How I use AI to tighten my writing, curate my back-catalogue, keep categories/tags consistent, and add polish—without losing my voice.
I don’t use AI to “write for me.”
I use it like a tough, fast editor who never gets tired—someone who can read a messy draft, spot what’s unclear, and help me turn it into something that actually lands. The bonus is that the same tool can help me do the unglamorous work that makes a photography blog grow over time: content curation, internal linking, and the steady discipline of categories and tags.
If you’ve ever done the SEO “plumbing” work—cleaning menus, fixing taxonomy, making your site make sense—you know the pain. I’ve written about that grind already (and why it mattered):
- Nine Days of Plumbing: Getting Serious About SEO
- How I Used ChatGPT to Rebuild My Site Taxonomy (and why I didn’t see the problem sooner)
- ChatGPT Website Critique: Action Plan
- The Website is the Home Base (everything else is just a flyer)
This post is the “how” behind that work—my repeatable process for using AI in the writing pipeline, especially where it shines: editing, curation, and consistency.
The part people forget: this used to be a team sport
If I’d been writing for a magazine in the past, “writing” wouldn’t have meant sitting alone at a keyboard and doing everything end-to-end.
There would have been people.
An editor who helps you find the story hiding in the draft. A copy editor who fixes the small stuff and catches inconsistencies. Someone who thinks about headlines and subheads. Someone who ensures the piece fits the publication’s voice and standards. Someone who keeps the archive coherent—categories, labels, continuity.
That’s not nostalgia. That’s how professional publishing worked: a writer plus an editing and production system.
As a solo blogger, you don’t get that system. Not unless you can pay for it.
This is where AI actually changes the game for someone in my situation. For roughly $24 a month, I can access something that behaves like an editing and curation staff I could never afford to hire.
- Not because AI is “better than people.”
- Because it makes parts of the professional workflow accessible.
AI serves many functions that, in a traditional publishing environment, would have been handled by humans. As a blogger, I could never afford those roles at scale. With AI, I can borrow the functions of those roles—on demand—while keeping my own intent, taste, and voice.
Where AI actually helps (and where it doesn’t)
AI is great at:
- Structural editing: “What’s the point?” “What’s redundant?” “What’s missing?”
- Line editing: clarity, flow, tightening sentences without changing meaning
- Consistency checks: headings, voice, terminology, capitalization, style rules
- Curation: suggesting internal links, series order, “related posts,” category placement
- Taxonomy hygiene: spotting tag duplication, near-synonyms, and category drift
- Production polish: helping with elements around the writing that make posts feel finished (like feature images)
AI is not great at:
- Being you. It will happily sand off your edges. Those edges are your voice.
- Truth. It can sound confident and still be wrong. You still verify facts and names.
- Taste. It can help you see options, but it can’t choose your direction.
My rule: AI can edit my work. It can’t own my intent.
The writing process (my version): draft → revise → edit → proof
Most writing frameworks break down into a few stages: getting the ideas out, reshaping them, polishing language, and then catching final errors. I treat AI like a tool I can plug into each stage—but I’m still driving.
1) Draft: I write the messy version first
I write like I’m talking to one person. I don’t worry about perfection. I aim for truth + momentum.
If I try to “sound publishable” too early, I get stiff. I’d rather produce a strong block of raw material and shape it after.
2) Revise: AI helps me find the actual post hiding inside the draft
This is where AI is the most useful, and it maps cleanly to what a real editor used to do.
I’ll paste the draft and ask for:
- A stronger structure (what should be moved, cut, expanded)
- A clearer throughline (what’s the main argument, and is it consistent)
- A better lead (does the opening earn attention)
Example prompt I use: “Act as an editor. Identify the core thesis in one sentence. Then propose a tighter outline using only the material already in my draft. Flag anything that’s repetitive or off-topic.”
3) Edit: AI helps me tighten, but I protect my voice
Once the structure is right, I’ll do a second pass for sentence-level clarity—closer to what a copy editor would do.
Key constraint: do not make it generic content.
I’ll literally instruct the model: “Keep my direct tone. Don’t add motivational fluff. Don’t add new claims. Make it clearer and tighter.”
