I Want To Show Up On AI Suggestions. Here’s How I Optimized My Website.

Successes, Redirections, AI SEO, Troubleshooting

My goal is to establish an internet presence as THE authority on “AI copywriting.”

And I wanted to measure it by how well I showed up in AI suggestions. Like ChatGPT and Google’s AI overview.

I could have spent over $30,000 getting the professional’s help. But right now, even the professionals are trying to figure out AI SEO.

I decided to figure it out myself.

Over the past three months, I implemented every suggestion I could find on building your authority on a topic, including: 

• Utilizing a web builder that lets you do more complex backend stuff (so I migrated from Squarespace to WordPress)
• Planned a content universe around “AI copywriting” (I used Ubersuggest and Google’s “People Also Ask” section to plan internally linking pillar pages and blogs)
• Created substantial pillar pages around core AI copywriting topics (a pillar page is a web page that covers a topic in depth – I wrote seven of them that were over five thousand words each)
• Stepped up my SEO game (Built custom GPTs to handle schema markup coding, metadata, and more)

My goal is to master AI authority building so I can

1. Offer it as a premium service
2. Show people what I learned
3. Create systems that reliably help people build AI authority

People’s search habits have changed a bunch over the past two years – so much so that it’s changing SEO. I want to keep up.

Even more than all that…I want to stay on top of AI so I can teach my two daughters the ropes. I don’t want to force them to figure it out alone.

My oldest is almost 3 years old, my youngest is 7 months old. I did this website authority-building overhaul over the last three months during their nap times and after putting them to bed.

There was a ton of troubleshooting and trail-and-error. It wasn’t easy, but it was super doable.

I’ll show you what I did, and you’ll save way more time.

Here’s How I Came Up With The Idea

The “become the AI Copywriting authority” idea was a shower thought. 

The next day, I researched authority-building elements and realized Squarespace (my web host) wasn’t customizable enough. I decided to switch to WordPress, since most of the sites ChatGPT viewed as authoritative were hosted there.

The next day, I mapped out the pillars and cluster blogs my site would need around the topic.

Another night putting the kids to bed, another night chewing away at the plan. The momentum built with each evening. 

I planned the wireframe, site map, keyword research, and pruned my old website. I migrated the website, re-wrote the web copy, and figured out how to navigate WordPress in general. 

I fed ChatGPT screenshots when I was stuck (which was a lot), and it walked me through all the weird tech stuff.

Then, I was ready to make these pillars and blogs. I was going to be a content-creation machine. I thought it was going to be just as fast, a month or so to make all the pillars and blogs.

Au contraire.

When I started creating my pillar pages, I knew I’d be leaning heavily on AI to speed up the process. And I knew there would be lots of editing and refining. 

What I didn’t expect was just how much course correction I’d need to do — not just of the content, but of my systems, tools, and expectations.

I started with a method I thought would work well. But about halfway through, I completely scrapped the pillars I’d written and edited and started over. It was really important to me that both the output and the process was consistent and high-quality.

Even with significant AI support, it took me 106 hours 18 minutes to outline, draft, and edit 8 pillar pages. These pillar pages, together, total over 45,000 words.

Here’s the walkthrough of my real process, including what didn’t work, what changed, and what ended up making all the difference.

Here’s My Original Content Framework (And Why It Changed)

I started out writing with SEO-style headings like “AI Copywriting Trends” or “AI Tools for Writers.” I used my custom blog-writing GPT by feeding it the brain-dump outline I’d created for each pillar. I tracked my time and wrote notes on what was working and what wasn’t.

It was the authority-building I knew. It was what worked before. I outlined, planned, and drafted these pages, and even began editing. 

But these pillars had much heavier research and length than my blogs (2K word blogs vs 5K+ word pillars). My notes and time tracking were proving that this was not a well-oiled machine. It wasn’t even the right machine.

Quality was decent but wasn’t getting better. They were not authority-building great.

At first, I knew I wanted to become an authority on AI copywriting — but the path wasn’t really clear. How would I measure authority? What did it practically mean?

This lack of clarity was showing up in the pillars. I kept reworking the structure, most AI “edits” were full rewrites, and the research process was a mess.

The longer I worked on these pillar pages, the more I realized that “authority” in this context meant showing up in:

• ChatGPT answers
• Google’s “People Also Ask”
• AI-generated overviews

There’s lots of other routes I could have gone, but I decided to focus on showing up in these three areas.

