The Agentic WordPress Prompt Book
Copy-paste Claude and ChatGPT prompts for editing your WordPress site in plain English. You don't learn tools. You talk to your site.
Get the ebook version
The whole book as a designed PDF, sent to your inbox. i'll also send you what's new in Respira and the best prompts customers discover, occasionally, and you can stop it with one click.
This is a book of AI prompts for WordPress: the exact wording that gets Claude, ChatGPT, or any MCP-connected agent to edit your real site well. Respira gives your assistant around 230 tools across 16 page builders, and you will never need to memorize any of them. You say what you want in plain English and the agent picks the right tool, in the right order, with the safety rails on.
This book is the companion for that: how to phrase things, what happens under the hood when you do, and the small set of habits that separate a great session from a frustrating one. Every example here was run against a real site before it was written down.
Start every session with the primer
Paste this first, once per session, before any real work:
You're connected to my WordPress site through Respira. Before doing anything, detect my page builder and read my site's structure, then follow Respira's safe workflow: edit through the builder's native modules (never raw HTML), and duplicate or snapshot a page before changing it live. Confirm what builder i'm on and what you can do.
Every builder stores content differently: shortcodes, block comments, JSON in hidden fields. When the agent detects your builder first, everything it writes lands in your builder's native format and stays editable in your visual builder. Skip the primer and the agent may guess, and a wrong guess is how pages end up as one opaque HTML blob.
The primer also lives in your WordPress admin under Respira → Prompts, next to 20 ready-made workflow prompts. Add one line to it when you know the session's intent: "today we're editing the About page, read its structure first", or "today is read-only: audit, analyze, report".
The four write workflows
Everything you ask for routes through one of four patterns. You don't have to name them, but knowing them helps you phrase asks that land on the right one.
1 · Small edit
Find it, change it
Text, dates, colors, one image, one button. The agent locates the element and updates only that element. "change the button on the contact page to say 'Get a quote'"
2 · Redesign existing
Read, modify, write back
The agent extracts the full structure, modifies the tree, injects it back in one atomic write. A snapshot makes it reversible. "restructure my services page so each service is its own section"
3 · Build from scratch
Describe the page
New pages build from a nested structure in your builder's vocabulary. For Divi you can say it the natural way: rows containing cols containing modules. If a build can't write every element, it fails with a clear error instead of pretending it succeeded.
4 · Two-pass build
Skeleton first, then fill
Pass one builds the skeleton with placeholder labels; pass two updates each module by label. The most controllable way to build content-heavy pages, and each request stays small.
Very large pages: build a section at a time ("add the pricing section to the draft"), not as one giant request. AI clients cap how much one request can carry. A dedicated streaming build mode is on the roadmap; i haven't shipped that yet.
The four anti-patterns
- Asking for raw HTML. It renders, but your visual builder can't edit it. Ask for the page or section instead; if you have an HTML prototype, say "convert this HTML into my builder's format".
- Skipping the read step. "Change the second section" makes the agent guess. "Read the page structure, then change the section with the testimonials" never misses.
- Editing live without a net. With default settings, edits stage on a duplicate you approve. If you enabled direct editing, say "duplicate the page first" on anything you'd mind losing.
- One giant ask. "Redesign my whole site" produces mush. "Let's start with the homepage hero" produces results, and momentum.
The safety net
Three layers, on by default: duplicate-before-edit (writes stage on a copy; you approve in Respira → Changes before anything replaces the original), snapshots (full-page writes snapshot first; "undo that" restores, "what changed?" diffs), and approval for destructive operations (deletions require an explicit yes).
The phrases that drive it: "undo that last change", "show me what changed", "restore the version from before we started". The full trust model is in the docs.
Worked recipes
Eight end-to-end recipes for the highest-value jobs. Copy, fill the brackets, paste.
Hero redesign
Same design, new message, staged for your approval.
read my homepage structure, then rewrite the hero section to target [audience]: new headline, one-line subhead, and a CTA that says [text]. keep the layout and images, stage it for my approval.
Site-wide find and replace
A reviewed list first, then one bulk pass across up to 100 pages.
find every page that mentions [old thing] and show me the list. then replace it with [new thing] everywhere except [page you want untouched].
SEO refresh on a key page
Mechanical wins applied, strategic calls surfaced for you.
run an SEO analysis on [page]. then fix the mechanical findings: title tag, meta description, heading hierarchy, image alt text, and internal links to related pages. show me anything that needs a judgment call instead of fixing it.
Accessibility pass
Alt text, labels and heading order fixed; judgment calls reported.
scan [page] for accessibility issues. fix missing alt text, unlabeled buttons and broken heading order. give me a short report of what you fixed and what needs a human decision (contrast, link text wording).
New page from scratch (two-pass)
A labeled skeleton you fill section by section. The most controllable way to build.
build a draft [type] page: [list the sections in order, one line each]. use placeholder labels for every text module. then read it back and confirm the structure before we fill in copy.
