Name: SEO Assistant with LLM
Tagline: Generate SEO-optimized focus keyphrases and meta descriptions using AI
Current status: Version 1.0.0 — functional WordPress plugin supporting three LLM providers
First commit / project start: Pre-migration commit March 27, 2026 (e65fdaf2); file creation dates not available
SEO Assistant with LLM is a WordPress plugin that generates AI-powered SEO recommendations — specifically focus keyphrases and meta descriptions — for any post or page. It works across all three major WordPress editing experiences (Classic Editor, Gutenberg Block Editor, and Elementor) by injecting a “Generate SEO” button into each editor’s toolbar. When triggered, the plugin sends the page title and body content alongside a built-in SEO best practices knowledge base to a user-selected LLM (OpenAI, Anthropic, or Google), and presents recommendations in a modal overlay with one-click copy functionality.
Platform: WordPress plugin
Deployment: Self-hosted WordPress (5.0+, PHP 7.4+)
Primary interface: In-editor button → modal popup with results and copy buttons
Primary user: WordPress site owners and content creators who need SEO optimization but lack SEO expertise
Secondary users: Elementor-based site builders
User environment: WordPress admin dashboard (post/page editor)
Creating effective SEO focus keyphrases and meta descriptions requires specialized knowledge that most content creators don’t have. Existing SEO plugins (Yoast, Rank Math) tell you what’s wrong but don’t generate recommendations. This plugin generates the actual recommended keyphrase and meta description using AI, applying SEO best practices automatically.
| Need | How the product addresses it | Source of evidence |
|---|---|---|
| SEO guidance without expertise | AI generates recommendations using embedded best practices KB | class-seo-assistant-api.php prepare_prompt() |
| Editor-agnostic workflow | Separate JS integrations for Classic, Gutenberg, and Elementor | classic-editor.js, gutenberg-editor.js, elementor-editor.js |
| Provider flexibility | Three AI providers with configurable model selection | class-seo-assistant-settings.php |
| Customizable SEO rules | Admin-editable knowledge base shipped with comprehensive defaults | class-seo-assistant-knowledge-base.php |
Manually writing keyphrases and meta descriptions based on personal judgment, or using Yoast/Rank Math’s analysis tools that critique but don’t generate content. Some users used separate AI tools (ChatGPT) and copy-pasted results back into WordPress.
Primary category: WordPress SEO plugins with AI content generation
Market maturity: Established (SEO plugins) but emerging (AI-powered SEO recommendation) [CLAUDE NOTE: inferred]
Key dynamics: Yoast and Rank Math dominate WordPress SEO; AI-powered generation is a differentiator [CLAUDE NOTE: inferred]
| Product / Company | Approach | Strengths | Key gap this project addresses | Source |
|---|---|---|---|---|
| ⚡ Yoast SEO | Analysis + scoring | Market leader, comprehensive analysis | Doesn’t generate keyphrases or descriptions | General knowledge |
| ⚡ Rank Math | Analysis + suggestions | Feature-rich, schema markup | AI features locked behind premium tier | General knowledge |
| ⚡ All in One SEO | Analysis + sitemap | Beginner-friendly | No AI generation in free tier | General knowledge |
| ⚡ SEOPress | Lightweight analysis | Performance-focused | No AI content generation | General knowledge |
Focused AI generation tool that complements (not replaces) existing SEO analysis plugins. Lightweight, single-purpose, multi-provider, and works across all WordPress editors. [CLAUDE NOTE: inferred]
Limited defensibility as a standalone plugin — the concept is straightforward. Value comes from the curated knowledge base, triple-editor support, multi-provider architecture, and ITI ecosystem integration. The n8n adapter path provides enterprise-grade orchestration when deployed with the ITI stack. [CLAUDE NOTE: inferred]
Requirements were documented in a single Requirements.md file using a hierarchical bullet format. The spec is concise (12 lines) and covers the complete product scope.