My audience also crystallized. I was always writing for copywriters (and those who write copy for their own business) who wanted to leverage AI, but do so ethically and actually see ROI. But I had to stop trying to make the content for a broader audience.

Even if many of them never buy my premium services, I knew this content (if successful) could:

• Position Check Copywriting as a trusted expert
• Drive organic traffic
• Serve as a case study for the quality and strategy I bring to my offers

The results needed to be great, but the process to make it happen needed to be great too. And repeatable.

So when pillars were taking around 20 hours to edit, I knew something earlier in the process had to change. I completely scrapped the pillars I’d created and started over.

With the new pillar pages, I abandoned some SEO standards. People’s search habits were changing, so I might as well evolve my SEO “rules.” 

I focused much more on user intent and conversational keywords.

I chose what seemed to be working better for AI crawlers:

1. Turning every H2 into an actual question – a high-intent query I pulled from Ubersuggest, Answer the Public, or Google’s “people also ask” sections.
2. Making sure the first sentence after each H2 directly answered that question, so it would be eligible for AI overviews and Featured Snippets.
3. Adding a dynamic table of contents and summary to the top of each page – the TL;DR to give both bots and human readers context, and the clickable Table of Contents made navigating easier. 
4. Only targeting one keyword per page. It was easier to implement and plan the pillar/blog cluster network. I treated the user search intent, FAQ answering, and my overall topic authority (which would be built with all the pillars and blogs) as more important than keyword targeting.
5. Bonus: Adding schema markup – a bit of code that helps bots understand what’s on your page to encourage them to crawl the whole thing. Don’t freak out, I can’t code either. ChatGPT does it for me.

Here’s what it looked like with Headings and their immediate answer:

• My old headings: No, AI Copywriting Isn’t Replacing Human Copywriters
• My new headings/answer: Is AI Replacing Copywriters? Yes, it’s replacing copywriters that don’t collaborate with it.

Here’s what my TL;DR section and Table of Contents look like:

Here’s what my keyword strategy looks like:

• Old strategy: Primary keyword (in title, subtitle, headings, and body), secondary keyword (in body), conversational keyword (in one heading).
• New strategy: Main keyword (used in title and throughout body). All headings were relevant FAQs, no adjusting to the keyword.

Here’s what the putting in the schema markup looks like. It’s just a block of code ChatGPT makes for me, and I copy/paste it into the website. I use the plugin RankMath to enjoy this nice user-friendly dashboard.

Does it all work? We’ll see. I do know that the old SEO keyword stuff has lost its mojo, so an experiment is certainly worth a shot.

This restructuring had major implications on how I planned – and wrote – the pillars.

Here are the main parts of the system I ended up building.

Here’s What Worked Well When Making the Pillar Pages

Over time, I created a repeatable set of internal rules that made editing much smoother. 

The process improvements cut off nearly 8 hours of editing time per page. Plus, they made the system way easier to repeat.

And I believe it’s the best copy I’ve ever made. 

I found that I like to:

• Use a custom Outlining GPT that did research on competitors and gap analysis. This replaced the manual outlines I was making, which was more like bulleted notes.
• Use DeepResearch (on ChatGPT) and a special prompt for pillar page drafting instead of my custom blog chatbot on ChatGPT or Claude.
• Use real case studies to keep things grounded, and only add fictional examples when needed – and never play them off as real.
• Pull in highly relevant quotes to add authority and break up the “AI voice.” People have such clever ways of distilling these concepts!
• Always re-write the hooks and conclusions (because the AI generated ones were always pretty weak — and when research was involved, they often triggered plagiarism tools).
• Remove generic AI-phrases like “game-changer,” “evolving landscape,” or “key takeaway”
• Use subtitles only to give context to the reader — not to chase keywords.
• Add a summary of the page in the form of a TL;DR (Too Long; Didn’t Read) section, as well as a Table of Contents.
• Create a formatting guide for myself so things, like citation, would finally stay consistent.
• Add a Table of Contents for context and easy navigation
• Use a Final Sweep Custom GPT to handle the meta data, title/subtitle, TL;DR, URL slug, extra places I can add my keyword, and specific image ideas.

I really strengthened my outlining process, which made for much better drafting downstream, then easier editing.

Finding the sweet spot for the research and drafting aspect took a lot of trial and error – so let me share what worked so you can avoid that.

Here’s How I Did The Research And Drafting Part

I think AI is really useful for drafting – not to write less, but because I believe the first draft is about building the structure you’ve already planned, not generating polished copy.

