Skip to main content

The Complete Guide to AI in WordPress Page Builders (2026)

Posted: May 1, 2026
Updated: May 23, 2026
52 min read

You asked Claude or ChatGPT to design a landing page. It nailed it. You copied the HTML, pasted it into your page builder, and watched the layout collapse into unstyled divs.

That’s not a prompt problem. It’s an architecture problem — and the way each major builder is solving it has split the WordPress AI market into two distinct camps. As of May 2026, the gap between them is the most important difference between WordPress page builders, and the third-party ecosystem around each has grown faster than the builders themselves.

This is where each tool stands, what changed in the last few months, the third-party plugins quietly doing more than the builders’ first-party AI, and the framework I’d use to evaluate any AI feature shipping into your stack.

How we got here: a brief history of AI in WordPress (2020–2026)

To understand why the WordPress AI landscape looks the way it does in May 2026, it helps to walk back six years and trace the inflection points that brought us here. The first thing to notice is that “AI in WordPress” as a phrase didn’t meaningfully exist until 2023. The category we’re discussing in this article is barely three years old.

Before the AI era (2020–2022). In 2020, WordPress page builders competed entirely on visual design tools, drag-and-drop UX, performance, and ecosystem. Elementor (launched 2016) had several million installs and dominated the marketing-page builder category. Divi (2014) was the Elegant Themes flagship with a loyal designer following. Beaver Builder (2014) and WPBakery held the developer-friendly and legacy-Visual-Composer markets. Oxygen had launched in 2019 as the performance-focused dev builder. Bricks Builder launched in 2021 with explicit ambition to take the developer crowd from Oxygen. YOOtheme Pro continued maturing as the design-quality option, still focused mostly on Joomla and WordPress crossover. None of these had AI on the roadmap. Differentiation was about which framework you preferred, which design tokens were exposed, and which third-party widget ecosystem you bought into.

Outside WordPress, things were stirring. OpenAI released GPT-3 in June 2020 via API access. The technology was impressive in developer demos — write code, write essays, answer questions — but it stayed mostly in research circles and dev experiments. Non-technical users hadn’t heard of it. Page builder vendors hadn’t either.

The first commercial application of large language models was marketing content. Jasper (originally Jarvis) launched in January 2021. Copy.ai and Writesonic launched the same season. The thesis: marketers would pay for AI-generated copy. They did. By late 2021, Jasper had over 50,000 paying customers and $35 million in revenue. By October 2022, Jasper raised a $125 million Series A at a $1.5 billion valuation. The AI writing category was a real business before WordPress users could spell “GPT.”

The 2022 picture: AI writing was a parallel ecosystem WordPress users dabbled in by copy-pasting content from Jasper into the WordPress editor. Notion was building Notion AI (launched February 2023 but in development through 2022). Hugging Face was the dominant model hub. Stable Diffusion launched in August 2022 and kicked off the AI image generation explosion. None of this had touched the WordPress builder market in any meaningful way.

Then on November 30, 2022, OpenAI released ChatGPT. It hit one million users in five days, a hundred million users in two months. The mainstream finally found out about large language models. The conversation about “AI in your tool” suddenly had a generic, easy reference point. Every SaaS product manager in the WordPress ecosystem started writing memos.

2023 — The catalyst year. GPT-4 launched in March 2023. The quality jump was significant enough that “use AI inside your product” went from speculative to obvious. The WordPress AI conversation started here:

  • April 2023: Automattic shipped Jetpack AI Assistant, the first major WordPress AI feature from a top-three WordPress company.
  • May 2023: Elementor AI launched, becoming the first AI feature shipped by a major page builder. Content and image generation only — credit-metered against Elementor Pro.
  • September 2023: Divi AI launched, similar feature set to Elementor’s.
  • Throughout 2023: AI Engine by Meow Apps saw explosive growth as the general-purpose WordPress AI plugin. JetPack AI Assistant added new modes. Third-party Elementor and Divi addon developers started shipping AI-powered widgets.
  • October 2023: Vercel’s v0.dev launched. Outside WordPress, but it set the bar for what AI-native design tooling looked like.

By the end of 2023, the WordPress builder market had broadly accepted that AI features were required to compete. But the implementations were all the same: content overlays. Fill this empty headline. Generate this image. Write this product description. The architectural conversation that’s defining 2026 — generation vs HTML→Builder vs agentic — hadn’t started yet.

2024 — The content-overlay era. AI in WordPress meant AI for content, full stop. Elementor AI had launched in mid-2023 and matured through 2024 as a credit-metered content and image generator. Divi AI shipped soon after. Jetpack AI Assistant arrived for the Gutenberg crowd. AI Engine by Meow Apps was the most ambitious general-purpose option, but even it was primarily about chatbots and content rewriting. Every AI feature in WordPress was an overlay on existing builders: fill this empty headline, generate this product description, write this blog post. Nobody was generating layouts.

Outside WordPress, the picture was different. Vercel’s v0.dev launched in October 2023 and showed what AI-native design tools looked like — type a prompt, get a working React component. Claude 3 arrived in March 2024 with a real jump in code and HTML quality. By mid-2024, AI-curious WordPress users were screen-recording v0 outputs and manually rebuilding them inside their builders. The workflow that would eventually become “paste HTML, get builder elements” existed only as developer pain.

2025 — The shift year. Three things happened in parallel.

First, frontier models got dramatically better at HTML. Claude 3.5 Sonnet (June 2024) and 3.7 Sonnet (late 2024) made AI-generated HTML production-quality for the first time. Designers stopped using AI as a starting point and started using it as a finishing tool — output was usable as-is.

Second, the AI-first design tool category exploded. v0 matured. Lovable launched (Anthropic-backed) and produced full applications from a prompt. Bolt.new followed. Claude Design appeared. Each of these tools generated clean HTML and Tailwind, and none of them spoke any page builder’s native format. The “design somewhere else, paste into WordPress” workflow became a real and recurring problem that nobody’s builder handled well.

Third, Anthropic released Model Context Protocol in November 2024 and quietly transformed what AI agents could do. By mid-2025, MCP servers were starting to appear in dev tooling. The WordPress core team started watching.

The Bricks community read the signals first. Code2Bricks emerged as a free HTML→Bricks converter on the web. BricksPilot launched as a paid plugin. Bricksfusion Studio bolted AI generation onto its component library. By the end of 2025, Bricks had the richest third-party AI ecosystem of any WordPress page builder, despite Bricks itself having no first-party AI. The pattern was clear: builders that treated AI as an external workflow and built strong receivers outperformed builders that tried to own the generation layer.

Elementor took a different bet — keep extending Elementor AI, keep iterating on container-layout generation, and start exploring agents. Divi doubled down on native generation, betting that the full-site prompt-to-website experience would be the winning UX. Both bets are still playing out.

Early 2026 — The infrastructure year. Four months ago, this article would have been impossible to write because the underlying primitives didn’t exist yet. The list of things that have shipped just in February and March 2026 is striking:

  • February 5: WordPress.com launches the first official Claude Connector, exposing 60+ abilities to AI agents
  • February 17: WordPress.com rolls out the broader WordPress AI Assistant in the dashboard
  • February 21: Builderius integrates Claude with stateful project-aware editing (the AI reads your GraphQL schema before acting)
  • February 26: Divi 5 launches with a multi-agent full-site generation flow
  • February (mid-month): WordPress Core AI team ships the official MCP Adapter for WordPress
  • March 18: Bricks 2.3 ships HTML & CSS to Bricks as the headline feature — the moment HTML→Builder became a canonical architecture
  • March 20: WordPress.com extends MCP from read-only to full write access. 43% of the web becomes natively agent-writable.
  • March (later): Elementor launches Angie, the first major builder-native agentic AI plugin

By May 2026, the WordPress AI conversation has shifted from “does your builder have AI” (a 2024 question) and “what does your builder’s AI generate” (a 2025 question) to “how does your builder integrate with the agent layer, the model providers, and the WordPress core AI primitives” (the 2026 question). The third-party ecosystem has caught up to and in many cases surpassed the builders’ first-party offerings. The conversation has moved up a level of the stack.

The rest of this article is a snapshot of what that looks like, builder by builder, plugin by plugin, plus a forecast of where it goes next.

Why AI in page builders is harder than it looks

Claude, GPT, Gemini — all exceptional at HTML and Tailwind, because the open web is HTML and Tailwind. “Describe a page → output clean HTML” is one of the things modern models do best on a cold start.

Page builders don’t render HTML. They render proprietary JSON. Elementor has its format. Bricks has its format. Divi has its format. YOOtheme Pro has its format. Each is a hierarchy of named elements with named properties, and each is mostly absent from any model’s training data.

