Overview
Since version 2.1.4, the Dashboard shows an estimate of what your chatbot costs in OpenAI credit: today, in the last 7 days and this month, per conversation and per feature. You don’t need to open your OpenAI account to know whether your chatbot costs cents or dollars.
You’ll find it in WPiko AI Chatbot → Dashboard:
- AI cost this month: the fifth card at the top. It shows this month’s estimate and, after 3 days, what the month is on track to cost. See details jumps to the full report.
- AI usage & cost: the full report, next to Recent activity.
What the Report Shows
| Part | What it shows |
|---|---|
| Today, Last 7 days, this month | The estimated cost and the number of AI replies. Days and months follow your site’s time zone. |
| On track for about… | What this month will cost at the current daily rate. Shown once there are at least 3 days of data in the month. |
| Last month | Last month’s total, for example “October 2026: $3.87”. Shown once the report was already running last month. |
| Per conversation, Per AI reply, Conversations | The average cost of one visitor conversation and of one AI reply this month, including knowledge search. Very small amounts keep two significant digits, for example $0.000072. |
| By feature | This month’s cost of each feature, with a bar that shows its share (see below). |
| Tokens and models | Input tokens (and how many were cached, which costs less), output tokens (and how many were reasoning) and the cost of each model you used. |
| OpenAI usage | Opens your usage page on OpenAI, where you see your real spending. |
Cost by Feature
A feature appears in the list once it has been used this month.
| Feature | What it counts |
|---|---|
| Chat replies | Every AI reply to a visitor, including streamed replies and automatic retries. |
| Knowledge search | Each time the chatbot searches your knowledge base (uploaded files, Quick learn pages, Scan Website content, products) before it answers. OpenAI charges $2.50 per 1,000 searches, on top of the tokens. With an affordable model such as GPT-6 Luna, this is often the largest part of the cost. |
| Conversation translations | The Translate button in Conversations. |
| Connection tests | The tiny test message sent when you save or test your API key. |
| Order and product lookups (Pro) | The extra request the chatbot makes after it looks up an order or products in WooCommerce. |
| Website scan (Q&A) (Pro) | AI-written Q&A from Scan Website, including automatic updates. |
| Contact form writing help (Pro) | Messages the AI writes or polishes for the contact form. |
The Pro features are counted only while WPiko AI Chatbot Pro is active. Without Pro, they don’t appear and aren’t included in the totals.
How the Estimate Works
- Every reply from OpenAI says how many tokens it used. The plugin multiplies them by OpenAI’s list price for that model, at the moment of the request. Later price changes don’t change past costs.
- Cached input tokens are counted at OpenAI’s lower cached price. Reasoning tokens count as output.
- Only the chatbot’s own requests are counted. Other plugins or apps that use the same OpenAI key are not included.
- The price list is updated with the plugin. If a model has no known price (for example one a developer added), its tokens are still counted, the cost isn’t, and a note says so.
Your OpenAI bill is the final amount. The estimate doesn’t include taxes, credits or discounts, the storage fee for knowledge files (OpenAI’s first 1 GB is free), or a reply a visitor stopped by leaving the page before it finished. Check your real spending with the OpenAI usage button.
When Counting Starts
Costs are counted from the day your site was updated to 2.1.4 (or the day you installed the plugin). Earlier chats aren’t included. During the first couple of months, while the periods shown start before that day, the report says “Costs are counted from [date].”
What Your Limits Can Cost
In Advanced → User Limits, the Cost check uses the average cost of one AI reply in the last 30 days to show what one visitor at the daily limit, and your whole site at the site-wide limit, can cost per day. The amounts change as you edit the limits, before you save. See User Limits.
Lowering Your Costs
- Choose an affordable model, such as GPT-6 Luna (the default), and a Low or Medium reasoning effort. Compare the Per AI reply cost after a change. See Setting Up AI Models.
- Keep Topic focus on “Only questions about this website”, so visitors can’t use your chatbot as a free general assistant. See Assistant Instructions.
- Review your User Limits and the site-wide daily limit. See User Limits.
- Set a monthly budget in your OpenAI account, as a hard stop.
Stored Data
Each OpenAI request is saved as one row in the {prefix}wpiko_chatbot_usage table: the time, the feature, the model, the conversation ID, the token counts and the estimated cost. No message text is stored. Rows older than 400 days are removed automatically. The table is deleted with the other plugin data when you choose Delete all WPiko AI Chatbot settings, conversations and logs before uninstalling. See Data Handling.
For Developers
Change the prices with wpiko_chatbot_model_prices and wpiko_chatbot_file_search_call_price, the features listed with wpiko_chatbot_usage_sources and how long rows are kept with wpiko_chatbot_usage_retention_days. The wpiko_chatbot_usage_recorded action runs after each request is saved. See Developer Hooks and Filters.
