My full-time job: how I work
Since May 2026 I've been a Senior SEO Manager at Optimal, a OneMagnify company, working on SEO and AI visibility for enterprise brands across many industries. This page is how I work: how I think, how I build, and how I prove it worked.
I build systems, not one-off deliverables. Every engagement gets the same three things: a clear way to think about the problem, a repeatable way to execute it, and a memory that makes the next round faster and cleaner than the last.
AI speeds up my thinking and covers more ground than I could alone. It helps me find opportunities, pressure-test ideas and take the repetitive work off my plate, so my time goes to what matters most. I decide what should exist and why, and I set the standards. Every number gets verified, every deliverable gets red-teamed, and every mistake becomes a rule.
How I keep the thinking separate from the building
Quality slips when one tool is asked to think, build and remember all at once. So I keep those jobs apart, and the decisions always sit in the first layer, with me.
-
1. Thinking
Strategy, prioritization and stress-testing. What pages should exist, what intent they serve, what the client actually needs to decide. AI helps me pressure-test my ideas and find the gaps I'd miss. The decisions stay mine.
-
2. Execution
Build scripts, data pipelines, audits, decks, workbooks, schema and content drafts, built to an architecture I've already decided. AI speeds this part up so more of my time goes to the thinking. The rule: when execution needs a new strategic decision, stop and go back to the thinking layer.
-
3. Memory
Every engagement has written project instructions, a decision log, standards and a list of known mistakes. Each working session starts from that context, so the work picks up where it left off instead of starting over.
People rotate, accounts change hands, and decisions get made in meetings that never get written down. A memory layer means the program survives all of that. The checks I run on my own sites work the same way: how I build.
What I do first on every account
Every search starts with a person trying to close a gap between what they know and what they need. The surfaces keep changing; the psychology doesn't. So intent comes first, and the other three serve it. They map onto my get found, get understood, get chosen system.
-
Start from the searcher's intent
Before any page, report or fix, I name who is searching, what they typed and what they need to do next. Every other choice on the account answers to that.
Get understood and get chosen. This is the spine; the other three exist to serve it.
-
Make the site easy for search and AI to find and read
Clean crawling and indexing, a structure that matches how people search, and fixes made at the template so one change reaches every page built on it. The right answer on the site does nothing for a searcher who never sees it.
Get found.
-
Measure what searchers did after the change
A baseline before the work, the same report structure every month, and AI visibility broken out by platform and by kind of buyer. Shipped is a date. Worked is a number someone can check.
The measurement layer, under all three outcomes.
-
Work with the teams who ship it
Digital PR, paid search, analytics and the client's own developers. A plan SEO writes alone ships slowly or not at all, so they're part of it before it's set, and the pages PR pitches are ready for the attention they earn.
Get chosen.
What the work looks like week to week
Strategy decks don't move rankings. These do, and they're the same on every account, which is why they're written down.
-
Start with the template
On a big site, one template sits behind thousands of pages. One careful change to its titles, headings, internal links or structured data lifts every one of them at once, so that's where I start. Technical SEO on complex websites
-
Give the pages that matter something no one can copy
Then I pick the pages that carry the most weight and add what a competitor can't scrape: first-party detail, real policies, real answers from the people who run the business. That's the moat, and in my experience it's what AI answers end up citing.
-
Product pages that answer the buying questions
Product page templates get the details shoppers check before they buy, in the order they check them, with product structured data underneath. One template change, every product.
-
FAQs from real questions, never invented ones
FAQ sections come from what customers actually ask: search queries, People Also Ask, support conversations. If nobody asks it, it doesn't go on the page, and it never goes into schema just to fill a slot.
-
Monthly reports that build themselves
The data pull, the comparisons and the first draft of the story are automated, with every number traced to its source. My time goes to what the numbers mean and what we do next.
How the SEO changes with the decision people are making
Industries matter less than the decision the searcher is making. The same brand can face several of these on one site, and each needs a different page. The page types behind each are in the four types of search intent.
-
A big purchase people research for weeks
They compare, read reviews and look for the catch. The pages that win are honest comparisons, real reviews and a straight answer to the question they'd be embarrassed to ask a salesperson. What is commercial search intent?
-
A quick buy from a big catalog
They know roughly what they want and need to find it fast. Templates, filters and product structured data do most of the work; the copy's job is to settle the last doubt.
-
A service people need now, nearby
Location, hours and availability decide it. Location pages, business profiles and the same details everywhere on the web matter more than any blog post. Why local SEO is underrated
-
A professional or regulated service
Trust decides it. Plain-language explanations, named credentials and a review step for every claim, because a wrong fact here costs more than a lost click.
-
A brand fans already love
Fans arrive with specific questions: release dates, editions, what's next. Answer those first and fast, and stay honest about what's confirmed. My fan sites run on exactly this.
Getting SEO work live in a big organization
On an enterprise account, the hard part is rarely knowing what to fix. It's getting the fix approved, built and live through brand review, legal review, development queues and a release calendar. So I plan the path to live as carefully as the SEO.
-
Make the ask small enough to approve
A template change with a clear before and after gets a yes faster than a sitewide overhaul. Big recommendations get broken into changes a reviewer can approve in one meeting. Repeatable SEO systems for site launches
-
Hand over work that's ready to build
Every recommendation says what to change, where, and how we'll know it worked, so it can go straight into a development ticket instead of waiting for someone to translate it.