Writing good copy with AI is like building a house. You’re the contractor handling planning and management, and you’re the interior designer that’s manually pulling the final elements together. You’re not the boss telling an intern to build a house and decorate it nicely.

Plus, these drafts weren’t coming from the void. I was putting a lot of original content, planning, and research into the outlines, which guided ChatGPT’s drafting.

Then, after a great deal of testing, I figured out how to add the research part more smoothly.

These pillars needed relevant stats, real-world stories, expert insights, and claim validations. Citations would make the pillar content more robust and create appropriate outbound links.

After lots of different systems, using DeepResearch for both the pillar page drafting and research was the winner. 

To do it, I opened a ChatGPT chat, switched on DeepResearch mode, gave it a custom prompt with the research task and format I wanted, and attached my outline (which I made with my custom Outliner GPT).

The draft took anywhere from 10 minutes to 30 minutes to generate. Compared to regular ChatGPT, the research was much more relevant, there was more of it, and the flow of the draft fit a pillar.

During editing, I still manually verified every source. 

I learned to:

• Find original sources (many “sources” were secondary or tertiary). I had to go on my own Google rabbit trail for several studies and stats.
• Reword study findings (ChatGPT tended to quote them almost verbatim, which flagged the plagiarism checker)
• Use the consultant blogs with unique case studies and see through SEO-fluff blogs
• See where stats were duplicated (popular stats were often featured twice from different sources quoting the same study)
• Include the right hyperlink in my citation (sometimes it went to a company’s main page, not the relevant case study or post)

I chose to format my research citations as (Hyperlinked source, Year) at the end of the relevant sentence. There wasn’t an industry standard. It just felt clean and consistent yet professional.

A few things surprised me during the process:

• DeepResearch could pull from super obscure sources, like LinkedIn posts, Reddit comments, Instagram captions, and YouTube videos. For example, when I needed an original quote from an industry expert, it found a perfect, super relevant one – in a 4-month-old Instagram caption with just 4 likes.
• It could sometimes crawl paywalled sources. Not sure how the Wall Street Journal feels about that.
• There is so much AI slop out there. I can’t tell you how often I came across blogs and articles with virtually the same opening paragraph. It was everywhere, from small company blogs to Forbes articles. The bar’s not as high as I thought.
• The research quality was significantly better with DeepResearch. It found more relevant and obscure sources than when I asked a plain ChatGPT chat to do research. It also didn’t hallucinate URLs.
• There was lots of enterprise talk, but not enough grassroots wins. I struggled finding case studies of AI copywriting helping small businesses increase revenue. Even though AI’s hyped up as a free copywriter, there’s not lots of real examples of the ROI.
• ChatGPT uses Bing as its search engine. And unlike Google, Bing factors in social media when judging authority – which has HUGE implications for anyone trying to get featured by AI.

I still did use ChatGPT (like, a plain thread) to do research to fill in the cracks. I often asked it for a relevant quote or statistic, and provided the respective section. 

Thankfully, the DeepResearch research and drafting made editing way easier. But editing still took the bulk of the production time.

Here’s How I Did The Editing

I relied on AI for a lot — outlining support, research, formatting help — but editing was where I really cleaned it up and polished it.

When I edit, here’s what I do:

• Manually go through each line, making sure it actually adds value
• Write an original hook for the intro
• Write stronger section openers and conclusions
• Click each link, making sure it goes to the right source, fixing it if it doesn’t
• Format all the citations and apply the right hyperlink
• Add, delete, and reword sentences
• Break up sentences and paragraphs
• Add bolds, italics, and bullets
• Rename the heading to an FAQ I want to target, then add the answer as a sentence that immediately follows it

While editing, I open a fresh ChatGPT thread that I used on-demand to:

• Simplify clunky sentences I added (I’d literally prompt “simplify: …….”)
• Break long sections into bullets
• Suggest topic transitions
• Do quick research
• Get help making analogies, metaphors, and examples stronger
• Get feedback 

After I’m done with the edit, I run it through a plagiarism checker, then re-write the sentences that needed it (usually the ones including research).

I struggled to find a checker that could handle 5K+ words at a time. I tried Copyleaks, but it didn’t catch plagiarism well at that volume, so I went with Quetext.

I don’t think I’d ever outsource editing to AI. Manual review and editing is such an easy way to make sure the brand voice and tone are in line. But more importantly – it’s the quality control.

Is it more work than letting AI take over? Definitely. Editing can be tedious – but it’s way easier to work through a solid draft than to fix a messy AI one or start from scratch without a plan.