Ask AI to “build a page in your builder’s native format” and you’re asking a fluent English speaker to write in a language with twelve known examples in the world. The output looks plausible. It breaks the moment it touches the builder’s importer.

There are two ways out, and the market has split cleanly between them.

Two architectures: native generation vs HTML→Builder

Native generation. The builder ships its own AI that generates layouts directly in its JSON format. Fine-tuned models, careful prompts, constrained schemas. Elementor AI and Divi AI both work this way.

The upside is a unified workflow inside the builder. The downside is the format problem — quality is bounded by how much you can teach a model your specific JSON vocabulary, and output skews toward templates the model has seen most.

HTML→Builder. The builder accepts HTML/CSS as input and converts it deterministically into native elements. The AI’s job is to produce HTML (which it’s exceptional at). The builder’s job is to translate.

The upside is that the source can be anything — Claude, v0, Lovable, Figma exports, hand-written code. Quality tracks the AI industry’s HTML quality, which compounds. The downside is the two-step workflow and the engineering cost of a robust converter.

Bricks went all-in on HTML→Builder in version 2.3 (March 2026). The richest third-party AI ecosystems across other builders converged on the same pattern. Elementor and Divi are still primarily in the native-generation camp, though Elementor’s new agentic plugin Angie is something different — more on that below.

Where each builder stands as of May 2026

Elementor + Elementor AI + Angie

The largest builder by install count (13M+). Elementor AI ships text, image, container layout, custom code, and image-to-layout generation, all credit-metered against your Elementor Pro subscription (text = 1 credit, image = 33, container = 40, with annual credit pools at Starter/Power/Visionary tiers). The standalone Elementor AI add-on runs $35.88/yr for unlimited features on top of Pro.

The 2026 development that matters is Angie, Elementor’s free agentic AI plugin built on Model Context Protocol. Angie reads your site’s plugins, settings, and database, then writes, tests, and deploys widgets, snippets, and CSS in a sandbox before pushing live. It’s the first credible agentic experience inside a major page builder — and Elementor releasing it as a free standalone plugin (rather than rolling it into Pro) signals they see the agent layer as strategically important enough to subsidize.

What Elementor doesn’t have: a native paste-HTML-and-get-Elementor-widgets pipeline. This is the most exploitable gap in their product right now, and the third-party market has filled it.

Web2Elementor (web2elementor.com) is the most polished of the gap-fillers — free tier at 1 use every two months, Basic at €9/mo for 50 conversions, supporting URL→Elementor, image→Elementor, and HTML→Elementor through a shared credit pool, plus a Chrome extension for capturing live pages. Quality is acceptable for landing pages; complex interactive components require manual cleanup. AI to Elementor (aitoelementor.com) positions specifically around converting v0.dev, Lovable, and Bolt output into Elementor.

On the agent side, the open-source Elementor MCP plugin (github.com/msrbuilds/elementor-mcp) exposes Elementor as 97 MCP tools — widget listing, page structure inspection, element settings read/write, template browsing, global design tokens, custom code, all addressable from Claude Desktop, Cursor, or any MCP client. It connects via WordPress Application Password over HTTP — no proxy, no Node.js. This is the kind of community infrastructure that arrives before the builder’s first-party version and shapes what the official version eventually looks like.

The rest of Elementor’s third-party AI scene is content-generation widgets — AI Content Generator for Elementor (free, BYOK OpenAI), XPro’s AI ChatGPT widget, LiveMesh’s AI Assistant for Elementor on CodeCanyon. Useful for filling copy, not for generating layouts.

Bricks Builder 2.3

Bricks shipped version 2.3 on March 18, 2026 (current 2.3.5 as of late April). The headline feature is HTML & CSS to Bricks: paste HTML/CSS from any source into the builder, and it converts into native Bricks elements, classes, and CSS variables. Standard HTML tags map to their Bricks counterparts. CSS margin, padding, width, and similar get mapped to Bricks UI settings where possible; unmappable styles preserve as custom CSS. A builder setting toggles between confirmation-on-paste and auto-convert.

Bricks has no first-party generative AI. The position is explicit: bring your AI from anywhere, paste, convert. The third-party ecosystem fills the generation side — and despite Elementor having ten times Bricks’ install base, the Bricks third-party AI ecosystem is structurally healthier and more competitive.

BricksPilot (brickspilot.com) is the most comprehensive paid option: $25/mo Pro (70 generations) or $50/mo Business (200 generations), with BYOK at $10/mo unlimited or $99 lifetime. Hosted plans run Claude Opus 4.6; BYOK supports Claude, OpenAI, Gemini, Grok, DeepSeek, and OpenRouter. Does both generation and HTML/CSS/JS conversion, pulls in your existing CSS variables so output matches your design system, supports BEM naming, and tracks token usage transparently.

Bricksfusion Studio AI (bricksfusion.com) takes a different angle — it’s primarily a component library (370+ wireframe sections) with an AI section generator bolted on, but uniquely auto-syncs generated elements with your Style Manager (colors, spacing, typography, global classes). The framework conversion is rare: it remaps sections between ACSS, Core Framework, AT Framework, Bricks Builder Framework, and Vanilla on the fly. €119/yr Essential, €279 lifetime Pro.

Bricks AI Studio (bricksaistudio.com) is a newer one-time-purchase entrant. Supports four AI providers (Claude, GPT-4o, Gemini 2.5 Pro/Flash, Grok 3), with two operating modes (managed credits or BYOK with zero markup). Generates sections from text prompts, converts HTML/CSS/JS, recreates designs from screenshots.

Code2Bricks (code2bricks.com) is the developer-leaning option — free web tool at code2bricks.sevalla.page, with a paid native plugin offering bidirectional editing (edits flow both ways between code editor and Bricks elements). No AI generation, conversion only, but the output is pure native Bricks with no frontend CSS/JS injection.

The free-only converters are worth knowing: Brickify (brickify.netlify.app — single-developer side project), bricksaiconverter.com (AutomaticCSS integration), and Move to Bricks (movetobricks.com — Elementor→Bricks migration plus HTML import). And ClonewebX (softlite.io/clonewebx) deserves a special mention as a cross-builder Chrome extension that captures any URL and exports to Webflow, Elementor, Gutenberg, Bricks, Breakdance, or Divi 5 — the only tool in this entire roundup that operates across builder lines.

The pattern across the Bricks ecosystem: BYOK is the default, multi-provider is the default, transparent token economics are the default. The Bricks community responded to Bricks 2.3 by building exactly the tooling the architecture invites.

Divi 5 + Divi AI

Divi 5 released February 26, 2026 — the biggest version bump in Divi’s history. Divi AI now includes a multi-agent flow that generates an entire site in 1–2 minutes: outlines pages, builds templates, fills content, generates images, picks fonts and colors. Beyond the full-site flow, you get per-module/section generation, AI-aware custom code, and AI text fields embedded throughout the builder.

Pricing is bundled into Divi subscriptions, with Divi AI as an add-on (unlimited usage on the top tier). No HTML import — Divi AI generates and renders within Divi’s own format end to end.

Divi is the strongest expression of the native-generation thesis. If you want “type a prompt, get a working website in your builder,” Divi 5 ships that experience further than anyone else right now. The tradeoff is the closed loop — what comes out is Divi format, and what goes in is text prompts, not external designs. If a better model arrives next quarter, Divi AI improves only as fast as Elegant Themes integrates it.

YOOtheme Pro + YOOforged

YOOtheme Pro is the design-quality leader among WordPress builders. Pro 5.0 shipped January 2026 with cookie consent management, lazy-load video, date filters, and language switcher updates. YOOtheme has signaled on social media that AI is coming to the core builder, though as of May 2026 nothing has shipped or appeared in the official changelog.

In the YOOtheme ecosystem today, AI ships through third-party plugins. The primary one is YOOforged, which I build. YOOforged was designed around the HTML→Builder architecture from the start — it accepts HTML from any source (Claude Design, v0, Lovable, Figma exports, hand-coded) and converts to YOOtheme Pro JSON. Beyond layout conversion and prompt-based generation, it ships translate (full-site translation in one pass), copywriter (refill/refresh content with brand-aware rewriting), and style modes.

YOOtheme’s smaller install base means a thinner third-party scene compared to Bricks or Elementor, but the design quality of YOOtheme Pro out of the box partially compensates — AI output looks polished by default because the theme itself enforces good design.

Beaver Builder + BeaverForge

No first-party AI. The Beaver team has been publicly skeptical of the AI hype cycle, and the product reflects that. The relevant 2026 development is BeaverForge AI (beaverforge.ai), a third-party plugin doing prompt-to-layout, post/page generation, design tokens, and Beaver Themer templates. BeaverForge runs BYOK — buy the plugin once, connect your own Claude, OpenAI, or Gemini key.

