Discipline 04 — AI & Automation

The repetitive work becomes a tool.

Python tools calling SERP, DataForSEO and OpenAI APIs: keyword clustering, entity research, image metadata, structured data. Each one removes a manual step from SEO delivery — and each one stops before the judgement call.

Python · Streamlit · OpenAI API · Node.js

Primary stack
Python · Streamlit · OpenAI API
Data sources
SERP APIs · DataForSEO · Google Autocomplete
Built for
Clustering, entities, metadata, schema
Principle
Automate volume, keep the judgement
01 — Tooling

What I have built.

Three public repositories and three internal tools. Each entry says what it replaces, because a tool that saves nothing is a hobby.

  • Open source · Python + Streamlit

    GitHub (opens in a new tab)

    SEO Clustering Tool

    Keyword clustering through SERP URL overlap, with cost-aware caching that reduces DataForSEO and SERP API spend on large keyword sets.

  • Open source · Python

    GitHub (opens in a new tab)

    Keyword Autocomplete Tool

    Long-tail keyword discovery at scale using Google Autocomplete and the DataForSEO API, built to shorten research cycles.

  • Open source · Python

    GitHub (opens in a new tab)

    Entity Intelligence Suite

    Entity gathering, entity SEO research and competitor intelligence to support semantic SEO and topical-authority analysis.

  • Internal tool · Python + Tkinter

    Internal · demo on request

    Internal Links Checker

    Audits supplied pages, classifies internal and external anchors, checks HTTP status and exports a four-column CSV for broken-link and orphan-page cleanup.

  • Internal tool · Python + ExifTool

    Internal · demo on request

    Image Metadata Processor

    Produces lossless WebP copies and writes SEO filenames, alt text, XMP/EXIF/IPTC and GPS metadata at scale, with read-back verification.

  • Internal tool · Node.js

    Internal · demo on request

    JSON-LD Schema Generator

    Generates paste-ready structured data for organisation, local business, service, article, how-to, FAQ, breadcrumb and person schema.

02 — Where the models go

Judgement calls, not the structure.

AI is used where the volume is high and the answer is fuzzy. It is not used to decide what the site should say.

  1. Research that a script cannot do alone

    SERP-overlap clustering and entity extraction run on data APIs. The model handles the calls that need judgement — intent, ambiguity, near-duplicates — across keyword sets no spreadsheet could hold.

  2. Drafting against the map, never inventing it

    Outlines, meta copy and internal-link suggestions are generated from the topical map, then read and corrected. The structure is decided before the model is involved.

  3. Chained into agentic workflows

    What I am building now: chains that carry one brief from keyword set to cluster to map slot to outline to link suggestions, with a checkpoint after each stage instead of one blind prompt.

03 — Rules

What does not get automated.

  • The map decides the structure. A model can shorten the research behind a row, not choose the row.
  • Nothing publishes unread. Generated copy is corrected against the source and the map before it goes live.
  • Client data stays out. Keyword sets, site exports and analytics are only processed where the engagement allows it.
  • Every script is documented. If I cannot hand it to someone else with its inputs and expected output, it is not finished.

All six tools run on my own input and print what they changed. None of them writes to a live site without a review step, and the client-facing versions are documented before handover.

Contact

Hiring for someone who automates the repetitive work?

I am open to remote SEO and technical SEO roles where tooling is part of the job. Ask for a walkthrough of any tool on this page and I will show it running.