How I Rebuilt DigiSavvy's Content with SEO Research & Positioning

Here’s the lazy way to relaunch a website: pour your old content into the new design, swap a few words, ship it. I’ve watched plenty of agencies do exactly that. For my own site? I couldn’t stomach it. I deleted almost all of it and started from a blank page.

Not because the old stuff was terrible — some of it ranked fine. It’s that twelve years of blogging on a WordPress site gives you content built on vibes: whatever I felt like writing that week, whatever a client asked about, whatever was trendy in 2016. There was no plan behind it, just sediment. So before I wrote a single new word, I did a bunch of unglamorous homework. Here’s what that actually looked like.

First, I went snooping on the competition

I pulled real data — not gut feelings — using DataForSEO and a fairly extensive SEO toolkit I’ve put together. The goal was simple: what are the agencies I respect (and the ones I don’t) actually ranking for, and where are the holes?

Competitive gap analysis sounds fancy. In practice it’s: pull every competitor’s ranking keywords, line them up against mine, and look for the topics where a bunch of them rank and I don’t show up at all. Those gaps are the opportunities. Some of them were obvious in hindsight. Some I’d never have guessed. That’s the whole point of looking at data instead of trusting your own brain — your brain is a liar with an ego.

Then I turned the same tools on myself

This part stung a little. I ran the same analysis on my own content to figure out what was actually pulling its weight. Which articles ranked and brought in traffic? Which ones were just… there, taking up space and diluting whatever authority the site had?

So I made some hard calls. Keep the top performers. Demote or cut the dead weight. And for everything I was removing, I set up redirects so I didn’t nuke the link equity or leave people staring at a 404. If you’ve ever wondered why a site relaunch tanks someone’s traffic, it’s almost always this step — skipped. I wasn’t about to do that to myself.

The 45-minute therapy session (a.k.a. positioning)

Here’s the one I almost skipped, and I’m glad I didn’t. Before writing the new pages, I sat down with a positioning exercise — a tool I built to force myself through the uncomfortable questions. How do I actually talk about what I do? What do I want to be known for? What do I want to promote, and what am I quietly tired of doing?

It took about 45 minutes and it was weirdly hard. Turns out it’s a lot easier to position a client than to position yourself — when it’s your own thing, every option feels true and you can’t see the forest. But coming out the other side, I knew exactly how I wanted to sound and what the site needed to say. Every page after that got easier, because I’d already done the thinking.

If you want the short version of where I landed, that’s basically the approach page in a nutshell.

Then, finally, the actual pages

With the research and the positioning done, the writing was the easy part. I did a real round of keyword research and built out a set of keyword-focused pages — the stuff people actually search for when they need what I do, written to match that intent instead of whatever I felt like rambling about.

Next on my list: industry and audience-focused pages. The keyword pages catch people who already know what they’re looking for; the audience pages are for the folks who just know they’ve got a problem and aren’t sure what to call it yet. Different doorways, same house.

(The build side of all this — Etch, ACSS, the AI workflow — I wrote about separately in Building With Etch and how I designed the thing with Claude. This post is about the words, not the wrapper.)

The part nobody’s figured out yet: getting cited by the AIs

Okay. Here’s where I stop pretending I’ve got it all figured out. There’s a whole emerging discipline around optimizing your content not for Google’s ten blue links, but for the AI answers — ChatGPT, Perplexity, Google’s AI Overviews, Gemini, Claude. The engines that read your page, decide whether you’re trustworthy, and either quote you in their answer or pretend you don’t exist.

And nobody can agree on what to call it. AEO? GEO? LLMO? Generative engine optimization? Answer engine optimization? I’ve counted five acronyms this month and I’m fairly sure two of them were invented by someone trying to sell a course. Whatever the hell it’s called, it’s real, the clicks are already showing up in my analytics, and I built the new site with it in mind. Here’s what I’ve actually learned so far.

Answer the damn question first

The single biggest shift: stop burying the lede. The old blogging instinct is to warm up — set the scene, build to the point, reward the patient reader. AI engines don’t read like that. They scan the first chunk of a page to decide whether you answer the question, and if you make them wait, you lose. So now I front-load. The first line under a heading answers the thing directly, then I get into the why. (Notice I’m doing it in this very post. Meta, I know.)

Write headings as the actual questions people ask

“Pricing overview” is a heading for me. “What does a WordPress care plan cost?” is a heading for a machine matching it to somebody’s question. Conversational, question-shaped headings line up with how people actually talk to these tools. It feels a little dumb to write them that literally. It also works.

FAQs are doing way more than I expected

A good FAQ section turns out to be disproportionately powerful here, because half of what people type into an AI is a question. I pull the real ones from Search Console and from what clients actually ask me, answer each in the first sentence, add a line of context, and mark it up with FAQ schema so the machines know exactly what they’re looking at. I built a little plugin to manage that schema across the site — that’s probably its own post.

Make your sentences liftable

This is the idea that reframed everything for me: if you can pull a single sentence out of your page, drop it into an answer with zero surrounding context, and it still makes complete sense — that sentence is citation-ready. Specific numbers, clean definitions, claims that stand on their own. Vague mush that only means something in context is useless to a model. So I write more declaratively now, with actual specifics, and I try to make every key passage able to survive on its own.

Be a real person with a name

The trust signals matter more here, not less. Named authors with real credentials beat “the editorial team” every time — the model is deciding whether to stake its answer on you. So the new site has real bylines, a real bio, a real face (mine, for better or worse), and links that prove I’m a human who’s done this a while. Twelve years of doing the work is the asset. I just had to make it legible to a machine.

Don’t accidentally lock the doors

The dumbest, most common own-goal: blocking the AI crawlers in your robots.txt without realizing it, then wondering why you never get cited. I checked mine. I’m also messing with an llms.txt file — basically a welcome mat that points the AI bots at my best stuff. Whether it does anything yet is genuinely unclear, but it’s cheap insurance.

It’s a layer, not a replacement

Here’s what the breathless “SEO is dead!” crowd gets wrong: none of this replaces SEO. A page that ranks well in Google tends to get cited well by AI, because they’re judging a lot of the same things — relevance, clarity, authority. GEO is a layer on top of solid SEO, not a substitute for it. If your fundamentals are garbage, no amount of FAQ schema saves you.

And measuring it is still kind of a mess

How do I know any of this is working? Imperfectly. I watch analytics for referral traffic from chatgpt.com, perplexity.ai, and the rest — that’s the cleanest signal an AI sent someone my way. And the low-tech method everyone sleeps on: I just ask the tools myself. Once a month I throw my target questions at ChatGPT and Perplexity and see if I show up. If I don’t, that’s the to-do list.

Is some of this overkill? Maybe. Ask me again in six months — half of it might turn out to be superstition and the other half table stakes. But most of what I’ve described isn’t some weird new dark art anyway. It’s good content hygiene wearing a new hat: answer the question, be clear, be trustworthy, don’t hide. Turns out the robots want roughly the same thing the humans always did.

The takeaway, if you want one

None of this is glamorous. It’s research, spreadsheets, hard cuts, an uncomfortable hour of self-reflection, and redirects — before you write anything good. But that’s the difference between content that exists and content that works. The writing is the fun part. The homework is what makes the writing count.

If your own site is twelve years of sediment and you’re staring down a relaunch — that’s a conversation I enjoy having.