-
Plan around review cycles, not against them
Work that needs brand or legal review starts first. Quick technical fixes fill the time while it's in review, so the calendar never sits empty.
-
Keep stalled work visible
Anything waiting on an approval shows up in the report with what it's waiting on, so a delay is a known cost instead of a quiet one.
How I work with clients
Clients bring me in for the expertise. They keep working with someone who makes their job easier. So every deliverable is written for the person who has to act on it, and to hold up when they forward it to their boss.
-
Simple beats impressive
Complex SEO and AI search ideas get translated into plain language and one clear recommendation, so the client can explain it upward without me in the room.
-
Agendas built from what they asked
Before each meeting I go back through our notes and open questions, with AI helping me make sure nothing slipped, so the agenda starts with their priorities, not mine.
-
Reports that answer "so what"
Every report leads with what changed, why it matters to the business and what happens next. The data is there for anyone who wants it; the story comes first.
-
Give more than was asked
The ready-to-build ticket, the one-slide summary, the answer before the follow-up question. The goal on every account is to make the client's week easier.
The prompts I reuse, and why they work
These are patterns I reuse, written generically. The real power isn't the wording. It's the context files each prompt points to. Better files, not better prompts. Each pattern sets the standard and the checks; I make the calls on what comes back.
-
Monthly SEO and AI visibility reporting
The problem: Monthly reports drift. Numbers get pulled from different windows, sources get blended, and nobody can trace a figure back to where it came from.
Read the prompt pattern
Build this month's SEO and AI visibility report from the project template and the decision log. Before any number goes on a slide, log it in the ledger with its source, date window, scope and raw count. Never mix date windows on one slide. Never average two sources that measure the same idea differently; name both. Frame early data as preliminary with no causal claims. Each slide follows three beats: what happened, why, what's next. Render every slide and visually check it before delivery, then run the number check against the ledger.
A story chain I use for every report: shipped, found, cited, chosen. It connects the work to how search systems responded, without overclaiming.
-
Technical SEO and GEO audit
The problem: Audits that flag everything as broken lose credibility fast, because some of what looks wrong was a deliberate choice.
Read the prompt pattern
Act as a senior technical SEO with a point of view, not a hedge. For every finding, separate what the site did intentionally from what is genuinely broken, and say why. Verify claims with direct evidence (crawl data, live page checks, AI citation data) rather than trusting summaries or a tool's own "verified" label. Rank findings by impact on getting found, understood and chosen. Flag anything you couldn't verify instead of citing it.
Opinionated, evidence-backed recommendations a team can actually act on.
-
AI visibility prompt strategy
The problem: Brands want to know how they show up in ChatGPT, Gemini and AI Overviews, but tracking the wrong questions produces confident-looking noise.
Read the prompt pattern
Review this AI visibility prompt set against our content pillars and topic clusters. Keep visibility scoring to non-branded questions real buyers ask; tag branded prompts for sentiment only. Flag prompts that are leading, duplicated or not tied to a pillar. Recommend a frozen core set for month-over-month comparison and a fixed baseline date. Keep internal critique separate from the clean, forward-looking version the client sees.
Measurement design before measurement. When the question set moves, the trend line means nothing.
-
Content built for search and answer engines
The problem: Content calendars produce volume, not visibility.
Read the prompt pattern
Two stages, outline before draft. Pull keyword volume from the research tool; when it isn't available, stop and tell me rather than estimating. Structure around how the reader actually decides, answer first, context second. Use the brand style file for terminology and mechanics. No symmetric bullets, no AI-sounding constructions, no self-serving listicles. Keep internal strategy notes out of customer-facing copy. Put time-sensitive figures in clearly dated snapshot sections.
Content as infrastructure. Each page has a job, a place in the architecture and a reason to be cited.
More of my real prompts, dated, are in the prompt library.
How I track results
Reporting is where trust is won or lost. These are the rules I hold every report to.
Every number has a passport. Source, window, scope and raw count, logged before it's used.
Say what the number covers. "18 of 30 pages with data," not just "60%."
Never blend what shouldn't be blended. Different engines, different windows, different sources stay separate.
Retrieved is not cited, and cited is not chosen. The words mean different things and stay that way.
Search data decides the KPI. AI visibility is diagnostic. It tells you where to look, not whether you won.
Baselines are fixed. Pick a start date, freeze the question set, compare against it.
Credit honestly. Check who built a page before claiming the result.
Weak data stays out of the big room. Unverified numbers get a one-on-one conversation, not a slide.
The same rules run on my own sites, in public: the measurement layer.
Rules I work by
Better files, not better prompts.
Context beats clever wording.
Fix the source, never the output.
One upstream change improves a thousand pages.
Every mistake becomes a rule.
Written down, carried forward, never repeated.
Verify, don't trust.
Not team summaries, not tool labels, not my own first pass.
Red-team before it goes external.
Stale links, stray comments, unverified claims.
Finish small before going wide.
Depth earns trust from people and search systems.
Know which document is the source of truth.
Everything else is context.
Audience first.
What belongs in this conversation, with this person, in this format?
Where you can see the work
The hard numbers on this site come from my own projects, where I can show the decisions, results and misses in full. The methods above also guide my full-time work.