide for Claude
A practical walkthrough of the Claude skills, prompt packs, scheduled audits, and release-day QA gates behind the downloadable guide.
If your AI search workflow starts with "make me a visibility report," pause for a second.
The report can look polished while the evidence is missing. That is the danger with Claude, ChatGPT, Perplexity, Gemini, or any other assistant. The output may sound operational before the workflow has proven what data it actually used.
The better question is simple:
What data did Claude see, and what rule decided the answer?
If you cannot answer that, you do not have an SEO operating system. You have a confident draft.
This guide is my attempt to turn Claude into something more useful for SEO teams: not a magic answer box, not a one-shot report writer, but a set of repeatable evidence-first workflows.
Download The Guide
The free package has three parts:
- The SEO Francisco AI Search and SEO Guide for Claude - the full operating guide.
- Prompt pack - copy-ready prompts for citation gaps, content decay, source packs, internal links, release-day QA, weekly reports, and YouTube metadata.
- Skill pack - 15 reusable skill specs you can adapt into Claude projects, Claude Code skills, or your own agent instructions.
The point is not to collect more prompts. The point is to stop repeating the same setup, same warnings, same source rules, and same QA gates every time you ask Claude to help with SEO.
What This Guide Is Not
This is not a promise that Claude automatically has access to Search Console, GA4, crawlers, logs, or AI visibility platforms.
That distinction matters. Claude can work with external context when you give it files, connect supported apps, use tools, or wire data through MCP-style workflows. Official Claude documentation covers connectors, MCP, tool use, and reusable skills. But access is not magic. If the data was not connected, uploaded, exported, or supplied, Claude should say data needed.
That is why the guide treats every workflow as evidence-first. Before a recommendation is allowed, the prompt has to identify the input it used and the confidence level behind the output.
The SEO Francisco System
The system has four layers:
- Evidence first. Source packs, crawl exports, GSC and GA4 data, server logs, AI answer snapshots, cited domains, screenshots, transcripts, and human notes.
- Skills second. Small reusable jobs with stable inputs, strict rules, and reviewable outputs.
- Schedule third. Daily, weekly, monthly, and release-day checks so the workflow does not depend on memory.
- QA last. No invented metrics, no unsupported citations, no unreviewed AI copy, no publication without media and parity gates.
That last layer is where many AI SEO workflows break. They produce a report, not a decision. They create content, not a release. They generate a chart, not a source trail.
Google's public guidance is still clear on the risk: automation used primarily to manipulate rankings violates spam policies, and scaled content abuse is about large amounts of unoriginal, low-value content regardless of how it is created. That is why the guide is built around evidence, review, and publication control rather than volume.
The 15 Skills Inside The Guide
The reference idea I saw had seven skills. That is useful, but not enough for the way I would actually run an SEO workflow. The SEO Francisco version has 15:
| Skill | What it prevents |
|---|---|
| Brand Brief Keeper | Repeating positioning, tone, proof rules, and claims to avoid. |
| Source Pack Builder | Writing from memory or secondhand summaries without claim boundaries. |
| AI Citation Gap Finder | Confusing competitor mentions with a measured visibility system. |
| Competitor Citation Teardown | Missing the domains and formats AI systems already trust. |
| Prompt Fan-Out Mapper | Building one page that misses the support questions behind the answer. |
| Content Decay Checker | Refreshing pages because they feel old instead of because evidence changed. |
| Internal Link Opportunity Mapper | Adding generic internal links that do not help the reader continue the task. |
| Schema Evidence Auditor | Publishing structured data that does not match visible page facts. |
| Rendering And JavaScript SEO Inspector | Assuming content is crawlable because a browser can show it. |
| AI Answer Snapshot Logger | Losing raw answer evidence before summarizing or reporting. |
| Entity And Author Evidence Gate | Publishing generic claims without author, brand, or experience proof. |
| YouTube SEO And Video Citation Mapper | Treating the video as separate from the article, transcript, and related videos. |
| Release-Day QA Gate | Calling an article done while images, video, parity, build, or indexing are still pending. |
| Weekly AI Visibility Report | Turning noisy prompt runs into fake precision. |
| Memory And Decision Log Keeper | Forgetting corrections, rejected assets, blockers, and final approved paths. |
The skills are intentionally narrow. That is the point. A giant "do SEO" prompt can sound impressive, but it is hard to review. A small skill with a clear input and output can be tested.
The Prompt Rule
Every prompt in the pack follows the same shape:
Role:
Context:
Data:
Rules:
Output:
The rules are the important part. For SEO work, I want prompts to say things like:
- Use only the provided evidence.
- If data is missing, say "data needed."
- Do not invent traffic, rankings, citations, revenue, dates, or conversion impact.
- Label confidence as confirmed, likely, or needs validation.
- Separate facts from recommendations.
- Name the source behind each claim.
- Prioritize by business impact and search risk.
This is boring in the best possible way. It makes the output less theatrical and more useful.
The Scheduled Audits
A good AI SEO system should not depend on remembering which prompt to run. The guide splits the work into four cadences.
Daily
Daily checks are for blockers and sudden changes: publication tasks, indexing issues, crawl problems, GSC drops, and new AI answer snapshots for priority prompts.
Weekly
Weekly checks are for patterns: citation gaps, competitor citations, declining pages, internal link opportunities, schema spot checks, and a report with three priorities.
Monthly
Monthly checks are for drift: brand brief updates, prompt set cleanup, source drift, schema quality, rendering patterns, and whether a content cluster is becoming too similar.
Release Day
Release day is stricter. For SEO Francisco, the gate includes source pack, AI-tells audit, humanization if needed, internal links, related articles, image QA, video QA, English and Spanish parity, build, deploy, IndexNow, and Search Console indexing when available.
That may sound heavy until you compare it with the cost of shipping something that looks done but is missing the actual publication steps.
How To Use This With Real SEO Data
Start with one workflow, not fifteen.
If I were handing this to a new SEO on day one, I would start with these five skills:
- Brand Brief Keeper
- AI Citation Gap Finder
- Content Decay Checker
- Internal Link Opportunity Mapper
- Release-Day QA Gate
Those five clean up most AI-assisted SEO work before you add more automation. They force Claude to know the brand, review AI search evidence, find pages that actually need attention, improve the site graph, and stop publication when a real gate is still open.
After that, add source packs, schema review, rendering checks, YouTube interlinking, and weekly reporting.
The Mistake To Avoid
The mistake is running one prompt on one platform one time and turning the answer into a visibility score.
That is not measurement. It is a sample.
A better audit repeats prompts, saves raw answers, tracks brand mentions, competitor mentions, cited domains, answer changes, and source formats. Then a human reviews what the evidence actually supports.
That is the difference between a prompt list and a publishing system.
Download The Package
You can download the free SEO Francisco Claude guide here:
Start small. Use the five-skill starter set. Then add the rest only when the evidence and QA process are stable.
FAQ
Is the SEO Francisco Claude guide free?
Yes. The PDF guide, prompt pack, and skill pack are available from this article.
Does Claude connect directly to Search Console and GA4?
Only if you provide the data through a verified path such as an export, upload, API workflow, custom connector, MCP server, or another configured tool. The guide does not assume Claude has data it was not given.
What is the best skill to start with?
Start with the Brand Brief Keeper and Release-Day QA Gate. One keeps Claude aligned before work begins. The other stops unfinished work from being treated as published.
Can this replace an SEO strategist?
No. It makes repeatable work faster and cleaner, but a human still needs to decide what evidence matters, what risk is acceptable, and what should not publish.
About the Author