Case study machine
A custom post type, its fields, a template, and post one, in one conversation.
create a "case study" custom post type with fields for client, industry, result and quote. then build a template page for it and create the first case study from these notes: [paste notes].
Divi 4 to Divi 5 migration check
A readiness report before anything changes.
run a Divi 5 migration readiness audit on my site. tell me which pages and modules would convert cleanly, which won't, and what to fix first. don't change anything yet.
The Monday morning audit
One prioritized list of what to fix this week. Read-only.
give me a site health snapshot: pages changed in the last week, anything that looks broken, SEO issues on the top 5 pages, and one prioritized list of what to fix this week.
Your builder, specifically
Every prompt in this book works on all sixteen builders; Respira translates to whatever your site runs. But each builder has one or two things worth knowing when you phrase an ask. Not sure what you're on? Start with: "what page builder is this site using, and what can you do with it?"
Gutenberg
Native WordPress blocks, edited in place. The most direct pipeline of all sixteen: what the agent writes is exactly what the block editor opens.
Tipname blocks by what they say, not by position: "the paragraph under the pricing heading" beats "the third block".
Elementor
Your page lives as a JSON tree in a hidden field, which is why hand-editing the visible content does nothing. Respira edits that tree through widget names, and can read and update your global colors and typography site-wide.
Tipsay "use my existing global colors and fonts" and new sections come out on-brand instead of hardcoded.
Divi 4
Shortcode-based. Builds accept your natural vocabulary (sections containing rows containing columns containing modules) and every module can carry an admin label.
Tipask for admin labels on every module while building. later edits become one-liners: "update the module labeled Hero Headline".
Divi 5
Block-based. Same prompts as Divi 4; Respira detects which Divi runtime your site is on and writes the right format. Builds carry styling, not just structure.
Tipplanning the upgrade? ask for a "Divi 5 migration readiness audit" first. it is read-only and tells you what converts cleanly.
Bricks
The deepest design-system support in the lineup: global classes, theme styles, reusable components, section presets, and query loops are all readable and editable.
Tipsay "reuse my global classes" or "follow my theme styles" so new sections extend your design system instead of piling up inline styles.
Oxygen Classic
Content lives in hidden fields as a JSON tree of ct_ elements. Structural edits, element updates and rebuilds all work.
Tipbe explicit about which page you mean; Oxygen templates and pages can share names.
Oxygen 6
A new engine under the hood (shared with Breakdance), storing a node tree in a hidden field. Respira speaks to its full element set, including the rich Section, Heading and Button elements.
Tipprompts that name elements plainly ("the heading", "the button row") map cleanly onto its element names.
Beaver Builder
A flat node map in a hidden field. Full read-modify-write: extract, restructure, inject.
Tiprows and modules have internal node ids; let the agent read the page first and reference elements by their text.
Breakdance
A JSON tree in a hidden field, written against the builder's own element schema so the visual builder opens everything cleanly.
Tipworks the same as Oxygen 6 in practice; the two share an engine.
Flatsome (UX Builder)
Shortcodes in the page content, common on WooCommerce shops. Reads, edits and rebuilds supported.
Tipfor shop pages, mention WooCommerce context: "the product grid section on the shop landing page".
Brizy
An envelope format in the page's meta. Full writes supported.
Tipstandard prompts work; read the page first, then edit by description.
WPBakery
Classic-era shortcodes. Great for audits, text refreshes and structure updates on long-lived sites.
Tipa lot of WPBakery sites are candidates for migration; there are ready-made skills for WPBakery to Gutenberg or Bricks (see the skills chapter).
Visual Composer
URL-encoded JSON in a hidden field. Full writes supported.
Tipnot the same product as WPBakery despite the shared history; Respira detects which one you actually run.
Thrive Architect
Serialized content in a hidden field, on conversion-focused landing pages.
Tipreference elements by their copy ("the CTA that says Book a call"), which stays stable across Thrive's markup.
Kadence · Spectra · GenerateBlocks
Block suites riding the Gutenberg pipeline with their own block families. Respira knows their block names and settings.
Tipyou can name the family ("use Kadence blocks for the accordion") or just describe the outcome and let detection handle it.
SeedProd
Read and audit only today. Respira reads landing pages for audits and reports but stages no writes.
Tipask for audits and content inventories; for edits, do them in SeedProd itself for now.
