WP-AutoInsight 3.4: No Black Boxes

The frustration is always the same: you set up the content scheduler, it seems to work, and then three days later you check the site and nothing ran. Or you switch API providers and can't tell if the new key is actually valid until a post fails to generate. Or you have no idea whether you're burning $0.02 per post or $2.

That's what the new release of WP-AutoInsight does. It stops making you guess about what's running on your automated content generation.

What Does a Post Actually Cost?

WP-AutoInsight supports four providers across a range of models. GPT-4.1 Mini, Claude Haiku, Gemini Flash-Lite, Sonar — they're not equivalent. Different strengths, different context windows, different costs per generation. If you're choosing between them, you should know what you're choosing between.

The AI Models tab now shows an estimated cost per post below each model card. It's not just a tier label anymore: GPT-4.1 Mini around $0.0002, Claude Opus around $0.015. These are estimates based on typical generation length, not guarantees, but they're in the right order of magnitude and give you enough to make a real decision. Will it be cheaper to use Claude Opus and Nano Banana, than letting OpenAi create the content and the images? Which setup outputs better results? You can test and decide, without surprises.

SaaS content tools don't show you this because their margin depends on you not thinking about it. The direct API costs are low enough that the economics are clearly on your side. The numbers make that case without needing to argue for it.

What Is the Scheduler Actually Doing?

The scheduling status used to be a frequency setting: set it, forget it, and hope. No confirmation that anything was queued. No indication of when the next post would run. No way to tell, at a glance, whether you'd accidentally disabled it six weeks ago.

It now shows human-readable text: "Next post scheduled in 6 hours — Saturday, March 15 at 9:00 AM." With a direct link to the scheduling settings if you want to change it. One line, always visible.

The Changes That Don't Have a Settings Page

A few things in 3.4 don't show up as new UI. But they make the output better.

SEO generation was rewritten to request structured JSON from the model instead of parsing free text. This sounds unnecessary, but it's the basic step to make WP-AutoInsight fully compatible with YOAST, RankMath and other SEO plugins.

The plugin now pulls your site name and description from WordPress settings and includes them in SEO prompts. This sounds minor, right? But now the generated SEO content is for your site.

Featured images now get alt text populated automatically. It was empty before, which is both an accessibility gap and an SEO issue.

OpenAI and Stability AI image generation now have the same size controls other providers already had.

Why This Release Is Called "No Black Boxes"

Every content tool running on someone else's infrastructure has structural incentives to obscure things from you. Hidden costs keep you from switching. Vague status messages keep you from questioning whether it's working. Opaque scheduling keeps you from noticing that nothing ran last week.

WP-AutoInsight runs on your server. The incentive structure is different. It's an automated content generator that you own. There's no reason to hide what a every post created costs. There's no reason to make your life more complex than it already is, with hidden settings and lack of information.

3.4 is the release that acts on that logic.

For now: Download WP-AutoInsight 3.4, validate your API keys, and check what your model actually costs per post.

If you're managing content at scale across multiple WordPress sites and want a setup that doesn't add a subscription on top of API costs you're already paying, that's the work I do