Hard constraints: WordPress 5.0+, PHP 7.4+ (per README)
Explicit non-goals: NOT FOUND — add manually
| Decision | Alternatives considered | Rationale | Evidence source |
|---|---|---|---|
| Multi-provider (OpenAI + Anthropic + Google) | Single provider | Flexibility; no vendor lock-in | class-seo-assistant-settings.php |
| Editable knowledge base in WordPress | Hard-coded prompts | Admin customization without code changes | Requirements.md line 8 |
| Modal popup for results | Sidebar panel, inline insertion | Non-intrusive, universal across editors | Requirements.md line 10 |
| JSON-only output contract | Free-form text | Reliable parsing for keyphrase + description | class-seo-assistant-api.php |
| Body truncation to 4000 chars | Full content | Token budget management | class-seo-assistant-api.php prepare_prompt() |
KB type: Editable text document stored as WordPress option (seo_assistant_knowledge_base)
Location in repo: includes/class-seo-assistant-knowledge-base.php (default content in get_default_content())
Estimated size: ~2000+ words of SEO best practices (shipped default)
The knowledge base is a single document (not a file tree) stored in the WordPress wp_options table. Default content covers:
SEO Best Practices Knowledge Base (default)
├── Focus Keyphrase Guidelines
│ ├── Length: 2-5 words
│ ├── Specificity rules
│ └── Intent matching
├── Meta Description Guidelines
│ ├── Length: 150-160 characters
│ ├── Formulas and patterns
│ └── Compelling copy rules
├── Content-Type Specific Guidelines
├── Quality Checklist (10 items)
├── Five Worked Examples
└── Notes for AI Generation (10 bullets)
| Category | Files / format | Purpose | Update frequency |
|---|---|---|---|
| SEO best practices | Single WordPress option (HTML/Markdown) | Injected into every LLM request as context | Admin-editable at any time |
The default knowledge base was authored as part of the plugin development, containing comprehensive SEO best practices, formulas for meta descriptions, content-type-specific guidelines, a quality checklist, and five worked examples. It ships as a PHP heredoc string and is stored in WordPress options on plugin activation. Administrators can edit it through a TinyMCE editor in the admin dashboard.
System prompt approach: Single prompt constructed in prepare_prompt() — instructs the model as an “SEO expert” with:
{"focus_keyphrase":"...","meta_description":"..."}Key behavioural guardrails: JSON-only output enforcement (parsed via regex)
Persona / tone configuration: “SEO expert” role instruction
Tool use / function calling: None — single-turn generation only
AI-assisted development using Cursor IDE. The plugin was built to spec from the 12-line Requirements.md document. The architecture follows standard WordPress plugin patterns (singleton, hooks, AJAX).
| Phase | Approximate timeframe | What was built | Key commits or milestones |
|---|---|---|---|
| Initial build | Pre-Mar 2026 | Complete v1.0.0 plugin with all three editor integrations | Files in repository |
| Pre-migration commit | Mar 27, 2026 | Full workspace committed to ITI monorepo | e65fdaf2 |
| Chat/API handler updates | Mar 31, 2026 | Updates to chat and API handlers | 69ea8003 |
| Challenge | How resolved | Evidence |
|---|---|---|
| Triple-editor integration | Separate JS files for Classic (jQuery), Gutenberg (wp-plugins API), and Elementor (panel API) |
Three dedicated JS files in assets/js/ |
| JSON extraction from LLM output | Regex-based extraction: /\{[^\}]"focus_keyphrase"[^\}]\}/s |
class-seo-assistant-api.php process_response() |
| Multi-provider API differences | Provider switch with separate HTTP call logic and response parsing per provider | class-seo-assistant-api.php call_*_api() methods |
| n8n adapter return format mismatch | When adapter returns text, raw text is returned without JSON parsing — may need normalization |