Worth noting the name parallel with YOOforged for transparency — “Forge” is a common dev-tooling pattern (SourceForge, GitHub Forge), and both YOOforged and BeaverForge land at the same metaphor independently. The products are doing similar things for different builders.

Oxygen + Breakdance

Oxygen and Breakdance now share the same core builder engine (the Oxygen 6 / Breakdance Studio era). No first-party AI from Soflyy. The third-party option is OxyDance Pilot (oxydancepilot.com), which generates native Breakdance/Oxygen sections from text prompts, design images, or HTML snippets, plus an “Edit with AI” context-menu action. BYOK across Claude, OpenAI, Gemini, DeepSeek, Grok, and OpenRouter.

Gutenberg — the AI ecosystem you might not have noticed

Every WordPress site has Gutenberg, the core block editor. It’s the default content layer underneath every page builder, and it has its own rapidly expanding AI ecosystem that often gets ignored in “page builder” conversations.

Spectra AI (wpspectra.com), formerly Ultimate Addons for Gutenberg, is the largest single Gutenberg AI plugin with over 1 million active installs. Free tier ships 30+ blocks and 100+ templates; Spectra Pro starts at $59–69/yr. AI features include content generation, persuasive copy generation, dynamic content, and AI-powered design tools in Pro. Performance-focused codebase — modular loading keeps it lightweight relative to the breadth of features.

AI Engine (Jordy Meow / Meow Apps) is the dominant general-purpose WordPress AI plugin and the most architecturally interesting of the lot. Supports OpenAI, Anthropic, Google, OpenRouter, Mistral, and more. Capabilities span ChatGPT-style chatbots, an AI Copilot in the editor for real-time rewrites/translations/improvements, voice + memory chatbots, AI forms, custom model training via embeddings, and — relevant to this article — an MCP server exposing posts/media/SEO/social management tools to AI agents. Token tracking and cost analytics built in. Free + Pro tiers. A 2026 security vulnerability affected approximately 100,000 sites running AI Engine (more on this below); the issue has been patched, but it’s the first significant security incident in this corner of the ecosystem.

Kadence AI is bundled into the Kadence Theme ecosystem rather than sold standalone. Pricing tiers from Express $69/yr to Ultimate Bundle Lifetime $899–999, with AI credit allocations from 250/yr (free) to 8,000/yr (Ultimate). One credit = one inline text edit; 70 credits = full starter site generation. Strong for content and site generation, weaker on AI editing inside individual blocks.

GenerateBlocks Pro is performance-focused and notably lacks a dedicated native AI generator as of May 2026 — pairs with the separate “AI Editor” plugin that adds AI-generated blocks. This is a competitive gap. Stackable ($49/yr to $499 lifetime, 47 custom blocks) has a similar AI gap.

The HTML-to-Gutenberg converter market is fragmented and competitive:

  • Gutenizer (gutenizer.com) — AI-powered Figma/HTML/v0/Lovable/Bolt to Gutenberg blocks with semantic recognition (hero/features/testimonials/CTA detection)
  • Parzero (parzero.ai) — converts any HTML template to a full WordPress theme with modular Gutenberg blocks
  • Telex — official WordPress AI-powered prompt-to-Gutenberg-blocks tool
  • html-to-gutenberg.com — developer-focused CLI/webpack for HTML→React+PHP/Twig
  • WP Block Converter (alleyinteractive on GitHub) — open-source PHP composer package
  • Block Format Bridge — lets AI output Markdown/HTML and converts server-side

No single tool dominates. The Gutenberg HTML-converter story is what Bricks 2.3 made native: six tools all chasing the same problem, none yet integrated into core.

And then there’s WordPress.com / Jetpack AI. WordPress.com launched the first official Claude Connector on February 5, 2026 — the first major host to ship a first-party Claude connector. Exposes 60+ abilities via OAuth 2.1, every change confirmed by user. Available on all paid WordPress.com plans plus any Jetpack-AI-or-Complete-connected site. Powered by WordPress.com’s own MCP server. On March 20, 2026, this was extended from read-only to full write access — draft posts, publish pages, moderate comments, bulk-update media metadata, with explicit confirmation per change and posts defaulting to draft. WordPress.com’s bet is that the agent layer is the platform layer, and self-hosted WordPress is going to need to catch up.

The MCP shift — and what WordPress 7.0 actually does

The most underrated 2026 development isn’t inside any individual builder. It’s happening at the WordPress core layer.

In February 2026, the WordPress Core AI team shipped the official MCP Adapter, formalizing the path from WordPress Abilities → MCP tools → AI agents. This isn’t a separate plugin layered on top; it’s becoming part of WordPress itself.

WordPress 7.0 — currently rescheduled (originally April 9, then May 20, then pushed back again on March 31 due to Real-Time Collaboration performance issues) — ships three things in core that matter for AI:

  1. The Abilities API — a standardized way for any plugin to expose its capabilities to AI agents
  2. The WP AI Client (php-ai-client) — a unified PHP interface for talking to OpenAI, Anthropic, Google, and others, so plugins don’t each ship their own provider integrations
  3. The Connectors API — the OAuth-based mechanism for AI agents to authenticate against a WordPress site

The implication: every plugin becomes a discoverable AI-agent surface. The official AI provider plugins (OpenAI, Google, Anthropic) are already in the Plugin Directory, ready for 7.0 launch.

Beyond core, the third-party MCP infrastructure has matured fast. Respira for WordPress (respira.press) ships 172 MCP tools across 12 builders — full intelligence support for Elementor, Divi, Gutenberg, Bricks, Flatsome, and Oxygen, with smart defaults for five more. The differentiating feature is a duplicate-before-edit safety pattern: nothing breaks because changes go to clones first, then promote on user confirmation. Pricing starts at €9/mo for one site. This is genuinely category-defining infrastructure that’s flying under most of the WordPress-AI coverage.

Builderius integrated Claude on February 21, 2026, and it’s the most architecturally novel piece of news in this space. Builderius’s AI reads the live project state before acting — element tree, GraphQL schema, resolved data variable values. It reads every CSS class/variable/rule before writing styles, and reuses existing classes rather than spawning new ones. New CSS goes into reusable template-level classes. It’s the only builder doing true stateful AI editing as of May 2026, where the AI knows what your site already has before it changes it.

On March 20, WordPress.com’s extension of MCP to full write access is, by some measure, the single most consequential ecosystem event of 2026 so far. When 43% of the web becomes natively agent-writable, the page-builder AI conversation shifts from “what can the builder do” to “what does the AI agent decide to do with WordPress as a target.”

The practical implication for any tool you use: ask whether it exposes an MCP server (first-party or community-maintained). The ones that do inherit the entire frontier model ecosystem. The ones that don’t will need to ship every AI feature themselves, indefinitely.

Cross-builder and site-generation tools

Beyond the per-builder ecosystems, a horizontal tier has emerged.

WDesignKit (wdesignkit.com) supports Elementor, Gutenberg, and Bricks from a single design kit. AI License Key support, AI-Powered Kit Import, AI-Generated Text & Descriptions, AI Image Library all shipping. Coming: “AI to Widget Builder” (describe and get a working widget for all three builders) and AI Code Snippets.

10Web (10web.io) generates full WordPress sites with an Elementor-based drag-and-drop editor underneath. $10/mo annual pricing, 7-day Pro trial, initial generation free. March 2026 launched the Agentic Website Builder with ten specialized agents (PM, Designer, Dev Team, QA, Content, Visual, DevOps, SEO, Analytics, WooCommerce). Recreate-from-URL is a notable competitive lever. Hosting on Google Cloud with PageSpeed Booster targeting 90+ scores.

Hostinger AI Website Builder for WordPress sits at the other end of the price curve — $2.99–3.99/mo, two tiers (Premium = 1 site, Business = 50 sites with AI image gen and AI blog gen). 45–60 second full-site generation, Gutenberg block editor underneath, auto-generated title tags/meta/H-hierarchy. Cheapest entry into AI WordPress site generation in this entire roundup.

ZipWP (zipwp.com) plans the site first — AI generates sitemaps, wireframes, and layouts before any building happens. Free tier available. The “plan, then build” workflow is rare and underrated.

AI across the rest of WordPress

Page builders are the headline category, but the WordPress AI story extends across several adjacent verticals that any site owner will eventually touch. Three of them — e-commerce, SEO, and translation — have their own mature AI ecosystems in 2026 that influence what builders need to integrate with.

