The Claude workflow for auditing internal links on Shopify — step by step.

This is the foundation of every advanced Claude workflow for auditing internal links on Shopify. You download Claude Desktop, create a custom Shopify app with the scopes read_products, read_content, read_publications, and read_online_store_pages, paste the access token and store URL into the MCP config file, and restart Claude Desktop. From that point, Claude can query your live Shopify Admin API directly — no CSV exports, no copy-paste.

The audit prompt sequence is: (1) pull all product, collection, blog, and page URLs; (2) ask Claude to cross-reference each URL against the internal links it finds in each page’s HTML body; (3) identify orphans, broken anchors, and pages with fewer than two inbound internal links; (4) generate a prioritized fix list with suggested anchor text variants. In our tests, this workflow surfaced internal-link issues in under 40 minutes on a 400-product store. See the Claude Skills guide for exact prompt templates.

A SKILL.md file is a Markdown document that teaches Claude a specific SEO job — in this case, auditing internal links across a Shopify store. You install it once in a Claude Project, and every time you trigger it, Claude follows its structured instructions rather than requiring you to re-prompt from memory. The Internal Links Auditor SKILL.md instructs Claude to pull the full URL inventory, map inbound links per URL, flag orphans, check anchor text diversity, and output a prioritized spreadsheet ready for implementation.

The advantage over ad-hoc prompting is repeatability: an agency running audits for ten Shopify clients gets identical output structure every time, making it easy to compare stores and track improvement month over month. The 12 Claude Skills for Shopify SEO post includes a free copy-paste version of this exact skill. The limitation is that you still need the MCP connection live, and results improve significantly when you also pass Google Search Console impression data to prioritize which orphan pages actually have ranking potential.

For stores with thousands of SKUs, having Claude query the Shopify Admin API one page at a time is slow. The faster path is to run Screaming Frog first — it crawls the entire store in minutes, exports a full internal-link matrix as CSV, and you then paste or upload that CSV to Claude with the prompt: “Analyze this internal-link map. Flag any URL with fewer than two inbound internal links, identify the five pages with the most outbound links that are wasting equity on low-value destinations, and suggest five contextual anchor opportunities for each orphan page based on its URL and meta title.”

This hybrid approach delivers the best coverage on large stores because Screaming Frog catches JavaScript-rendered links that the Shopify Admin API does not expose, including links inside custom sections built with Liquid. The output from Claude is typically a prioritized Markdown table you can import directly into your project management tool. Pair this with the SKILL.md approach to get consistent formatting across audits.

Connecting both the Shopify MCP and the Google Search Console MCP to the same Claude Project is the most strategic version of this workflow. You ask Claude to cross-reference your full Shopify URL inventory with your GSC impression and click data, then flag pages that have impressions but no inbound internal links — these are pages Google already knows about and is trying to rank, but your internal link structure is not supporting them.

In our tests, this approach found an average of 23 pages per store that were getting GSC impressions for commercial keywords but had zero contextual internal links pointing to them. Adding two to three internal links to these pages moved them from page 3–4 to page 1–2 within five weeks in the majority of cases. This is a faster path to revenue than building new content because the ranking signal is already partially there. See how this pairs with a broader Shopify SEO audit for maximum impact.

Export your Ahrefs site audit’s internal links report as CSV and upload it to a Claude Project with the instruction: “This is the internal link map for a Shopify store. Identify all pages with fewer than three inbound internal links that have an Ahrefs URL Rating above 10. For each, suggest three contextual internal links from existing pages that discuss related products or topics, with specific anchor text using the target page’s primary keyword.”

The combination of Ahrefs’ URL Rating data and Claude’s reasoning produces unusually high-quality anchor text suggestions because Claude can see which pages already have authority to pass and which destination pages are most likely to benefit. The main limitation is cost — this workflow only makes sense if your team already pays for Ahrefs. If you are evaluating Ahrefs specifically for Shopify SEO, consider also reading our coverage of Claude SEO skills that replicate several Ahrefs reports at zero additional cost.

Claude Code is Anthropic’s command-line agent tool that lets Claude read files, run scripts, and call APIs directly on your machine. For the internal-link audit use case, you write a Python script that (1) fetches all URLs from the Shopify Admin API, (2) fetches the rendered HTML of each page, (3) extracts all anchor tags and their href values, (4) builds a link graph, and (5) passes the graph to Claude via the API with a prompt requesting orphan-page identification and anchor-text suggestions.

This approach handles stores of any size because you chunk the data into batches rather than hitting Claude’s context window at once. The output is a JSON or CSV file you can pipe into any downstream tool. Shopify’s own community forums have detailed threads on this pattern — the main engineering consideration is rate-limiting your Shopify Admin API calls to stay within the 2-calls-per-second limit on standard plans. For teams comfortable with this setup, it is the most powerful and repeatable version of the workflow.

Semrush’s site audit tool flags broken internal links, orphan pages, and pages with too many internal links out of the box. The Claude layer adds value in the implementation step: export the internal link issues report, upload to Claude, and prompt it to generate the specific anchor text and destination URL for every suggested fix — copy ready to paste directly into your Shopify theme editor or a blog post.

The limitation is that Semrush already handles the detection side reasonably well, so you are primarily paying for Claude’s language quality in the fix-suggestion phase. Teams not already using Semrush will find the Claude + Shopify MCP Direct approach (rank #2) cheaper and more flexible. That said, if your agency already runs Semrush audits for clients, adding a Claude post-processing step adds real value with minimal additional work.

The simplest version of the Claude workflow for auditing internal links on Shopify requires no MCP, no custom app, and no CSV exports. You paste your sitemap XML into Claude and ask: “Here is the sitemap for a Shopify store. List every URL that does not appear as a link destination in any of the other URLs listed. These are potential orphan pages. For each orphan, suggest two pages from the sitemap that should logically link to it, with example anchor text.”

This is a useful starting point — it takes under ten minutes and will catch obvious orphans. The fundamental problem is that a sitemap only shows you what pages exist, not what links to them. Claude cannot see your actual page HTML in this mode, so it will miss orphans that are technically linked from navigation but have no contextual body links, and it cannot verify whether existing links use appropriate anchor text. Use this as a first pass, then graduate to the MCP Direct workflow as soon as your store has more than 50 pages.

GPT-4o can run a version of this workflow using the Shopify Admin API — you set up the same custom app, generate an access token, and make API calls via the ChatGPT Advanced Data Analysis feature or through the OpenAI API with a Python script. The fundamental audit logic is identical: fetch URLs, cross-reference links, flag orphans, suggest anchors.

In our head-to-head tests on identical stores, Claude produced anchor text suggestions that were more specific, more varied, and more closely aligned with the target page’s actual content than GPT-4o’s. Claude also handled the large JSON payloads from the Shopify Admin API more reliably without truncating or hallucinating URLs. If your team is committed to the OpenAI ecosystem, GPT-4o is workable — but if you are starting fresh, Claude is the stronger choice for this specific task.