The full archive.

Everything below is work I did the thinking for, not just the words. It's grouped by skill rather than industry, because the method carries across subjects. Client brands are withheld under NDA and described by site type instead.

Software and CRM reviews E-commerce category pages Regulated YMYL content Crypto guides UK, US, EU, NZ and Romanian audiences

Rankings & recovery

Pages rebuilt until they ranked.

Diagnosis from search data first, then rewrites built around what the searcher actually wanted.

↗ 100+ to #57 in one week/ commercial listing page/ NZ market

Turning around a page ranked outside the top 100

Rankings

A commercially important listing page was ranking beyond position 100 and producing nothing. I rebuilt the content completely around search intent and applied targeted on-page SEO to the offer structure, terms presentation and internal linking. It climbed to position 57 within one week of going live.

♻ Core Update recovery/ regulated YMYL site/ ✓ still live 2026

Trust pages rebuilt after a Google Core Update

Recovery

After a traffic drop from the March 2024 Core Update, I led the rebuild of the three pages carrying a UK site's trust architecture, applying E-E-A-T and Helpful Content principles:

  • Homepage: restructured around the review methodology and reader safety
  • Publishing policy: editorial credentials, authorship, corrections and independence
  • Review process: a documented 8-factor scoring system

Rankings and organic traffic recovered across both release waves. Re-verified July 2026: the structure is still live, with several sections running my original copy word for word.

↗ position 34 to 16/ pricing page rebuild/ OfficeAtlas

Rebuilding a pricing page around what buyers ask

Refreshes

I rebuilt the Pipedrive pricing page with usage limits, hidden costs, plan-change rules and a team-size cost table, all verified in a live trial account. “pipedrive pricing” moved from position 34.1 to 15.7, and “pipedrive subscription plans” from 31 to 6 in 28 days. The page is now cited in Google's AI Overviews for high-volume searches like “pipedrive pricing” and “pipedrive plans” (checked manually, October 2026).

⚖ cannibalization fix/ evaluation cluster/ Search Console

Diagnosing and fixing keyword cannibalization

On-page SEO

Two pages were eclipsing each other on shared evaluation intent, with a template outranking the guide for the guide's own target term. I confirmed the overlap in Search Console, retargeted one page to a separate vocabulary and split the contested cluster into two clean intents.

⇄ internal linking/ three topic clusters/ multi-market site

Cluster interlinking and search intent alignment

Architecture

Topic cluster analysis across a multi-market review site, rebuilding internal link maps to improve crawlability and topical authority, and rewriting headings to match what users searched for rather than internal category names. Titles and meta descriptions were refreshed to lift click-through rate.

monthly program/ highest-value pages/ UK market

Monthly refresh cycle across commercial pages

Refreshes

I ran a monthly update cycle across a site's most commercially important listings and guides, deciding each month which pages justified the work, so the content adapted to search updates as they landed rather than after a drop.

Research & testing

Evidence before opinions.

Reviews and guides built on first-hand testing and verified data, in whatever subject the work needs.

⚙ four CRMs tested/ scripts, datasets, evidence logs/ OfficeAtlas

Hands-on software reviews on OfficeAtlas

Product testing

OfficeAtlas is my own CRM review site, built from zero. Pipedrive, HubSpot, Zoho and folk were each tested in a real trial account with a written test script, a test dataset and an evidence log, then scored against a published 14-category framework. The Zoho review alone rests on 187 screenshots.

In three months, site impressions more than doubled and average positions rose sharply: Pipedrive vs HubSpot from 50 to 27, the Pipedrive review from 53 to 34, Best CRM for Small Agencies from 63 to 35. Impressions in Google's AI features grew from about 1,000 to over 4,500, and the CRM evaluation template is linked directly in several AI Overviews.

Testing surfaces what vendor pages leave out: Zoho's import read European number formats as the wrong amounts, and Pipedrive trial accounts get no enrichment credits, even on its top plan.

✎ guest article/ HubSpot Diamond Partner/ IntegrateIQ

A technical guide built on controlled tests

Technical guide

Published by IntegrateIQ in September 2026: what to check after a HubSpot CSV import, written from controlled test imports rather than documentation.

▦ interactive tool/ 14 criteria, five weights/ OfficeAtlas

An interactive CRM evaluation matrix

Helpful content

A weighted evaluation matrix readers can use to score CRMs themselves, with a guide, a worked example and a matching Google Sheets template. GA4 events confirm visitors use it.