AI in WooCommerce

WooCommerce powers approximately 30% of all online stores worldwide. Its AI ecosystem has matured fast in 2026, though notably without a Shopify-Sidekick equivalent shipping inside WooCommerce core. Automattic’s strategy is platform-first: WooCommerce 7 (released April 9, 2026) added Model Context Protocol support, deeper Analytics hooks, and broader store-data exposure so third-party agents can plug in. Merchants can already point ChatGPT Connectors or Claude Desktop at their store today, but no first-party conversational assistant ships in the Woo admin yet. That absence is, itself, the story.

The third-party Woo AI scene fills the gap. Klaviyo is the most aggressive: their Spring 2026 release added a Customer Agent for email and WhatsApp, Audience Optimization (predictive unsubscribe scoring that auto-prunes high-risk profiles), and brought Customer Hub to WooCommerce in Q1 2026. Pricing is volume-based on contacts, starting around $20/mo. AI product description plugins are now a crowded mature category — WriteText.ai, AI Product Tools (with multi-provider routing across OpenAI, Gemini, Claude, OpenRouter), Describely, ContentGen, and StoreAgent all compete on bulk descriptions, titles, tags, FAQs, and meta — typically a small monthly fee plus your own API key.

AI chatbots for WooCommerce are split between general-purpose tools and Woo-native challengers. Tidio’s Lyro AI is the most visible but doesn’t natively read live Woo product/stock data (you build flows). Woo-native challengers like StoreAgent, Amio, and MxChat are positioning specifically on “knows your catalog out of the box.” AI recommendation engines have largely been absorbed into chat/cart-recovery agents — recommendation as static sidebar widget is giving way to recommendation in conversation.

AI product image generation is shipping in the Woo marketplace as first-class. AI Product Image Generator and AI Product Image Builder lead the category, with the latter routing across OpenAI’s GPT Image 1.5 and Google’s Imagen 4 with bulk CSV upload. ContentGen and Product Photo AI round out the list — all BYO-API-key. Abandoned cart AI moved in 2026 from email-only post-abandonment to in-session conversational recovery, with vendor benchmarks claiming AI-assisted flows recover 15–30% of carts vs Omnisend’s 7% rule-based baseline (treat with appropriate skepticism — these are vendor-published numbers).

The headline takeaway for the WooCommerce slice of the WordPress AI story: WooCommerce 7’s MCP-readiness is the most important infrastructure shift of 2026 for e-commerce site owners, and the third-party AI ecosystem is healthier than the Woo core suggests. The gap with Shopify Sidekick is real but narrowing through plugins.

AI in SEO plugins

Every WordPress site touches an SEO plugin. The AI features in those plugins have shifted from “nice to have” in 2024 to “the main reason to upgrade” in 2026.

Yoast SEO has shipped the most strategically interesting moves of any SEO plugin in 2026. Yoast Premium contains AI Generate (titles, meta descriptions, summaries for posts/pages/products/taxonomies) and AI Optimize (suggested changes against the SEO analysis). The bigger 2026 play is Yoast SEO AI+, launched April 7, 2026, which tracks how your brand appears in ChatGPT, Gemini, and Perplexity answers — i.e., generative engine optimization (GEO/AEO), not classic SEO. Available to Premium and WooCommerce SEO customers with no usage cap. Premium runs approximately $99/yr per site.

RankMath Content AI migrated off credits to feature-based monthly limits in 2025. Three tiers (Starter $6.99/mo, Creator $11.99/mo, Expert $18.99/mo on annual billing) include 40+ tools, AI Link Genius for internal-link suggestions, RankBot assistant, and one-click article generation. Unlimited seats differentiate it from Yoast and Surfer.

AIOSEO now ships bulk AI generation of meta titles, descriptions, and alt text; an AI Schema Generator; an AI Assistant block for inline drafting; an AI Image Generator; and an AI Writing Assistant via SEOBoost integration pulling competitor SERP data. Pricing from approximately $50 to $400/yr.

SEOPress AI is multi-provider by design — generates titles, descriptions, OG tags, and alt text via OpenAI, Gemini, Claude, Mistral, or DeepSeek. The standout 2026 feature is “Agent Readiness”: one toggle exposes plain-text content, an llms.txt file, and structured signals for ChatGPT/Claude/Perplexity/Gemini crawlers. Used by 350K+ sites.

Squirrly SEO leans into AI with three proprietary models: Ranking Vision AI, Keyword Research AI, and ASK AI. Positioned as an AI SEO consultant generating custom to-do lists rather than just text.

The category-level story in 2026 SEO plugins is GEO/AEO — optimizing for how LLMs answer queries, not just Google rank. Yoast AI+ and SEOPress’s Agent Readiness are the clearest expressions. Expect AIOSEO and RankMath to follow within a release or two.

AI in translation plugins

Multilingual WordPress is a huge market, and translation plugins have been quietly using AI since well before the LLM era (Google Translate and DeepL underneath since the late 2010s). The 2024–2026 shift was adding LLM-based translation on top of the existing NMT layer for marketing-tier output.

WPML is still dominant. Multilingual CMS $99/yr (90K credits), Multilingual Agency $199/yr (180K credits). Four engines under a unified credit system: Microsoft (1 credit/word), Google or DeepL (2 credits/word), and WPML’s own AI / Professional Translation Cloud (up to 4 credits/word). DeepL is the most-used engine in their ecosystem — WPML reports 13B+ words translated via DeepL in 2025 alone.

TranslatePress AI is explicitly hybrid by design — combines LLMs (Gemini, GPT) with NMT engines (DeepL, Google) and routes per language pair and context. Detects new content automatically including SEO metadata. Free tier 2 languages and 2K AI words; Personal ~€89/yr; Business ~€169 for 3 sites; Developer €349. Supports 132+ languages.

Linguise is AI-native SaaS that bolts onto WordPress (and Shopify, etc.). Pricing from $15/mo (20K words) up to $159/mo (1M words) with unlimited languages on all tiers. Marketed against Weglot as more affordable for high-volume sites.

Weglot uses a mix of DeepL, Google, and Microsoft under the hood plus an “AI translation” tier that markets LLM-quality output for marketing copy. Pricing scales by word volume, language count, and pageviews — fastest cost growth among the four for large multilingual sites.

Polylang + Polylang Pro ships DeepL integration and XLIFF export but no native LLM features. The AI layer is bolted on via separate plugins — AutoPoly and AI Translator for Polylang — popular with developers who want Polylang’s free-tier foundation plus à-la-carte AI.

DeepL underneath everything. DeepL Pro remains the de facto NMT layer across WPML, TranslatePress, Polylang Pro, and Weglot. DeepL Pro Starter approximately $8.74/mo, Advanced $28.74/mo, Ultimate $57.49/mo per user; API pricing at $25/mo + ~$25 per 1M characters. The hybrid pattern most vendors recommend in 2026: DeepL for first-pass NMT, LLM for brand-voice proofread and marketing-tier output.

YOOforged’s translate mode sits in this category as a YOOtheme-specific option — full-site translation in a single pass with brand-aware copy adjustments, rather than the per-string flow most plugins use. Different use case; same broader trend.

Practical risks worth knowing about

Two things in the May 2026 landscape that deserve mention before you ship AI features into a client site.

AI Engine vulnerability. A flaw in AI Engine — the most widely installed general-purpose WordPress AI plugin — exposed approximately 100,000 sites to attack. The vulnerability has been patched, but it’s the first significant security incident in WordPress AI plugins, and it’s a useful reminder that AI plugins frequently hold API keys, manage tokens, and execute code in privileged contexts. Audit AI plugins the way you’d audit security plugins. Keep them updated.

Hosts blocking AI bots. Managed WordPress hosts have started silently blocking AI bot traffic at the infrastructure layer, breaking MCP and connector workflows in ways customers can’t see. Before relying on any agentic feature, confirm with your host whether AI bot traffic is allowed. Search Engine Land flagged this as a growing pattern in 2026. Hosts have legitimate reasons (bandwidth, compute costs), but the friction means an Angie or Claude Desktop workflow that works on one host may simply fail on another.

Which AI models actually power what

Most articles in this space treat “AI” as a single black box. It isn’t. Different tasks reward different models, and the WordPress AI plugins that ship multi-provider routing (BricksPilot, Bricks AI Studio, SEOPress AI, ContentGen, AI Product Tools) do so because their authors have learned the same lesson the broader industry has: the right model for layout generation isn’t necessarily the right model for product descriptions, and neither is the right model for translation.