4) Proof: I do the final human pass
This is where I slow down and read it as a reader would. Out loud if needed. I’m looking for:
- typos
- missing words
- awkward rhythm
- broken links
- headings that don’t match the section
AI can assist here too, but the final proof is still on me—because I’m the one publishing under my name.
The hidden win: AI-assisted content curation
This is the part most people skip, and it’s where blogs quietly die.
A blog isn’t a timeline. It’s a library. And libraries need:
- shelves (categories)
- labels (tags)
- cross-references (internal links)
- featured paths (series + “start here” routes)
In a magazine environment, this was partly handled by editorial structure and production workflow. As a solo blogger, it’s easy to ignore because it’s invisible work. AI makes it faster, and that means it actually gets done.
Curating internal links (without turning posts into SEO spam)
I’ll ask AI to suggest internal links based on themes, not keywords. If I’m writing about editing and consistency, these are natural internal links from my site:
- How I Used ChatGPT to Rebuild My Site Taxonomy
- Nine Days of Plumbing: Getting Serious About SEO
- Keeping Score on My Visibility (January/February 2026)
Prompt I use: “Suggest 3–5 internal links from my existing posts that genuinely deepen this topic. For each, tell me why it fits and where it should appear (intro/middle/close).”
That “why” is the filter. If the reason is weak, the link doesn’t go in.
Building “reading paths”
AI can also help me assemble curated sequences, like:
- “Start here if you’re building a photography practice”
- “Start here if you’re serious about B&W / chiaroscuro”
- “Start here if you want the fundamentals (Adams)”
That turns older posts into a guided experience instead of a dead archive.
Feature images: I’m a photographer, not a designer
Here’s another place AI helps in a very practical way: feature images.
I’m a visual artist, but my medium is photography—not graphic design. And the reality of running a modern blog is that posts need a consistent visual “cover” in archives, on social shares, and in previews. A strong feature image doesn’t just look nice; it signals that the post is finished and worth clicking.
When I need it, I use ChatGPT to generate feature images—usually simple, clean designs that match the tone of the article. Sometimes it’s typography over a minimal background. Sometimes it’s a concept image that reinforces the theme. Either way, it adds real polish to my posts without me having to become a designer or lose hours fussing with layouts.
It’s the same theme again: in a professional environment, design support existed. As a solo blogger, it usually doesn’t. AI makes that “production layer” accessible.
Tags and categories: definition beats invention
Tags and categories only work if they mean the same thing consistently.
My category rule
- Categories are shelves. Few, stable, long-lived.
- Tags are labels. Many, but controlled.
A category answers: “Where does this belong?”
A tag answers: “What is this about?”
Since this post is about workflow, SEO hygiene, and site organization, it belongs in:
Category: Social Media Visibility
Tag rules that keep me sane
AI helps most when I treat tagging like a system, not a mood—again, the kind of maintenance a publication would have handled with people and process.
Rules:
- Prefer reuse over novelty. If a tag already exists, use it.
- Kill near-duplicates. “SEO” vs “Search engine optimization” — pick one.
- Avoid one-off tags. A tag used once isn’t navigation; it’s clutter.
- Use tags to connect clusters. (Editing, taxonomy, internal linking, workflow)
Prompt I use: Here is my draft + my existing tag list. Suggest up to 10 tags, prioritizing existing tags. Also flag duplicates/near-duplicates and propose one canonical wording.”
That last line—canonical wording—is where AI becomes a taxonomy assistant.
My repeatable AI workflow (the version I can run every time)
Here’s the loop I aim for:
- I write the draft (messy, honest)
- AI gives structural notes (editor function)
- I revise (human judgment + intent)
- AI line-edits with voice constraints (copy editor function)
- AI suggests internal links + “related posts” (curation function)
- AI suggests categories/tags — I enforce consistency (production function)
- AI helps generate a clean feature image when needed (design support function)
- I proof + publish (final accountability)
- I measure (because visibility is part of the work)
This is the point: AI doesn’t replace craft. It reduces friction. And friction is what stops consistency.
The standard I’m aiming for
Not more content. Better signal.
- posts that read clean
- navigation that makes sense
- tags that actually connect ideas
- categories that don’t sprawl
- internal links that create a path
- a site that behaves like a home base, not a feed
- presentation that looks intentional, not accidental
Because the long game isn’t going viral. It’s building something that holds together.