class-seo-assistant-api.php Anthropic path [CLAUDE NOTE: potential issue noted] |
| Model / API | Provider | Role in product | Integration method |
|---|---|---|---|
| GPT-4o | OpenAI | SEO recommendation generation (default option) | POST https://api.openai.com/v1/chat/completions |
| GPT-4o-mini | OpenAI | Lighter/cheaper alternative | Same endpoint |
| GPT-4-turbo | OpenAI | Higher-quality alternative | Same endpoint |
| GPT-3.5-turbo | OpenAI | Budget option | Same endpoint |
Claude Sonnet 4.5 (claude-sonnet-4-5-20250929) |
Anthropic | Default Anthropic model | POST https://api.anthropic.com/v1/messages |
Claude Sonnet 4 (claude-sonnet-4-20250514) |
Anthropic | Alternative | Same endpoint |
Claude Haiku 4.5 (claude-haiku-4-5-20251001) |
Anthropic | Lightweight option | Same endpoint |
Claude 3.5 Sonnet (claude-3-5-sonnet-20241022) |
Anthropic | Legacy option | Same endpoint |
| Gemini Pro | Default Google model | Generative Language API | |
| Gemini 2.0 Flash | Fast alternative | Same endpoint | |
| Gemini 1.5 Pro / Flash | Additional options | Same endpoint | |
| ITI Workflow Adapter (optional) | ITI / n8n | Anthropic requests routed through n8n when shared library present | ITI_Workflow_Adapter::run_agent('seo-assistant', ...) |
| Tool | Category | Purpose |
|---|---|---|
| ITI Workflow Adapter (optional) | Orchestration | Routes Anthropic calls through n8n when ITI shared library is present |
| WordPress AJAX | Transport | Frontend-to-backend communication for generation requests |
| Tool | How used in build |
|---|---|
| Cursor IDE | Primary development — plugin code, JS integrations, knowledge base authoring |
| Version / Phase | Date | Summary of changes | Significance |
|---|---|---|---|
| v1.0.0 | Pre-Mar 2026 | Initial release: three editors, three providers, editable KB, modal UI | Full feature set |
| API handler update | Mar 31, 2026 | Chat and API handler updates across products | 69ea8003 |
NOT FOUND — the product appears to have been built to spec from Requirements.md without significant pivots.
yourusername, yourwebsite.com)| Artifact | Path | Type | What it shows |
|---|---|---|---|
| Admin CSS | assets/css/admin.css |
Stylesheet | Full modal, spinner, results, copy buttons, responsive rules |
| Gutenberg panel | assets/js/gutenberg-editor.js |
JS | DocumentSettingPanel “SEO Assistant” in block editor sidebar |
| Document | Path | Type | Status |
|---|---|---|---|
| README.md | README.md |
User documentation | Complete — install, configure, usage, FAQ, changelog |
| Requirements.md | Requirements.md |
Product spec | Complete — 12-line hierarchical bullet spec |
| Artifact | Path | Description |
|---|---|---|
| Default SEO knowledge base | includes/class-seo-assistant-knowledge-base.php (heredoc) |
~2000+ word SEO best practices document |
| Nested plugin copy | seo-assistant-plugin/ |
Duplicate directory for zip packaging |
The plugin addresses a gap between SEO analysis tools (which tell you what’s wrong) and AI tools (which can generate content but aren’t integrated into the editor workflow). The requirement to support all three WordPress editor types (Classic, Gutenberg, Elementor) suggests targeting the broadest possible WordPress user base. [CLAUDE NOTE: inferred from Requirements.md and product design]
NOT FOUND — add manually. The product fills an obvious gap in the WordPress plugin ecosystem where SEO analysis exists but AI-powered generation is typically locked behind premium tiers.
That a lightweight, focused WordPress plugin for AI-generated SEO recommendations — working across all editors and supporting multiple AI providers — is more useful than adding AI features to a monolithic SEO plugin. [CLAUDE NOTE: inferred]
Requirements → single-purpose plugin → multi-editor support → multi-provider architecture → editable knowledge base → ITI ecosystem integration (Workflow Adapter).
