Jay's Ideas
Whenever tech commentators talk about artificial intelligence and the future of the web, WordPress is almost reflexively dismissed. The fashionable narrative says it's a bulky, twenty-year-old CMS with no place in a world where AI can vibe-code an entire web application over a weekend. I think that dismissal is a mistake. WordPress has a serious shot at becoming a strong open-source foundation for AI on the web. Not an inevitable winner, but a contender worth taking seriously.
If you want proof major infrastructure players see this coming, look at Cloudflare’s EmDash. Cloudflare recognizes WordPress’s potential as an open, common layer for AI applications. By attacking WordPress’s historic soft spot: plugin security, Cloudflare is attempting to solve the bottleneck of manual reviews while leveraging an ecosystem that powers over forty percent of the web. Their aggressive move suggests WordPress is a prize worth fighting over.
Much of the current developer conversation revolves around the Model Context Protocol, but the real connecting glue is the MariaDB relational database and the open integration of plugins. Together they form a decentralized web operating system that is not biased toward any single corporate giant. Salesforce, for example, is pushing Agentforce as a shift toward autonomous agents that execute workflows inside corporate guardrails. But proprietary platforms exist to lock you in and mine your data. An open foundation with no owner is far more appealing.
The bullish case for WordPress is straightforward. It possesses an enormous, decentralized network effect that no competitor can buy. Its data model is a relational schema that every AI model has seen millions of times, so agents can read, write, and extend it with almost no training. Its plugin ecosystem offers thousands of ready-made integration points for commerce, membership, and SEO. Furthermore, it already provides the roles, workflows, and editorial collaboration tools that are not trivial to recreate from scratch.
The bearish case is equally weighted. Many applications thrive as purpose-built tools, unencumbered by a CMS. Shared protocols and open standards like MCP, REST, and GraphQL APIs allow heterogeneous systems to interoperate without a common platform. As a recent WP Minute+ Podcast discussion pointed out, LLMs often don't recommend WordPress because installation is complex and many projects simply don't need a database. If a project doesn't require one, AI will rarely recommend an overly complex solution.
Add to that WordPress’s legacy liabilities: a massive attack surface, PHP hosting constraints, and two decades of accumulated cruft. While not fatal, these are reasons a rational architect might choose something else. WordPress has been losing the brochure-website market for years because those sites don't need structured data or content workflows. AI accelerates that shift, leaving WordPress better suited for complex publishing operations and sophisticated business sites.
Running underneath this is the question of how AI evolves. On one hand, models are trained overwhelmingly on popular patterns, which favors the dominance of PHP and WordPress. On the other hand, models are improving at learning unfamiliar frameworks on the fly. If AI becomes a translation layer that lowers switching costs, the existing training set stops being a ceiling and becomes a floor.
It is also worth remembering that businesses are often more conservative than developers. Many will wait and see rather than jump into unfamiliar territory. When the cost of AI inference, oversight and "prompt fatigue" exceeds the cost of hiring a developer who can guide agents from a position of expertise, the existing solution wins.
Skeptics often point to PHP as a liability, but that objection is weaker than it once was. AI models write PHP exceptionally well because they have been trained on vast oceans of it. Modern PHP 8 and runtimes like FrankenPHP have optimized speed, and agents can refactor bottlenecks on demand. The technical objections are simply not as decisive as skeptics assume.
