University of Greenwich
Clearing: titles, URLs, IA, canonicals, redirects, internal links and breadcrumbs across gre.ac.uk.
- organic page views, clearing goal
- +0%
- form submissions
- +0%
- ranking clearing keywords, from 246
- 0
Work · index
Things I’ve built, measured and shipped.
Not just an SEO. The strategy is the easy part to talk about, so here is the stuff that exists: the products, the data pieces, the dashboards, and the numbers behind the client work.
01 · Built
Products and pieces that exist.
Some for clients, some for the agency, some because the tool did not exist yet. All of them real, most of them still running.
Free children’s books online
Lil’ Shelf
Stories, rhymes and poems from the public domain, read aloud by age and browsable by kind. Free to read, free to keep. A public site plus a reading app that lives on the home screen, on my phone and on friends’ phones.
Next.jsPWASupabaselilshelf.com- In use with clients since July 2026
AI answer tracking for PR agencies
Totally Answers
Which questions AI answer with a client in them, which sources those answers lean on, and what to do about it. Built as a product behind our managed service: question baskets, daily checks, source classification, insight cards, and a report to hand to a client.
Next.jsTypeScriptSQLiteChrome extensionPlaywrightDocker Public data, charted and kept current
totally.digital/data
Data journalism section for the agency, built by me. A&E waits by hospital, ambulance response times by service, pay by council area set against house prices, business openings and closures, and private rents by area, each from the primary NHS England or ONS release with the method on the page. The ambulance piece picks up where Totally's FOI postcode map, covered by the Guardian Datablog, BHF.org.uk, Institute for Government, the Planning Inspectorate, confused.com and several councils left off.
NHS EnglandONSSVGVanilla JSWordPress RESTtotally.digital/data- Running weekly
The charts update themselves
Release watcher
A scheduled job that reads the NHS England and ONS release pages, one request at a time, spots a new edition, rebuilds the affected piece in a staging copy, checks it, re-renders the images and only then promotes it. A second script updates the live pages over the WordPress API, hub last, and clears the cache. It runs to the release calendar rather than a guess: NHS on the second Thursday, ONS on its own Wednesdays.
PythonScheduled runsWordPress REST Chrome extension
Labrador
Captures real chatgpt.com answers and the sources they cite to CSV, then pushes them into Totally Answers. The thing the API will never show you: what a signed-in human actually sees.
Chrome MV3TypeScript- Running monthly
Eight clients, one builder, drafts on the first
Monthly AI answer reports
The PR-led AI answer report, built once for every client instead of by hand each time: where the client is included, the pages the answers draw on, the targets and the fixes. A decisions file per client carries every judgement forward so a month never quietly reverses one, and an independent check reads each draft before it lands in the client's Drive folder for review.
PythonTotally Answers APIGoogle Drive - In progress
A live experiment
Does truth matter to AI citation?
A statistics site built in three arms on one template: real figures from primary UK sources with the source beside every number, derived estimates with the method shown, and seeded invented figures in the gaps no source fills. Pre-registered hypotheses, a private register of every figure, bot logs and AI answers tracked to see which arm gets fetched and which gets cited. Where it lives stays unnamed until the results are in.
PythonStatic HTMLServer logsTotally Answers - In progress
A second live experiment
Would an AI quote a parody bank?
An anonymous reviewer rates the big investment banks and twenty-two private banks on terms, fees and service. Every figure about a real bank is sourced and dated, with the cheapest and dearest called out above each table. Beside them sits a panel of six invented banks with their own fingerprinted terms, visibly apart, so the answers engines can be watched for the day one of them turns up as fact. Satire in the desk notes, none in the numbers. Unnamed until the results are in.
PythonStatic HTMLCloudflare PagesTotally Answers A web archive, rebuilt
Now Then BMX
The old lost websites of five Northern UK BMX brands, 2000 to 2012, pulled out of the archives and rebuilt so you can click around them again. Six hundred pages and nine hundred images, a capture timeline, an image wall, every lost image marked as lost, and a sourced history of the people and the scene.
PythonWayback MachineStatic HTMLCloudflarenowthenbmx.comOne template, one config per client
Client dashboards
A monorepo that renders a password-protected dashboard for every client. All client data on one page. Improve the template once and every client gets it.
PythonVercelGSC APIGA4dashboards.totally.digital- In daily use
A kitchen that remembers
The Beans
Recipes imported from a photo, a screenshot or a link, a fuzzy cupboard that knows roughly what is in, a weekly plan, and a shopping list built from the plan minus what you already have. Lives on the home screen as an app, hooked up to Alexa so I can speak to my app; is in daily use in my house and a few friends’ houses.
Next.jsSupabaseClaude APIAlexa skillPWA Gin ratings, serves and bars
Ginapper
Rate the pour, not just the bottle. 155+ gins, classic cocktails and UK gin bars on the public site, and an app for logging serves and finding the pairing that sings.