A quick frontier-model state as of May 2026:

  • Claude Opus 4.7 (released April 16, 2026 — $15/$75 per million tokens, 87.6% SWE-bench Verified) is the current quality leader for structural reasoning, complex code/HTML output, and tasks that require following detailed schemas without drift.
  • Claude Sonnet 4.6 ($3/$15 per million tokens, 79.6% SWE-bench) is the cost-sensible workhorse tier. Most production WordPress AI pipelines that need Claude-class quality at volume use Sonnet, not Opus.
  • GPT-5-class (OpenAI’s current flagship at $1.75/$14 per million tokens, with version naming somewhat fluid across third-party trackers) leads on price-performance for high-volume content generation.
  • Gemini 2.5 Pro / 3.1 Pro (Google, $2/$12 to 200K tokens, $4/$18 above; 2M-token context, native video) is the choice when you need to feed massive context — a full design system, an entire codebase, hours of source material.

What each task actually rewards:

Layout generation (HTML → builder). This is the task that rewards structural reasoning and clean JSON/HTML output at scale. Claude Opus 4.7 and Sonnet 4.6 are the de facto choice for serious builder-targeted layout work — the SWE-bench gains transfer directly to layout schema fidelity. Gemini 2.5 Pro is the alternative when 2M-token context lets you stuff a full design system into the prompt. GPT-5-class is competitive on price and shows up widely in cheaper tiers.

Content and copywriting. WordPress AI plugins explicitly multi-route here (SEOPress, AI Product Tools, ContentGen expose OpenAI/Claude/Gemini side-by-side). The de facto split: GPT-5-class for default volume content because of price-performance, Claude Opus 4.7 or Sonnet 4.6 when brand voice and nuance matter, Gemini when content needs to be grounded in long uploaded source material.

Translation. DeepL still benchmarks best for European pairs at the NMT layer and is the underlying engine for WPML, TranslatePress, Polylang Pro, and Weglot. For brand-voice marketing translation and Asian languages where idiomatic nuance matters, Claude (Opus/Sonnet) and GPT-5 consistently rate higher than DeepL on idiom, register, and keigo-class problems. The 2026 production pattern is hybrid: NMT for first-pass, LLM for proofread/marketing-tier output. TranslatePress AI’s whole pitch is automating exactly that routing.

Image generation. No single winner; routing is the norm. OpenAI’s GPT Image 1.5 leads on instruction-following and is what most Woo bulk-image plugins ship with. Google’s Imagen 4 leads on photorealism and is the engine behind Nano Banana routing in WooCommerce image plugins. Flux 2 Pro (Black Forest Labs) is the API-first choice for indistinguishable-from-photo realism. Midjourney v8 holds the aesthetic crown but the lack of a serious API keeps it out of most WP plugins.

A note on responseSchema and the validation tradeoff

One technique deserves explicit mention because it’s central to how serious AI-in-WordPress engineering works and because misunderstanding it produces most of the brittle, broken-on-import AI features you encounter in the wild.

Gemini’s responseSchema parameter — and its equivalents on the OpenAI stack (structured outputs) and Anthropic stack (tool-use schemas) — lets developers pass a JSON Schema that constrains the model’s output. The model biases token generation so the output conforms to that schema. The appeal is obvious: parsing arbitrary LLM output is brittle, and a model that occasionally returns fenced code blocks, trailing commentary, or hallucinated keys breaks any downstream pipeline expecting deterministic JSON. Constrained decoding eliminates that failure mode entirely.

The tradeoff is real and worth naming explicitly: structured output guarantees syntactic validity but not semantic correctness, and overly strict schemas measurably degrade reasoning quality. Google’s own data shows JSON-Schema-constrained Gemini scoring 86% on a shuffled-objects reasoning task vs 97% for unconstrained natural language output. Complex schemas — long enums, deep nesting, many optional fields — compound this. Google warns explicitly that using structured output on tuned models can reduce quality.

The practical engineering rule that emerges across serious AI products: keep the schema lean enough that the model still “thinks” freely, then validate semantics in application code after parsing. The schema gives you parseable JSON, not correct JSON. The closer you push toward “schema enforces every detail,” the more the model burns capacity satisfying the schema rather than reasoning about the design. The closer you push toward “no constraints,” the more brittle imports you see.

Finding the right balance is most of the engineering work behind any AI page-builder feature. It’s the difference between “the AI generated something that imported cleanly and looks good” and “the AI generated something technically valid that crumbles in the builder.” Every team working on this — Bricks, Elementor, Divi, YOOforged, BricksPilot, all of them — is solving the same fundamental tension, with different answers that show up as different output quality. I’ll cover how this works in practice in a video walkthrough later this week — link will land here when it’s up.

How to evaluate AI in your page builder

The rubric that survives across builders and across model releases. Score whichever tool you use against these eight questions.

1. Does it accept HTML/CSS from external sources? The single most predictive question. AI is best at HTML. Tools that accept HTML scale with the AI industry; tools that don’t lag every six months as the frontier models improve.

2. How brittle is the import or generation? Run the same prompt three times. How many produce a usable layout vs broken DOM? Reliability beats peak quality. A brilliant output one in five times is worse than a competent output every time.

3. What sources does it import from? A converter that only handles its own AI’s output is a closed loop. A converter that handles Claude Design, v0, Lovable, Figma, and hand-coded HTML is a category-wide tool.

4. How does it handle validation? Too much schema constraint chokes output creativity. Too little produces invalid layouts that break on import. The right pattern is post-generation validation with smart auto-correction.

5. What’s the pricing model? Per-credit, per-generation, monthly subscription, or BYOK. BYOK gives direct cost control and frontier-model access. Bundled credit pools hide the economics and make per-generation costs hard to optimize. Across the third-party WordPress AI market in 2026, BYOK has won — every serious third-party tool (BricksPilot, BeaverForge, OxyDance Pilot, Bricks AI Studio, Builderius) offers a BYOK tier.

6. Does it cover more than layout? Layout-only AI is incomplete. Translation, copywriting, alt-text, content refresh, brand-aware rewriting — these save more cumulative time than another marginal layout improvement.

7. Is there an MCP server (or roadmap for one)? The agent layer is real and already shipping. Builders without an MCP surface will have to bolt every AI feature on as discrete product work. Builders with one inherit the entire frontier model ecosystem.

8. What’s the security posture? AI plugins hold API keys, manage tokens, execute privileged operations. Audit them. Update them. The AI Engine incident in 2026 is the new baseline expectation: plan for AI plugins to be a security surface, not just a feature surface.

Outside WordPress: the broader AI website landscape

WordPress is 43% of the web. The other 57% is doing AI website generation too — and most of what’s shaping the WordPress AI conversation in 2026 originated outside it. The AI-first website builders, the design tool category, the “feeder” tools that produce HTML for WordPress imports, the dev tooling that pressures everything — all of this is the context within which WordPress AI builders are being evaluated, whether the WordPress ecosystem acknowledges it or not.

Direct WordPress competitors with first-party AI

Webflow + Webflow AI is the most directly WordPress-competitive of the AI-first builders. The Webflow AI Assistant generates entire responsive sites from prompts, builds CMS collection items in bulk, and produces new sections that automatically match an existing site’s design system. In 2026 Webflow now markets the assistant as capable of generating and deploying full-stack web apps from natural-language prompts — directly competing on Lovable’s turf. Pricing runs $0 to $212+/mo (CMS at $23/mo, Business at $39/mo). The structural advantage over WordPress builders is that Webflow’s canvas is exact-CSS, designer-grade, with AI riding on top of a structured design system rather than bolting onto an opinionated theme.

Framer + Framer AI simplified to Free / Basic $10 / Pro $30 / Scale $100 tiers in early 2026. AI features include Wireframer (layout generation), Workshop (coding assistant), built-in translation, and live on-page editing in the browser. Distinctive vs WordPress: design-tool-first aesthetic with motion and interaction baked in at the primitive level, and a single ecosystem from prompt to publish — no plugin hunt.

Wix + Wix AI Website Builder. Wix retired the original ADI (Artificial Design Intelligence) on November 10, 2024 and replaced it with the conversational AI Website Builder, now branded under “Wix Harmony.” The platform exposes 20+ AI tools — site generation from a single prompt, AI customer chat, automated product descriptions, blog/email/Google Ads generators, and GPT-4o vision for image captioning. Wix has the most opinionated end-to-end AI flow on the market: site, content, marketing, and support all generated in one conversational thread. This is also the closest competitor to what WordPress.com is building with its Claude Connector and AI Assistant.

Squarespace + Blueprint AI. Blueprint AI Builder is free to use (publishing requires a $16–$99/mo annual plan) and walks users through a guided five-step flow producing custom layouts, fonts, palettes, AI copy, and AI imagery in 5–10 minutes. It sits under a broader “Design Intelligence” suite that includes Beacon AI (business assistant), Brand Identity tools, and the AIO Scanner. Distinctive vs WordPress: design taste curated tightly — the AI’s job is matching brand to a small set of expert-built layouts, not unbounded generation.

