Internal linking · Image SEO · Structured data · Python · Node.js

Internal SEO Toolkit: Links, Image Metadata and Schema

Three internal tools for the jobs that are pure volume — auditing internal links across supplied pages, writing SEO filenames and alt text into images at scale, and generating paste-ready JSON-LD for eight schema types.

Client
Own tooling · internal
Period
2025 — present
Role
Builder & maintainer
Status
In use · internal tooling
Typographic cover card for the Internal SEO Toolkit — three scripts for links, image metadata and schema.
Three scripts for the work that is volume, not judgement.

Results at a glance

Tools
3

Links · metadata · schema

Link audit output
4-column CSV

Anchor type, status, page

Image pipeline
Lossless WebP

With read-back check

Schema types
8

Org to breadcrumb

The problem

Three jobs on a 100+ page site are pure volume and no judgement: checking internal and external anchors across a list of pages, writing SEO filenames and alt text into hundreds of images, and producing structured data for every page type in a consistent shape.

Done by hand they are slow, and slow work gets done unevenly — the first fifty pages get careful internal linking and the last two hundred get whatever fits. Alt text is the clearest example: it is either written for every image or it is written for none, and the middle state is what actually ships.

Why we decided to solve it

Volume work has a property that judgement work does not: it can be made correct once and then repeated. Every hour spent scripting it is repaid across the next site, and the output becomes consistent, which is worth more than any single careful afternoon.

It also makes the result checkable. A tool that reads back what it wrote can prove the metadata landed; a person clicking through pages cannot.

How we did it

The internal links checker takes supplied pages, classifies each anchor as internal or external, checks HTTP status and exports a four-column CSV. That is enough to find broken links, orphan pages and anchors that point at the wrong target without running a full crawl against a client site.

Write image metadata at scale, then verify it

The metadata processor produces lossless WebP copies and writes SEO filenames, alt text, XMP/EXIF/IPTC and GPS fields in one pass — then reads the result back to confirm each field was written. The read-back is the part that makes it usable on a real site: a silent failure on image metadata is invisible until an audit.

Generate schema from the page type

The JSON-LD generator produces paste-ready structured data for organisation, local business, service, article, how-to, FAQ, breadcrumb and person — the eight types this work actually required. Consistent output per type beats improvised markup per page.

Keep a person on the final check

All three run against supplied input and print what they changed. Nothing writes to a live site silently, and the CSV or image set is reviewed before it is used.

The result

Three jobs that used to consume days now run as one step each, with output that can be diffed and re-run. They were built for the production site work documented elsewhere in this portfolio, where the same tasks repeat across five country sites.

The code is internal rather than public, and can be demonstrated on request. What it shows is the habit behind the other tools here: when a task repeats, it becomes a script.

All projects

Cleared · My own internal tooling. No client page lists, image sets or site exports are published; the tools are described and can be demonstrated on request.

Contact

Want to talk through a project like this?

I am open to remote SEO, web design and performance roles. If you need more detail on a specific project, ask — I will share what the client allows.