Here’s How I Handled The Tech And SEO Setup

Once the pages were edited and cleaned (I work on a Google Doc), I downloaded the pillar page as a Plain Text (.txt) format and gave it to my Final Sweep GPT.

It’s a custom ChatGPT chatbot I made that’s essentially a checklist for all the finishing touches. 

It gives me:

• A TL;DR (“Too Long; Didn’t Read” summary)
• Meta title & description
• URL slug (i.e. http://www.checkcopywriting/ai-copywriting)
• Title and subtitle suggestions
• Suggestions for images, and where to put them
• A list of all the headings (and first sentences)
• (If requested) Suggestions on extra sentences I could add my keyword

Once I get that info, I take that list of headings/the following sentences and feed it into my custom Schema Markup GPT. 

One of the top tips I found for getting content into AI snippets is using schema markup. It boosts SEO without changing what visitors actually see.

It’s a small bit of code that gives bots context about your page. Kind of like posting a menu outside your restaurant so people know if you’ve got the desert they want before stepping in. In this case, it helps AI see if your blog answers a user’s question.

I do not know how to code. Not even a little. But I don’t have to. My Schema Markup GPT just takes a list of headings and their answers and turns them into the correct JSON schema markup.

I can then copy that code and paste it into the backend of my website. That’s it.

And speaking of the tech side, I tried out RankMath. The free version was great, but I upgraded to access some of the other features.

Here’s what RankMath makes much easier to add:

• Meta title/meta description
• Pasting the schema markup
• URL slug
• Noting little things, like whether I had internal links, keyword density, keywords in alt text, etc.

It’s got a ton more stuff I plan to explore, but even just these have been super helpful.

Once I got those little details all arranged into their nice little homes on the backend of my website, I submitted my sitemaps to Google & Bing for indexing using their respective tools

Google: Search Console

Bing: Webmasters

Then, I built a full KPI tracking spreadsheet to measure how each pillar performs—tracking things like baseline SEO metrics, impressions, whether the FAQ appears in AI results, and more. I even have a prompt I use with DeepResearch to fill it out automatically.

Here’s What Didn’t Work (And Why I Dropped It)

Obviously, I want you to know what worked. But I’d also like to share what didn’t work. A lot of things that felt like an obvious solution just didn’t work. 

Here are a few things I tried… and eventually abandoned:

Experiment	Why It Didn’t Work
Created banks of research to reference while drafting or editing	Splitting the research phase from drafting created too much double-backing. It was better to research mid-edit – and even better to get DeepResearch to research mid-drafting.
Used custom longform blog GPTs to draft whitepaper-style pillars	Too generic, even when I used my custom blog-writing GPT on Claude and ChatGPT. Required heavy rewrites and struggled to break through 2K words.
Made a custom GPT to draft each pillar, putting the outline and research bank in the training	The GPT consistently only read the first page of the docs I attached to its instructions. It would use the same few studies from the research bank over and over.
Created a loose outline, tightening it as I drafted and edited	Weak outlining was like weak leadership. It made ChatGPT guess what I wanted, which I went behind and fixed. Really prioritizing the outlining process made drafting/editing so much smoother.
Trained AI on “my voice”	The effort it took did not proportionally show up in the end result. Regular editing naturally refined the tone and voice. Plus, these were pillars, not emails or SM posts. I felt that personality was relevant, but not a top driving factor for this kind of copy.
Attached files to my prompts as PDFs or DOCs	These pillar pages were 20-30+ pages each in a Google Doc. I realized ChatGPT was struggling to read past 20ish pages when I uploaded it as a .pdf or .docx. Once it was in Plain Text (.txt), it didn’t have trouble.
Gathered & assigned keywords in one go before writing	I used Ubersuggest’s free trial to find a slew of keywords relevant to my content universe. But the bank was hard to wade through while planning, even with AI help, especially when my SEO plan wasn’t nailed down. I ended up getting the Ubersuggest lifetime access for $99.
Used a plugin for code-free schema markup	It was too much copy/pasting each FAQ heading and answer into a template. It was actually easier to give ChatGPT the list of them and tell it to “make an FAQ schema markup in JSON-LD.”
Manually added the SEO details (meta data, etc) without help from a plugin	There’s a ton of scattered SEO details if you’re putting things in manually on WordPress.  I got the RankMath Plugin, which creates a tab where it’s all consolidated, and it makes suggestions. It’s about 100 bucks a year, 100% worth it.
Creating SEO details on the front end	I used to make the title, meta data, etc. before drafting. It was way easier to draft, feed it into a Final Sweep custom GPT, and have it spit out that data.