Shopify Sidekick + Magic. Commerce-focused but architecturally instructive. Sidekick is conversational; Magic is the embedded “Generate with…” buttons throughout the admin. The 2026 standouts: Brand Voice Cloning (trains on up to 1,000 past posts), Sidekick Pulse (proactively flags performance issues versus market trends), and auto-tagging from product images. Sidekick can also propose theme settings, palettes, and Shopify Flow automations directly in chat. The WordPress/WooCommerce equivalent doesn’t yet exist as a unified experience — and Shopify’s bet on “agentic commerce” is the model WooCommerce will eventually have to answer.

The feeders: AI-first design tools producing HTML for WordPress imports

These tools aren’t WordPress competitors so much as they’re the source of the HTML that ends up inside WordPress sites. Understanding them is essential because the HTML→Builder architecture only works if the upstream output is good — and these tools are where the upstream lives.

v0 by Vercel is the most-used feeder for WordPress builders in 2026. Five pricing tiers from Free ($5/mo credits) up to Enterprise, now using metered input/output token credits instead of fixed message counts. Models are proprietary v0 Mini/Pro/Max plus v0-1.5-lg for Pro/Team tiers, all fine-tuned for React and Tailwind. The February 2026 update added Git integration, an in-product code editor, database connectivity, and agentic workflows — repositioning v0 from a prototyping toy into a deployment-grade tool. The common WordPress workflow: prompt a marketing page, export the HTML/Tailwind, paste into a Gutenberg block, a Bricks Code element, or a custom Elementor widget via Web2Elementor.

Lovable is the breakout of this cycle and the single most important data point in this entire article. ARR crossed $400 million in February 2026 — having grown from $100M to $200M to $300M to $400M in successive months from mid-2025 — with just 146 employees. Series B of $330 million closed at a $6.6 billion valuation in December 2025 (led by CapitalG, with Menlo Anthology participating). It is the fastest software ARR ramp ever publicly documented. Lovable routes prompts to OpenAI and Anthropic models. Its Series B funds full-stack infrastructure — database, payments, hosting — so it can ship live apps end-to-end rather than just designs. WordPress users primarily use Lovable as a feeder: prompt a landing page, export HTML, hand-port into Elementor/Bricks/Breakdance containers.

Bolt.new (StackBlitz) hit $40M ARR by March 2025 (roughly five months from launch), 5M registered users by May 2025, and is reportedly profitable. Total funding sits at $135M; Forbes valued StackBlitz at ~$700M in August 2025. StackBlitz said it would announce a new milestone after crossing $100M ARR — that announcement has not landed publicly as of May 2026, which is itself notable. Bolt’s output skews more app-shaped than page-shaped, so it appears less in WordPress workflows than v0 or Lovable.

Claude Artifacts, Skills, and Claude Design. Anthropic’s product story here moved fast in 2026. Artifacts can now call Claude’s API directly, use persistent storage, and connect to external services via MCP — effectively a micro-app dev environment inside Claude. The Anthropic-published “frontend-design” Skill had 277,000+ installs by March 2026. The big April news: Claude Design launched April 17, 2026 as an Anthropic Labs research preview for Pro/Max/Team/Enterprise users. It generates prototypes, slides, and one-pagers, reads your codebase or design files to apply your existing design system, and is powered by Claude Opus 4.7. Figma stock dropped 7% the day it launched — three days after Anthropic CPO Mike Krieger stepped off Figma’s board. For WordPress users this matters in two ways: Claude Design is rapidly becoming a feeder for HTML imports, and the company that powers most WordPress AI plugins (Anthropic) is now also a direct design-tool company.

Figma Make. First released May 7, 2025; the May 2026 update added voice-to-text in chat, question cards (structured tradeoff choices mid-generation), full version history with one-click revert, “clear context” session reset, and Make kits/attachments — which seed generations with real components, data, and constraints rather than blank-slate prompts. Figma Make is the only feeder tool starting from a real design-system context window, which makes it disproportionately useful for teams that already have design systems in Figma.

Replit Agent 3 restructured to effort-based credits in early 2026: Free Starter, Core $20/mo, Pro $100/mo (for up to 15 builders, launched February 2026), plus Enterprise. Simple checkpoints typically under $0.25; complex bundled tasks cost proportionally more. Agent 3 adds extended thinking mode, high-power model toggle, and web-search-grounded documentation lookup. Less common as a WordPress feeder, more common as a parallel dev environment.

AI code tools that shape everything else

Cursor offers six plans from Free to Enterprise, with Pro+ ($60) and Ultra ($200) tiers providing 3× and 20× credits respectively. Agent mode supports multi-file edits and parallel background agents — the parallel workflow is the standout vs every other IDE in 2026.

Claude Code has become the de facto standard CLI and agent harness in the Anthropic ecosystem — Skills, Artifacts, and Claude Design all share its underlying lineage (Opus 4.7). It is the primary route by which Anthropic competes with Cursor in serious engineering workflows, and it’s increasingly used by WordPress developers building custom plugins and themes.

GitHub Copilot Workspace / Coding Agent added cloud-agent sessions launching from inside the IDE, custom user-level agents, and a Debugger agent that validates fixes against live runtime. The cloud coding agent runs in GitHub Actions, researches the repo, plans, codes, self-reviews via Copilot code review, and opens a PR.

Windsurf (formerly Codeium) was acquired by OpenAI in 2026 for approximately $3 billion — OpenAI’s largest acquisition to date. OpenAI took the user base plus Cascade (the agentic system) and SWE-1 (purpose-built coding model).

Devin (Cognition AI) was in talks to raise hundreds of millions at a $25 billion valuation as of late April 2026. Enterprise traction became serious in 2026 — Cognizant and Infosys both announced strategic partnerships to scale Devin across enterprise software engineering. The autonomous-PR-from-Linear/Jira workflow is what’s selling.

Adjacent AI-native categories

Worth knowing for context, even if not directly competing with WordPress builders:

Bubble remains the heaviest visual-programming no-code platform; 2026 updates added AI-generated workflows and screens but the full database/workflow editor stays exposed. Glide added image-to-text, audio-to-text, and in-app AI support agents in 2026 — still spreadsheet-data-first. Softr repositioned as “first AI-native no-code business-software platform” with an AI Co-Builder that generates database, app, and business logic from a description. Penpot (open-source Figma alternative) is free core with a $7/user/mo Unlimited tier; first-party AI is roadmap, not shipped.

The AI-native headless CMS category emerged in 2025–2026: Sanity leads with its Content Operations Agent (schema-aware research and generation), CrafterCMS positions as “content + agents + orchestration,” and newer entrants like NomaCMS and ElmapiCMS appeared in 2026. Market projections call this category $3.94B in 2026 growing to $22.28B by 2034 at over 21% CAGR — which is a hint at how much capital believes the CMS layer itself is about to be rebuilt around AI.

The funding context

The capital flowing into this space is the reason it’s moving this fast. OpenAI raised $122 billion on March 31, 2026 (Amazon $50B, Nvidia $30B, SoftBank $30B). Amazon added another $5 billion to Anthropic in late April, bringing total Anthropic commitment to over $33B. Lovable’s Series B was $330M at a $6.6B valuation. Cognition was raising at a $25B valuation. OpenAI bought Windsurf for $3B. These numbers shape what’s possible in the space: model costs keep falling, agentic features keep shipping, and the rate of new tool launches accelerates because the cost of building one keeps dropping.

WordPress sits inside this landscape, not outside it. The frontier models powering WordPress AI plugins are the same ones powering Cursor and Devin. The “feeder” tools that produce HTML for WordPress imports are sitting on top of the same Claude and GPT and Gemini APIs that Elementor AI and Divi AI run on. The agent layer being built into WordPress 7.0 uses the same MCP protocol that Claude Desktop and Cursor and Replit Agent already use. This is one market with one underlying substrate, segmented by interface.

What WordPress builders should be learning

A few patterns from outside WordPress that haven’t fully landed inside it yet:

Single-prompt-to-deployed-artifact latency is the user expectation now. Lovable, v0, and Claude Design have trained users to expect “describe it, see it running” in seconds. WordPress builders that require multiple steps (prompt → review JSON → import → check styling → fix issues) feel slow by comparison even when the output quality matches. Divi 5’s multi-agent flow is the closest WordPress equivalent and it’s still a generation behind.

Design-system-aware generation is the differentiator. Claude Design reads your codebase. Figma Make reads your real components. Bricksfusion Studio AI syncs with the Style Manager. These tools generate output that fits the existing system instead of fighting it. WordPress builders mostly still generate output that looks generic and then make the user style it.