What works well: Clean single-purpose architecture; triple-editor support; multi-provider flexibility; editable knowledge base with comprehensive defaults
Current limitations: Placeholder author/URI metadata; no automated tests; JSON extraction via regex (fragile); n8n adapter path may not fully normalize output for modal display
Estimated completeness: 85% — functional product, needs polish and testing [CLAUDE NOTE: inferred]
_Manual input required — this section cannot be populated automatically._
| File / Path | What it contributed |
|---|---|
README.md |
Product description, features, installation, configuration, FAQ, changelog |
Requirements.md |
Original 12-line product specification |
seo-assistant.php |
Plugin header (v1.0.0), singleton bootstrap, AJAX handler, admin menu |
includes/class-seo-assistant-api.php |
Multi-provider API calls, prompt construction, JSON parsing, ITI Workflow Adapter path |
includes/class-seo-assistant-settings.php |
Provider/model configuration, available models list, test connection |
includes/class-seo-assistant-knowledge-base.php |
Default KB content (~2000+ words), CRUD methods |
admin/settings-page.php |
Settings UI with provider/model/key configuration |
admin/knowledge-base-page.php |
TinyMCE editor for KB, save/reset, tips |
assets/js/admin.js |
Modal UI, generation workflow, copy functionality, character counting |
assets/js/classic-editor.js |
Classic Editor toolbar button integration |
assets/js/gutenberg-editor.js |
Gutenberg DocumentSettingPanel registration |
assets/js/elementor-editor.js |
Elementor panel footer integration |
assets/css/admin.css |
Full UI styling |
.gitignore |
Standard WordPress/Node/env exclusions |
| Git log (2 path-specific commits) | Commit history |
The WordPress SEO plugin category sits at the center of two colliding forces: the vibe coding explosion and the convergence of frontier LLM capabilities. With 82% of developers now using or planning to use AI coding tools, and platforms like Bolt.new and Lovable enabling non-technical builders to ship functional applications in hours, the barrier to creating an “AI-powered SEO plugin” has dropped to near zero. Any developer — or motivated marketer with Cursor open — can build a plugin that sends content to Claude or GPT and returns a keyphrase and meta description. That’s essentially what SEO Assistant v1.0.0 does, and it’s no longer a differentiating capability.
The more significant shift is what’s happening to the AI models themselves. Claude 4, GPT-5, and Gemini 2.5 now score within 10% of each other on production tasks, and pricing has compressed dramatically (Gemini 2.5 offers roughly 80% of frontier capability at 8% of competitor cost). Multi-provider support — which SEO Assistant listed as a differentiator — is now a configuration toggle in most frameworks. The “uses AI” or “powered by Claude” positioning that worked in early 2025 has lost its force. When every competitor integrates the same models at similar cost, differentiation has to come from the domain expertise embedded in the product, the workflow integration, and the judgment about where AI helps versus where it misleads.
The third dynamic is the emergence of Generative Engine Optimization (GEO) as a distinct discipline from traditional SEO. Content optimization is no longer just about ranking in Google’s blue links — it’s about being cited by ChatGPT, Perplexity, and Google AI Overviews. Surfer SEO pivoted explicitly to “AI Search Guidelines” in early 2026, and Clearscope added AEO-specific content grading. This shift changes what an SEO plugin needs to do: it’s no longer enough to generate a good keyphrase and meta description. The plugin needs to evaluate whether content is structured for AI extraction — entity clarity, factual density, question-answer formatting, source citations. SEO Assistant doesn’t do any of that yet.
Since SEO Assistant v1.0.0 shipped, the competitive environment has changed materially:
| Competitor | What Changed | Direct Impact |
|---|---|---|
| Rank Math | Content AI v2.0 with real-time NLP suggestions on free tier; AI Link Genius (v3.0.108, March 2026) for automated internal linking; llms.txt support and AI search traffic tracking | Our core keyphrase generation is now a free feature in Rank Math |
| Yoast SEO | AI Optimize button rewrites meta descriptions and titles via GPT (Premium, $99-119/year); AI-assisted internal linking workouts | Direct competition for our primary use case |
| AIOSEO | Full AI Suite (v4.9.4, Feb 2026) including AI content generator, FAQ blocks, social posts, E-E-A-T author profiles, and Writing Assistant via SEOBoost integration | Far broader AI scope than our single-feature generation |
| Surfer SEO | AI Search Guidelines for optimizing AI citation; Workspaces + Brand Kit per domain ($89-219/mo) | GEO/AEO capabilities we entirely lack |
| Clearscope | AEO optimization with content grading for AI platform citations ($129-399/mo) | Addresses the GEO shift we haven’t responded to |
| SEOPress | Now supports 5 AI providers | Matches our multi-provider approach |
Eroded differentiators:
| Feature We Claimed as Unique | Current Status |
|---|---|
| AI keyphrase generation | Available free in Rank Math, in Yoast Premium, and in AIOSEO AI Suite |
| AI meta description generation | Yoast AI Optimize, AIOSEO AI Content Generator, and Rank Math Content AI all do this |
| Multi-provider LLM support | SEOPress offers 5 providers |
| Editable knowledge base | Still somewhat unique — AIOSEO’s SEOBoost provides similar customization, but no competitor has our admin-editable KB approach |
What remains ours:
The roadmap is organized in four tiers, prioritized by competitive urgency.