Builder cheat sheet
You never write these payloads by hand; your agent does. This table exists so that when you look over its shoulder, you know what "correct" looks like for your builder. One universal rule: the canonical nesting key is children, and for Divi the natural vocabulary (rows, cols, modules) works too, at any depth, on both Divi 4 and Divi 5.
| Builder | Where content lives | Nesting key | Settings key |
|---|---|---|---|
| Divi 4 | shortcodes in the page content | children (+ rows/cols/modules aliases) | attrs / settings |
| Divi 5 | block comments in the page content | children (+ the same aliases) | attributes |
| Elementor | JSON in a hidden field (_elementor_data) | elements | settings |
| Bricks | flat JSON list in a hidden field | children | settings |
| Gutenberg | native block comments | innerBlocks | attrs |
| Beaver Builder | flat node map in _fl_builder_data | children | attributes |
| Breakdance | JSON tree in _breakdance_data | children | properties |
| Oxygen Classic | JSON tree of ct_ elements in hidden fields | children | options |
| Oxygen 6 | node tree in _oxygen_data | children | options |
| WPBakery | shortcode-derived flat map | children | attributes |
| Brizy | envelope in the brizy meta field | children | attributes |
| Flatsome (UX) | shortcodes in the page content | elements | attributes |
| Thrive Architect | serialized content in tve_updated_post | children | attributes |
| Visual Composer | URL-encoded JSON in vcv-pageContent | children | attributes |
Builder-specific notes worth knowing: Divi 4 has a per-attribute writer, so give modules admin labels while building and later edits become one-liners. Elementor keeps everything in a hidden field, which is why hand-editing the page content does nothing. Bricks has the deepest tool set here (global classes, theme styles, components). Gutenberg is WordPress-native and round-trips most forgivingly. The ebook has the full sheet per builder with payload examples.
Skills: prompts someone already perfected
Before writing a long prompt from scratch, check the skills catalog. A skill is a ready-made, tested playbook for a whole job: you copy it into your AI client once and then trigger it with a sentence. Respira ships 37 of them, in every plan, browsable under Skills in your dashboard.
- Migrations (16 skills). Divi, Elementor, WPBakery, Oxygen, Beaver Builder, Brizy, Thrive or Visual Composer to Gutenberg, Bricks, Breakdance or Oxygen: each pairing gets a step-by-step skill that reads the source pages and rebuilds them natively.
- Audits. conversion-audit, technical-debt-audit, woocommerce-health-check, mobile-experience-report, stale-content-detector. Read-only reports you can run monthly.
- Building. html-to-bricks, figma-to-elementor, build-oxygen6-page, page-template-library, custom-post-type-architect, design-system-synthesizer.
- Content and SEO. seo-aeo-amplifier, internal-link-builder, brand-voice-synthesizer, wordpress-ai-image-optimizer.
- Meta. prime-the-agent (the primer from chapter one, packaged), wordpress-site-dna, respira-setup-assistant, activity-report-composer.
The pattern: install or paste the skill, then ask in plain English. "run the conversion audit on my homepage". The skill carries the expertise so your prompt doesn't have to.
Playbooks: teach your site a repeatable job
When you and your agent work out a workflow you'll want again, you don't have to re-explain it next month. Ask the agent to save it as a playbook: a named, typed sequence of steps stored on your site itself. Once saved, it shows up as a tool the agent can run on demand.
save what we just did as a playbook called page-outline-report, so next time i can just say "run the page outline report".
Playbooks are validated when they're saved, not when they run: a step that references a missing tool or a missing argument is rejected on the spot with a message telling the agent exactly what to fix, so a broken playbook can't sit there looking healthy. "List my playbooks" and "delete the playbook called X" work the way you'd expect. This is a newer feature; skills are the curated library, playbooks are the ones you grow yourself.
Frequently asked questions
What are AI prompts for WordPress? +
They are plain-English instructions you give an AI assistant that is connected to your WordPress site, so it can read and edit real pages instead of just generating text for you to paste. With Respira installed, prompts like "rewrite my homepage hero" or "fix the SEO on my services page" become live edits in your own page builder, staged for your approval.
Do these prompts work with Claude and ChatGPT? +
Yes. Respira connects your WordPress site to Claude, ChatGPT, and any other AI client that speaks MCP (Model Context Protocol). Every prompt in this book is written in plain English, so the same wording works in whichever assistant you use.
Which page builders do the prompts work with? +
All sixteen builders Respira supports: Gutenberg, Elementor, Divi 4 and Divi 5, Bricks, Oxygen Classic and Oxygen 6, Beaver Builder, Breakdance, Flatsome, Brizy, WPBakery, Visual Composer, Thrive Architect, and the Kadence, Spectra and GenerateBlocks block suites. The agent detects your builder and writes in its native format, so the same prompt works everywhere.
What is a priming prompt for WordPress? +
The first message of a session. It tells the agent to detect your page builder, read your site structure, and follow the safe workflow before making any change. Starting every session with the primer in chapter one is the single biggest quality improvement you can make.
Is it safe to let AI edit my WordPress site? +
With the right rails, yes. Respira stages edits on a duplicate you approve, snapshots full-page writes so "undo that" works, and requires an explicit yes for destructive operations like deletions. The safety net chapter of this book explains all three layers.
Do i need to know code to use these prompts? +
No. Every prompt in this book is plain English. You describe the outcome ("change the button on the contact page to say Get a quote") and the agent picks the right tools. The cheat sheets exist only for the day you want to check its work.
Want this as a book?
The designed PDF version, with the full per-builder sheets and payload examples.
Email me the book