Conversational iteration beats panel-based property editing. Wix and Shopify Sidekick let you say “make the hero darker and more aggressive” in chat. Most WordPress AI tools still expect you to open a properties panel and adjust values. The conversation is the editing surface.

The AI’s output should be the canonical artifact. In Lovable and Claude Design, what the AI produced is the project. In most WordPress builders, the AI’s output is a first draft that the human then immediately starts modifying via the traditional builder UI. Inverting this — making the AI’s output the canonical state, with the visual builder as the override — is the UX shift that’s coming.

The wedge for WordPress builders in 2026 isn’t “add AI to the builder.” It’s “make the AI’s output the canonical artifact and let humans nudge from there,” which is exactly what Lovable, v0, and Claude Design have demonstrated works at scale.

Why this matters: the stakes for WordPress builders

This is the section where I tell you why a category that didn’t exist three years ago now justifies a ten-thousand-word article. The short answer: WordPress’s position in the broader web infrastructure is genuinely threatened for the first time in over a decade, and the page builders are the layer where the fight gets decided.

The threat has three distinct angles, each of which would matter on its own. Together they reshape the calculus for every builder, every plugin author, and every site owner in the WordPress ecosystem.

Customer expectations have shifted permanently

The single most important data point in this article is Lovable’s $400 million ARR with 146 employees in February 2026. Not because Lovable is going to kill WordPress — it isn’t — but because that revenue ramp is evidence of demand that didn’t exist before. Hundreds of thousands of people who would have bought a WordPress site three years ago bought a Lovable subscription instead. Many of them didn’t even consider WordPress.

When a non-technical user wants a website in 2026, the question they ask is no longer “which page builder should I learn?” It’s “which AI tool can build this for me?” The first answer they hear is Lovable, v0, Claude Design, Webflow AI, Wix, Squarespace, or 10Web. WordPress shows up in that consideration set only when the user has a specific reason to want WordPress — content ownership, the plugin ecosystem, the brand, a developer’s recommendation. Increasingly, those reasons need to be defended explicitly, because the default expectation is now AI-first.

The site owners who already use WordPress are doing the comparison too. They open Claude or ChatGPT, prompt a page design, marvel at it, then realize they have to manually rebuild it inside Elementor or Bricks. Some of them keep doing that. Many of them quietly start migrating sites or new projects to Lovable, Webflow, or Framer. The migration isn’t dramatic and it doesn’t make headlines, but it’s continuous, and the larger the gap between AI-first tools and WordPress builders gets, the faster it accelerates.

The traditional WordPress moat is eroding

For fifteen years, WordPress has won the “I want to own my website” argument. Self-hosted, open-source, plugin ecosystem, no vendor lock-in, your content is yours. Those advantages are real and they still matter to a meaningful percentage of users. But they matter less than they used to, for two reasons.

First, the AI-first builders are converging on hybrid models. Webflow exports clean HTML/CSS. Framer offers code export. Lovable’s full-stack apps include a database and hosting you can theoretically self-host. The “I want to own my site” argument used to mean WordPress; now it means “you can leave most platforms if you have to.” The lock-in penalty for picking an AI-first builder is smaller than it was.

Second, the cost equation has flipped. Self-hosted WordPress used to be cheap (a domain + $5 hosting + a free theme). AI-first builders started expensive. In 2026, a Hostinger AI WordPress site is $2.99/month and a 10Web site is $10/month, while a serious WordPress build (managed hosting + premium theme + page builder + plugins + AI plugin) routinely exceeds $30/month before any custom work. The cost moat that benefited DIY WordPress is gone.

What remains is the ecosystem moat — 60,000+ plugins, the largest theme market in any CMS, deep WooCommerce integration, the developer talent pool. That moat is still substantial. But it’s not enough to overcome a 10× experience gap on the AI workflow. And the experience gap is currently 10× or more between Lovable and any WordPress page builder for the specific task of “generate me a working site from a prompt.”

The page builder is the layer that decides this

Here’s why this article focuses on page builders specifically: the page builder is the layer in the WordPress stack where the AI experience lives or dies.

WordPress core is doing its part. The MCP Adapter shipped. WordPress 7.0 brings the Abilities API, the AI Client, and the Connectors API into core. The Plugin Directory is filling with provider plugins for Claude, OpenAI, Gemini. The platform-level work is real and substantial.

But the platform doesn’t render pages. The builder does. When a non-technical user opens their WordPress site to make changes, they don’t see the Abilities API. They see Elementor’s panel, or Bricks’s structure tree, or Divi’s modules, or the Gutenberg block editor. Whatever AI experience exists in that interface is the AI experience of WordPress for that user. The MCP infrastructure underneath is invisible — what’s visible is whether the builder feels AI-native or not.

This is why every major builder’s 2026 decisions matter so much. Elementor shipping Angie. Bricks shipping HTML to Bricks. Divi shipping the multi-agent flow. YOOtheme signaling first-party AI. Beaver Builder staying skeptical. Each of these is a bet on what the builder’s UX should be in 2027. The builders that bet correctly survive and pull market share. The builders that bet wrong lose users to Lovable, Webflow, and the next AI-first competitor.

For site owners reading this: this isn’t an abstract market analysis. If you’re committed to a WordPress builder that doesn’t have a credible AI roadmap in 2026, you’ll be planning a migration in 2027. For plugin authors: the AI-first plugins shipping into established WordPress builders right now (YOOforged for YOOtheme, BricksPilot for Bricks, BeaverForge for Beaver, OxyDance Pilot for Oxygen/Breakdance, Web2Elementor for Elementor, the entire Gutenberg AI ecosystem) are the bridge the WordPress ecosystem is currently crossing. They’re filling the gap between what the builders ship natively and what users expect from outside-WordPress AI tools.

The opportunity inside the threat

The same dynamic that makes this an existential moment for WordPress builders also makes it the largest opportunity in the WordPress economy since the original page builder explosion of 2014–2016. WordPress has 43% of the web. The frontier AI tools have, collectively, low single-digit market share. If WordPress builders ship competitive AI experiences in the next eighteen months, they win because they have install base, distribution, and ecosystem the AI-first tools don’t have.

The “AI-first WordPress builder” position is open for whoever takes it. WordPress.com is moving fast on the platform side (Claude Connector, AI Assistant, MCP write access). Bricks took the architectural bet on HTML→Builder. Elementor took the agentic bet with Angie. YOOtheme is signaling. Beaver and Divi are still defining their plays. Whoever closes the gap first captures a generational shift in market share.

For users, the practical takeaway is simpler: ask hard questions about your builder’s AI roadmap. Not “do you have AI features” (everyone says yes). But “do you have HTML import, do you have an MCP server, do you have an agent layer, do you have a BYOK option, where are you investing in 2027.” The builders that answer those questions confidently are the ones still standing in three years.

Where this is going: a 2027 forecast

Forecasting AI in WordPress in 2026 is hard because the rate of change keeps accelerating. Two years ago, the conversation was about how to write blog posts faster. Six months ago, it was about whether layout generation would ever be reliable. Today, the conversation is about agent-writable infrastructure shipping into WordPress core. The horizon keeps moving up the stack.

Here’s what I’d bet money on for the next eighteen months, ordered from most to least confident.

The visual builder becomes the override, not the default

Today, you open your page builder and drag elements to make something. AI helps fill gaps. By end of 2027, that ordering flips for most users: you open an AI assistant, describe what you want, and the visual builder becomes the place you go to refine what AI produced. This is the Cursor pattern — a code editor where the AI pane is the primary surface and the editor pane is the manual override — applied to page builders.

Elementor’s Angie and Divi 5’s multi-agent flow are the early shapes of this. By 2027, the AI assistant pane is the default workflow in every major builder, and “manual drag-and-drop building” becomes the secondary mode users invoke when AI doesn’t produce what they want. The skill that used to be “knowing your way around the builder” becomes “knowing how to prompt and refine.”

HTML→Builder is universal table stakes

Bricks 2.3 made it canonical in March 2026. By the end of 2026, expect Elementor to ship native HTML import — the competitive pressure from Web2Elementor and the broader third-party converter market is too high to ignore for much longer. By mid-2027, every major builder accepts HTML/CSS as a first-class input format. The differentiation moves to how well they handle imports — semantic mapping quality, framework awareness (ACSS, Tailwind, design tokens), framework conversion (taking one builder’s HTML output and importing it cleanly into another).

The corollary: AI-first design tools like Claude Design, v0, Lovable, and Bolt become the default frontends for designing WordPress sites. Builders become the rendering and management layer. Designers don’t open their builder to design — they open it to deploy.