Tier 1 (next build cycle) closes the gaps that matter most. AI SEO title generation (S effort) extends our existing prompt and modal to produce titles alongside keyphrases — minimal code change for a feature every competitor now has. Content SEO scoring (L) adds a 0-100 score against keyphrase usage, readability, heading structure, and meta completeness. GEO/AEO readiness scoring (L) is the strategic move: scoring content for AI citation likelihood — entity clarity, factual density, question-answer formatting, structured data presence. This cross-pollinates with the AEO Optimizer product, sharing scoring methodology rather than duplicating effort. Real-time keyword/NLP suggestions (L) responds directly to Rank Math Content AI v2.0. JSON extraction hardening (S) replaces our fragile regex parsing with provider-specific JSON mode flags.
Tier 2 builds on the scoring foundation. Internal linking suggestions (L) — Rank Math shipped AI Link Genius in March 2026 and we have nothing comparable. Section-level rewrite suggestions (L) go beyond scoring to generate specific, actionable improvements using Claude’s reasoning. Yoast/Rank Math integration (M) reads their existing SEO data and enhances it with AI generation, reinforcing the complement-not-replace positioning. Bulk generation (M) handles multiple posts from a dashboard view.
Tier 3 adds strategic differentiation: AI-powered schema recommendation and JSON-LD generation (L), iterative multi-turn refinement (M), E-E-A-T content guidance (M), content gap analysis via Claude (L), and SEO performance tracking (XL).
Tier 4 explores llms.txt generation, AI visibility monitoring, multi-language generation, content briefs, WordPress.com compatibility, and WooCommerce product SEO.
The sequencing logic: Tier 1 items come first because without content scoring and GEO readiness, we’re a single-feature plugin in a market where that feature is now free. Scoring is the prerequisite for everything downstream — internal linking, section rewrites, and schema all require content analysis infrastructure.
Several Skills created as part of the April 2026 roadmap cycle are directly relevant to SEO Assistant’s development:
Current strengths: SEO Assistant does one thing and does it well within a clean, lightweight architecture. The triple-editor support (Classic, Gutenberg, Elementor) is genuinely rare. The editable knowledge base gives administrators control over the SEO best practices that inform generation. The n8n Workflow Adapter integration demonstrates enterprise-grade orchestration. And the complement-not-replace positioning is smart — users don’t want to rip out Yoast or Rank Math.
Acknowledged gaps: The product’s core feature — generating a keyphrase and meta description via AI — is now available for free in Rank Math and in Yoast Premium. We have no content scoring, no internal linking suggestions, no GEO/AEO readiness evaluation, no schema generation, no readability analysis, and no bulk capabilities. The JSON extraction via regex is fragile and needs hardening. The plugin is a v1.0.0 single-feature tool in a market that moved to multi-feature AI SEO suites in Q1 2026.
What we’re watching: The GEO/AEO shift is the most consequential market change. Whether content is “good for SEO” is increasingly secondary to whether content is structured for AI extraction and citation. We’re also watching the pricing compression in AI APIs — as model costs drop, the cost barrier to offering AI-powered features in free-tier plugins disappears, further commoditizing the generation capability.
This product demonstrates how we approach WordPress content optimization — with a focus on workflow integration across all editor types and a knowledge-base-driven architecture that can evolve as SEO best practices change. The roadmap reflects what we’ve learned about where the market moved while we built v1.0.0, and the shared Skills infrastructure means the next build cycle can move faster.