Next.jsSupabaseVercelginapper.comData journalism for a PR campaign
ONS business demography, visualised
Six charts and a council-area lookup of business openings and closures across the UK, 2019 to 2024, built straight from the ONS tables for 365 Finance. Self-contained HTML, no libraries, light and dark, embeddable by any outlet that covers the story.
ONS dataSVGVanilla JSPythonSee it liveCompanies House surveillance
BTR Capital Radar
Watches UK build-to-rent capital markets for charge registrations, repayments, new SPVs and accounts filings, writes each signal up in a debt analyst’s voice and emails the alert. Cron jobs, no dashboard needed.
Next.jsCompanies House APIClaude API- Pilot
Edtech experiment
Deckhand
Turns a primary teacher’s English planning into fully scripted lesson decks, with a tick-or-fix ‘purple pen’ self-marking panel and an exit-ticket loop into the next lesson. Piloted with a classroom teacher in summer 2026.
Next.jsSupabaseClaude API
And the small stuff
- AIVis · DIY AI brand-visibility tracker: prompts in, mentions, citations and share of voice out as CSV.
- PMax Asset Studio · Programatically, turns a brand’s real images into Performance Max ratio with copy, logo and colours.
- Raw Search Console pulls · Datawarehousing script that pulls query by page by country by day straight from the API in a minute, before any tool gets a say. Historical querying of SC data that would otherwise be lost.
- Domain and uptime watch · RDAP drop watcher plus homepage monitors across every client site.
- Crawler access checks · Finds the bot blocks a site owner cannot see, using cloud probes and AI fetchers.
- Monthly leadership decks · GA4, Search Console and HubSpot into a PowerPoint a marketing director will read.
02 · Receipts
Numbers, with the sources attached.
Figures as published in Totally Digital case studies, which I led. Each card links to the write-up so you can check the method.
M&C Saatchi Performance
Domain migration, new site sections, landing page templates designed in house, global reporting dashboards.
- organic traffic
- +0%
- non-branded traffic
- +0%
- top 10 rankings
- +0%
National Garden Scheme
Programmatic SEO on a bespoke WordPress build fed by the garden database, with internal linking at scale.
- search impressions, from 5.3M
- 0.0M
- top 10 rankings
- +0%
- bookings
- +0.0%
Percentages compare the periods stated in each case study. Growth like this depends on the starting point, the sector and the client’s own team, so treat them as what happened there, not a promise of what happens next.
03 · The SEO bit
What I actually do.
Since 2012, from agency floors to the founder’s chair, and still in the crawl files.
01
Technical SEO
The bit most people mean when they say SEO. Still hands-on.
- Site migrations and re-platforms without losing the rankings
- Information architecture, internal linking at scale, programmatic pages
- Crawl, log file and index coverage diagnostics
- Structured data, canonicals, hreflang, Core Web Vitals
- JavaScript rendering and crawler access, including AI fetchers
02
Measurement
If it is not measured honestly it did not happen.
- Raw Search Console API pulls: query by page by country by device by day
- GA4 and GTM, server-side tagging, consent-aware setups
- Looker Studio and bespoke dashboards built from the APIs
- Attribution sanity checks: what the CRM says versus what the data says
03
AI search
Where search is going, measured rather than guessed.
- Tracking whether brands are cited or recommended in ChatGPT, Gemini and Perplexity
- Classifying the sources AI answers lean on and turning them into PR targets
- Capturing real consumer answers, not just API output
- Reports an agency can present to its client
04
Content systems
Briefs that tell a writer exactly what to write.
- Keyword to page mapping and content calendars
- Full-brief formats with intent, structure, entities and internal links
- Competitor and SERP ownership research
- Editorial processes that survive the client’s own team
05
Digital PR and links
Coverage that moves rankings and AI answers, not just a clippings list.
- Data pieces outlets can embed, like the ONS work
- Placement lists built from where AI answers already look
- Coverage audits: who linked, where, follow or not, what it did
- Working alongside the client’s PR agency rather than around it
06
Building the tools
When the tool does not exist, make it.
- Next.js and TypeScript products, Python pipelines, SQLite and Supabase
- Phone apps as installable PWAs, shared with friends, with no app store in the way
- Chrome extensions, cron jobs, launchd automations
- Claude API where it earns its place, rules where it does not
- Playwright test suites so clients never see the bug
04 · Toolbox
The stack, more or less.
Next.jsTypeScriptPythonSQLiteSupabaseVercelDockerPlaywrightWordPressShopifyTailwindNodeGitHubGitLablaunchdCloudflareNext.jsTypeScriptPythonSQLiteSupabaseVercelDockerPlaywrightWordPressShopifyTailwindNodeGitHubGitLablaunchdCloudflare
A few of the names
Clients whose work Totally Digital has published case studies on.
- University of Greenwich
- M&C Saatchi Performance
- National Garden Scheme
- CCGroup
- Case Luggage
- 365 Business Finance
Got a messy middle?
Technical problems, tracking that lies, content nobody owns, PR that does not move anything. That is the bit I am usually dropped into.