MCP becomes the universal protocol for WordPress plugins

WordPress 7.0 ships with the Abilities API, the WP AI Client, and the Connectors API in core. Once that’s released — likely summer or fall 2026 if the current slippage pattern continues — every meaningful WordPress plugin will expose an MCP surface within 12 months. Page builders, e-commerce platforms (WooCommerce), forms, SEO plugins, security plugins, analytics — all of them. The plugin author who doesn’t ship MCP tools falls behind the plugin author who does, because users with agents will simply route around the laggards.

By end of 2027, “Does it have an MCP server?” is a baseline expectation when reviewing any WordPress plugin, the way “Does it work with PHP 8?” is today. The plugins that get it wrong (clunky tool definitions, slow servers, poor error handling) become the ones agents avoid even if they technically support them. The plugins that get it right become disproportionately powerful because agents can compose them.

Agentic site management becomes the new differentiator

One-shot generation (type a prompt, get a layout) is becoming commodity. The next axis of competition is ongoing AI work — agents that operate sites continuously rather than building them once.

Concretely: an agent that audits your site weekly for dead links, accessibility regressions, image alt-text gaps, slow Core Web Vitals, and SEO drift, then opens prioritized drafts of fixes for you to review. An agent that reads incoming form submissions and auto-categorizes them. An agent that ingests your support inbox and auto-drafts knowledge base articles. An agent that A/B tests headline variants on landing pages and promotes winners.

This is the layer where WordPress.com’s Claude Connector is most ahead. Their March 2026 expansion to full MCP write access positioned them to be the platform where agentic site management actually works end-to-end. By 2027, expect 10Web, Hostinger, and one or two major managed WordPress hosts to ship “AI Site Manager” products that bundle agentic features at the platform layer. Self-hosted WordPress catches up once 7.0 ships and the broader plugin ecosystem builds MCP servers.

Frontier model integration gets formal

Anthropic’s existing WordPress.com Claude Connector is the first major frontier-model-lab-meets-CMS-platform partnership. By 2027, expect at least one more: OpenAI partnering with a major hosting provider or builder, Google integrating Gemini more deeply into Google’s WordPress hosting (via Google Cloud), or Anthropic deepening its WordPress presence beyond WordPress.com.

These partnerships matter because they create native integrations that BYOK plugins can’t match — preferred-customer pricing, fine-tuned models for WordPress-specific tasks, direct connector experiences inside the model lab’s own products (Claude Desktop, ChatGPT canvas, etc.). The model labs see WordPress as 43% of the web; the WordPress platforms see model labs as the operating system for their next decade. The incentives align.

Cost per generation drops 5–10x

Model providers have shipped consistent cost reductions every year since 2023. By end of 2027, the cost of generating a full-site layout drops to the point where AI generation is functionally free for end users. This kills the credit-pool business model first-party AI tools have relied on. Either those tools shift to flat-rate (which compresses margins), or they shift to BYOK (which exposes the markup), or they shift to bundled-with-something-else (which is what’s happening to Jetpack AI, Divi AI, and Elementor AI today).

The third-party BYOK ecosystem benefits the most from this trend. By 2027, BYOK is the dominant pricing model for serious AI in WordPress, with bundled credits surviving only at the platform-managed tier (WordPress.com, Hostinger AI).

Builder consolidation accelerates

WordPress has too many page builders. AI compresses the relevant differentiation between them — when AI generates the design and AI manages the content, the builder’s job narrows to rendering and managing. The builders that survive 2027 will be the ones with (a) strong agentic/MCP integration, (b) high design quality out of the box, and (c) a viable third-party ecosystem.

My informed guesses on who wins and loses, with the caveat that this is the most speculative part of the forecast:

  • Bricks survives strongly. The HTML→Builder bet positioned them correctly; the third-party ecosystem is the healthiest in the market.
  • Elementor survives by scale alone but loses share at the high end. Angie helps; missing HTML import for two years has hurt.
  • Divi survives via brand and Elegant Themes’ ecosystem lock-in. The full-site multi-agent flow is genuinely good for beginners.
  • YOOtheme Pro survives via design quality and a smaller but loyal base. The shipping of first-party AI (signaled but not yet executed) determines whether they grow or stagnate.
  • Beaver Builder depends on whether the BeaverForge ecosystem keeps growing. First-party AI skepticism worked in 2024–2025; it’s expensive in 2027.
  • Oxygen/Breakdance depends on Soflyy executing on the shared engine and on OxyDance Pilot continuing to fill the AI gap.
  • Gutenberg wins by default because it’s in WordPress core. AI Engine, Spectra, Kadence, and the HTML-to-blocks converters all become more relevant as WordPress 7.0 ships and Gutenberg gets MCP-native.

At least one current top-five builder is acquired or sunsets by end of 2027. I don’t know which.

What’s NOT going to happen by 2027 (with high confidence)

A few negative predictions worth committing to:

  • AI does not replace page builders. Builders become more important, not less, because they’re the rendering and management layer for AI output. The “AI killed page builders” headline doesn’t happen in this window.
  • A single dominant “WordPress AI” product does not emerge. No equivalent of “Cursor for WordPress” wins the space. The market stays fragmented because WordPress itself is fragmented. Multiple winners coexist.
  • MCP doesn’t get displaced by a competing standard. It’s won. Every credible competing protocol either dies or gets absorbed.
  • Self-hosted WordPress doesn’t lose to fully-managed AI hosts. WordPress.com and 10Web grow, but they don’t dominate. The decentralized, self-hosted majority of WordPress is reinforced by the WP 7.0 AI primitives, not displaced.

The thesis underneath all of it

Page builders are converging on the rendering and management layer. AI is the design and content layer. HTML and MCP are the interchange formats between them. WordPress core is becoming agent-aware infrastructure. The builder you choose matters less than how cleanly it sits inside that picture.

The bet I’m making with YOOforged — and the bet Bricks is making, and the bet WordPress.com is making — is that this picture holds. The bet Elementor and Divi are making is that their first-party AI experience can keep enough users in the closed loop to matter. Both bets are reasonable. The next eighteen months tell us which one was right.

Frequently asked questions

Is AI ready to build production websites in 2026?

For content-heavy sections — heroes, features, testimonials, FAQs, blog templates — yes. AI output quality matches or exceeds template-driven design. For interactive components, complex dynamic data, or anything beyond marketing surfaces, AI still needs human refinement and the builder’s manual tools.

Will AI replace page builders?

No. AI generates designs well; page builders manage them — content updates, responsive states, dynamic data, themes, version control, team workflows, role-based editing. The combination is stronger than either alone. Builders are becoming the rendering and management layer for AI-generated design, which is a more valuable position than they had before AI.

Which AI page builder is best for beginners?

Divi 5 with Divi AI has the smoothest first-time experience — type a prompt, get a working site. 10Web is the strongest standalone option if you’re starting from zero and not committed to a specific page builder. For Gutenberg-based starts, WordPress.com’s AI Website Builder plus the Claude Connector cover the full beginner workflow.

Which AI page builder is best for developers?

Bricks 2.3 with its native HTML/CSS converter and the BricksPilot/Bricksfusion/Code2Bricks ecosystem. YOOtheme Pro with YOOforged. Builderius for stateful Claude integration. All three reward technical users with cleaner output, BYOK pricing, and source flexibility.

What about Gutenberg AI? Is the core editor a viable AI workflow?

Yes, and increasingly so. AI Engine, Spectra Pro, Kadence AI, and the WordPress.com Claude Connector together cover most of what a page builder’s AI does. The HTML→Gutenberg converter market is fragmented (Gutenizer, Parzero, Telex) — no single tool dominates — but the trajectory is clear. WordPress 7.0’s Abilities API in core will accelerate this further.

Can I migrate AI-generated layouts between page builders?

Not directly via native formats. Each builder’s JSON is incompatible. But layouts that originated as HTML — the HTML→Builder workflow — can be re-imported into any builder that accepts HTML. ClonewebX is the only tool that operates element-level across builders today (Webflow, Elementor, Gutenberg, Bricks, Breakdance, Divi 5). For everything else, HTML is the interchange format.

What about Model Context Protocol (MCP) for WordPress?

The WordPress Core AI team shipped the official MCP Adapter in February 2026. WordPress.com runs a Claude Connector (read access since Feb 5, write access since March 20). Builder-specific MCP servers exist for Elementor and Bricks. Respira covers twelve builders via a single server. If your builder doesn’t have an MCP surface, expect one within a year — it’s becoming a baseline expectation, and WordPress 7.0 makes it part of core.

Are there security risks with AI page builder plugins?

Yes — and the May 2026 baseline reflects this. The AI Engine plugin had a vulnerability affecting approximately 100,000 sites earlier this year. AI plugins routinely hold API keys, manage tokens, and execute privileged operations. Treat them as part of your security surface, not just your feature surface. Audit, update, and scope API keys narrowly.