WordPress guide
Core's new AI Client gives every plugin shared access to your AI provider keys and a view-only request log. This guide covers the practical layer on top: which plugin is spending, what it costs, and how to cap it before the invoice surprises you.
Ownership disclosure
ChatbotEdge and Axtolab are operated by the same owner. This guide mentions two Axtolab plugins, so treat those mentions as a first-party explanation of how the tools work, not an independent third-party review. The WordPress 7.0 core behavior described here applies no matter whose plugins you use, and the links to Axtolab carry no affiliate tracking.
WordPress 7.0 (released May 2026) ships a core AI Client: a shared layer that stores AI provider keys once and lets any plugin on the site use them. That is convenient, and it is also how AI spend gets away from small teams. A chatbot, an SEO assistant, and an image generator can all bill quietly against the same key, and the provider invoice arrives as one undifferentiated total.
Core's answer is observability: Tools > AI Request Logs shows the calls. What core does not ship is attribution in dollars, per-plugin budgets, or any way to stop a misbehaving plugin at a limit. If a plugin update introduces a retry loop on a paid model, the log will faithfully record every request while the bill grows.
If you are still choosing the chatbot itself, start with the chatbot scaling-costs guide or the WordPress chatbot guide; this page is about the spend layer underneath whatever you install.
Step 1
List every plugin and theme feature that can call an AI provider: chatbots, content assistants, image generation, SEO helpers, translation, and automation plugins. On WordPress 7.0 they can all share the same provider keys, so each one is a line on the same bill.
Step 2
WordPress 7.0 ships Tools > AI Request Logs as part of the core AI Client. It is view-only, but it tells you whether AI calls are happening at all, how often, and against which provider. If the log is busy and you cannot explain why, stop and investigate before adding more AI features.
Step 3
The provider's billing page shows one total for a shared key. To see which plugin is responsible, add a monitoring layer that records each call with tokens, provider, model, and the plugin that made it, then estimates the cost per source over time.
Step 4
Decide what each plugin and the site as a whole may spend per month. Warnings are useful; an enforced stop is safer. WordPress 7.0 includes a core filter that lets a plugin block AI prompts before they run, which is what makes a real budget cap possible rather than an after-the-fact alert.
AI Spend Monitor by Axtolab (free) handles attribution: it listens to the core AI Client, records tokens, provider, model, and the calling plugin for each request, estimates USD cost per plugin from published list prices, and shows a 30-day dashboard under Tools. It makes no external calls and does not store prompt content. It has been submitted to the WordPress.org plugin directory and is in review there; until that listing is live it is available from axtolab.com.
Axtolab AI Spend Governance (paid) handles enforcement on top of the Monitor: per-plugin and sitewide monthly budgets, warning emails at a threshold you choose, and a hard stop that blocks AI calls through the WordPress 7.0 core filter once a budget is reached. Well-behaved plugins see AI as unavailable and hide their AI features instead of erroring; blocked calls are recorded so you can see what would have spent. Details and current pricing are on the product page.
The same-owner disclosure above applies to both. If you prefer a different vendor, apply the same checklist: does it attribute calls to a named plugin, does it estimate cost per source over time, and does it block at the core AI Client level rather than only alerting after the spend has happened?
Core stores provider keys site-wide, so every AI plugin draws from the same account.
Attribute per plugin before trusting any single plugin's own usage screen.
Core's AI Request Logs show activity but set no limits and assign no dollar amounts.
Add cost estimation and budgets on top of core, or check provider billing often.
Scheduled jobs, retries, and automations can spend without anyone in the admin.
Review the per-source table weekly until the baseline is boring.
On-site cost figures come from list prices and token counts, not your invoice.
Treat the provider's billing dashboard as the source of truth for money decisions.
Spend control is the boring half of AI on WordPress, and it is the half that keeps the interesting half affordable. Set the budget before the bot gets popular, not after the first surprising invoice. When you are ready to pick the customer-facing tool, the small-business chatbot picks page is the next step.
FAQ
WordPress 7.0 ships an AI Client layer and a view-only request log under Tools > AI Request Logs. That answers "is something calling AI?" but it does not budget, cap, or attribute spend in dollars. Core stores shared provider keys that any plugin on the site can use, so usage from different plugins lands on the same bill. If you need to know which plugin is spending what, or stop a plugin at a monthly limit, you need a plugin layered on top of core's AI Client.
Reviewed
Because WordPress 7.0 shares provider keys across all plugins, the provider's billing page shows one total, not a per-plugin split. To attribute spend you need a monitoring layer on the site that records each AI call and resolves which plugin made it. AI Spend Monitor by Axtolab (free; operated by the same owner as ChatbotEdge) records tokens, provider, model, and the calling plugin per request and estimates cost per plugin. Whatever tool you use, confirm it attributes calls to a named plugin rather than only logging raw requests.
Reviewed
Core itself does not enforce budgets, but WordPress 7.0 includes a filter (wp_ai_client_prevent_prompt) that lets a plugin block AI calls before they run. Axtolab AI Spend Governance (operated by the same owner as ChatbotEdge) uses that filter for per-plugin and sitewide monthly budgets with warning emails and a hard stop. Any plugin built on the same core filter could do this; check that whatever you choose blocks at the core AI Client level rather than only warning after the bill arrives.
Reviewed
Treat any in-WordPress cost figure as an estimate, not a bill. Plugin-side numbers are computed from token counts and published list prices, which drift as providers change pricing, and they cannot see negotiated rates, free tiers, or non-WordPress usage on the same key. Use the on-site estimate to compare plugins against each other and catch runaway growth early, and use the AI provider's own billing dashboard as the source of truth before making budget decisions.
Reviewed