Here’s How I Saved Around $30K – $60K Doing Things Myself

According to ChatGPT’s research, handling all the tech, writing, and design solo saved me anywhere from ~$32,950 to ~$65,100.

Was I as efficient as the professionals? No.

But did a website get migrated, designed, and written, with an awesome SEO strategy? And I didn’t have to hire childcare? YES.

I really believe this is the kind of leverage AI can give the small business owner, especially the one that’s 100% bootstrapping. As a stay at home mom of two under three, how else could I have done things like DNS setup or coding schema? Even if I could have, how would I have found the time to move this quickly?

You’ve got access to this kind of leverage too.

Estimated Cost Breakdown of Work I Did:

Task	Professional Equivalent	Estimated Market Rate	Qty/Time	Est. Cost
Copywriting (pillar pages)	Senior SEO Copywriter	$0.50–$1 per word	4 pages x 5,000 words	$10,000–$20,000
Copywriting (blog updates)	Mid-Senior Copywriter	$0.30–$0.50 per word	15 blogs x 2,000 words	$9,000–$15,000
Website Design (pages)	Web Designer	$1,000–$2,500 per page	4 main pages	$4,000–$10,000
Image Design (text overlays, layout)	Graphic Designer	$100/hr or $200–$400/page	4 main + 15 blog pages	$2,000–$4,000
WordPress Site Build + Migration	WordPress Dev	$75–$150/hr	~20–30 hrs	$1,500–$4,500
Domain Hosting Migration & Google Workspace Transfer	Tech Admin or IT Pro	$100–$150/hr	~2–3 hrs	$200–$450
DNS Setup (including email + integrations)	Technical SEO or IT Admin	$100–$150/hr	~2–3 hrs	$200–$450
Schema Markup (manual coding)	Technical SEO Specialist	$100–$200/hr or $150 per page	~20 pages	$3,000–$4,000
Pillar Page Strategy/SEO Plan	SEO Strategist	$1,000–$3,000 per strategy	1 comprehensive strategy	$2,000–$5,000
Citation-based Research	Content Researcher	$50–$100/hr	~15–20 hrs	$750–$2,000
Google & Bing Analytics Setup + Sitemap Submission	SEO Tech or Consultant	$300–$700 flat	1 site	$300–$700

People are getting nervous about AI taking jobs. But for a lot of us small business owners, it’s giving us access to services we couldn’t have afforded anyway.

You’ve got access to tech-power that used to be just for enterprise-level businesses.

Here’s My Biggest Takeaway: Keep Moving & Take Notes

The best lesson I learned: Just make a decision and move forward.

And document the process.

Instead of over-researching every next step, I got the most leverage by trying it, noting what worked or didn’t, and adjusting with the help of ChatGPT.

For example, I needed to decide:

Did I want to run a plagiarism check? Yes.

What tool should I use? After research, I opted for CopyLeaks.

I used it. It said everything was plagiarism free. I was suspicious. 

Another decision – should I try another tool?  I decided “yes.”

Which one? I decided on Quetext.

It was just hundreds of little decisions like this. Ones that didn’t have a “best” answer. Decisions don’t take but a moment to make, but they have to be made.

My notes and time tracking helped me realize which steps made the biggest impact — like investing more time up front on refining the outline to dramatically cut down editing.

And I let go of the idea that every decision had to be “right” up front — some of them just had to be made. 

I’m excited to track the progress, because I know there’s no way I won’t learn something useful. I should show no problem getting great info that’ll help me course correct.

Even though it’s not made a dent in my business yet, it’s been a huge season of personal growth. Problem solving, sitting with the uncertainty of stuff I don’t understand, making executive decisions and handling the outcomes – all of it has improved my life outside business.

Food’s been more exciting because I’ve been figuring out how to make my favorite restaurant dishes at home. I got all my personal and business tax stuff pulled together by February 1 (never been this fast!). I stopped doomscrolling at night and acquired the taste for replacing it with challenging, rewarding work.

Highly recommend spending a whole quarter focused on one big business goal.

Who can help ME get AI visibility?

I’d love to help! My newsletter is an actionable tip to boost your AI visibility – and you can read it in less than 60 seconds. You can sign up for it (the Checkmate Minute) here.

Want me to do the AI visibility stuff for you? Let’s talk. Here’s my calendar.