⚲ cold start/ 100+ guides/ up to speed in a day

Writing on subjects new to me

Research method

I have written more than 100 guides on subjects ranging from familiar to entirely new, including crypto guides on payments, wallets and how to use crypto for beginners. The method: gather and summarize sources fast, verify the mechanics first-hand wherever possible, and study the strongest competing pages to find what they miss.

⏱ tight deadline/ unfamiliar market/ US freelance

Delivering in a market new to me

Adaptability

A freelance commission for a US client under a tight deadline: I researched state-level context and audience behavior from scratch and delivered a ranked listing page. The client reported high satisfaction with how well the work adapted to the market.

Audits & localization

Finding what's wrong, in more than one language.

On-page reviews, content audits and native-language checks, with problems flagged even when they sit outside the brief.

⇄ native Romanian review/ ~120 pages a week/ e-commerce

Romanian SEO review for e-commerce category pages

Audit & localization

For an SEO agency, I review Romanian e-commerce category pages in weekly batches of around 120 URLs: primary and secondary keywords, titles, meta descriptions and H1s, rewritten for natural Romanian and real search intent. Every keyword is checked in Google Keyword Planner against the products live on the page.

  • Replaced keywords that weren't Romanian, didn't match the page or had no search volume
  • Put primary keywords at the start of titles; wrote meta descriptions under 160 characters with a hook and a call to action
  • Beyond the brief: flagged URLs that didn't match their products, foreign-language slugs, keyword cannibalization and categories whose products didn't match their name
country versions/ UK, EU, NZ, global/ one brief, several executions

Content for different country versions of the same site

Multi-market

I wrote content for the different country versions of the same websites, adjusting spelling, regulatory framing and tone for each audience rather than reusing one text.

⚠ duplicate content/ hundreds of URLs/ WordPress

Writing around repeated page elements

Technical content

Sites built on templates repeat the same boxes, disclosures and notices on every page. I accounted for them in how each page was written, so unique content stayed unique and no page was diluted by boilerplate it couldn't remove.

Editorial systems, AI & automation

The operation behind the content.

Writing is the visible part. Underneath sit the briefs, the voice, the writers, the tools and the automation that keep quality steady at volume.

⚙ AI editorial system/ built over three months/ adopted by an eight-person team

An AI-assisted editorial system for regulated content

AI & editorial systems

Over three months I built a prompt system the whole product team came to use daily. By the end, roughly 90% of each output was usable as written, my own estimate of my own workflow rather than a measured figure.

  • Compliance enforced at generation, not corrected afterwards
  • House voice as parameters: active voice, second person, geo-targeting
  • Anti-fabrication: missing values flagged in bold instead of invented
  • Facts never from the model: every prompt supplied the data or pointed at the live page
  • Quality rubric: each heading scored 0 to 10 on intent, depth, originality, trust and UX, then a harsher second pass
Read the full write-up →
↗ 20-40% to ~90% usable/ around ten writers/ five sites

Running and developing a freelance writer network

Editorial management

As one of four editors across five sites, I briefed, reviewed and edited a rotating network of around ten freelancers and made retention decisions. At least three went from a 20-40% usable baseline to around 90% after a few months of structured feedback.

▢ brief architecture/ 10-20 briefs a day/ cross-department

Brief architecture and house voice

Editorial systems

I designed the brief formats the portfolio published through, tiered by content type and writer capability, and devised a friendlier, non-promotional house voice that became the site standard. The role ran across departments: briefing designers and escalating template faults to developers.

⏱ ~200 updates in five days/ three sites/ 3,500-5,000-word pages

Working at volume

Throughput

Rebrands and audits meant production at pace: one project ran to roughly 200 page updates in five days across three sites. My longest pieces, 3,500 to 5,000-word listing pages, were usually researched and written in a single day.

⚙ two AutoCAD plugins/ 20+ hours saved per project/ land registry work

Automating repetitive work wherever I find it

Automation

Outside content, I document land registry plans. I built two AutoCAD plugins with Claude: one automates topology testing and correction, a step that took more than 20 hours of manual work per project, and the other fixes a long-standing fault where coordinates drifted without anyone editing them. The same instinct runs through my content work: whatever I keep doing by hand gets turned into a system.

Next step

If any of that maps to what you need, get in touch.

Send me the role, the site or the problem. I'll tell you honestly whether I'm the right fit before quoting anything.