Message Credits & AI Tokens: Tracking and Optimising
What message credits and AI tokens are, what uses them, where Scope Wiser reports both, and the changes that reduce use without reducing results.
Message credits and AI tokens both go down on their own as your bots work. This page shows what uses each one, where to read the figures, and how to use fewer.
What each resource pays for
Message credits cover outbound messages on every channel — WhatsApp, Facebook, Instagram and Webchat. Every outgoing message uses credit: a template, a plain text reply, an image, and a broadcast to five thousand people counts five thousand times. Incoming messages do not use credits.
Two things surprise people. Automation counts — a sequence that sends four follow-ups to every new subscriber spends four credits per person. And WhatsApp adds Meta's own conversation charges on top of your Scope Wiser allowance, so a WhatsApp broadcast has a cost in two places.
AI tokens are used every time the AI writes a reply; longer conversations and longer instructions use more.
In short: message credits grow with how many people you contact, AI tokens with how much you say to each of them.
Where the numbers are
The top bar, on every screen. Three ring meters — Subscriber, Message and AI Token — each show usage against your allowance. 238/500.0K on the Message meter is 238 messages against a 500,000 ceiling. 1.1K/3.1K on AI Token is just over a third used. An infinity sign, as in 557/∞, means that allowance is uncapped on your plan.
Glance at it before you send anything large.
The Account page. Click your avatar at the top right and choose Account. Two panels in the right-hand column carry the detail: AI Token Usage Log — "Recent AI credit usage for your account." — and Usage Log — "Module limits and current usage for your account."
Read the AI Token Usage Log when the meter has moved and you do not know why. The Usage Log is the wider view, covering every module your plan includes.
Note the figures at the start of each month so a sudden jump stands out.
Reducing message credit use
Segment before you send. This is the single largest saving. A broadcast to your whole list when a label would have reached the relevant quarter costs four times as much and performs worse. Labels, Lists & Segments: How Scope Wiser Groups Subscribers covers building those groups.
Prune sequences. Every step of a follow-up sequence sends to everyone enrolled. Read your longest sequence end to end and remove steps that do not earn their place.
Watch for loops. A flow that can re-trigger itself, or two automations that answer each other, spends credits faster than any campaign. If use rises with no campaign to explain it, look here first.
Clean the list. Contacts that never respond still cost the same to message. Removing or excluding an unengaged segment cuts spend and lifts your delivery quality with Meta.
Reducing AI token use
Keep the instructions short. Keep the AI agent's instructions (AI → Agents) short — they are sent with every reply.
Ask for shorter answers. Tell the AI agent to reply briefly — long answers use more tokens.
Do not use AI where a fixed reply works. Opening hours, delivery times and returns policies do not need generating. A Keyword Reply answers them instantly and for no tokens.
Keep knowledge sets tight. A knowledge set stuffed with everything gives the agent more to read per question. Relevant content produces better answers and cheaper ones.
When you are close to a limit
If the Message meter is nearly full and a campaign is due: narrow the audience, postpone the non-urgent part, or top up — in that order.
If AI tokens are nearly gone, turn the AI Agent off on lower-value bots rather than letting it fail mid-conversation, and let Keyword Replies carry the load until the allowance resets or you buy more. A bot that sends a fixed reply looks fine to a customer; a bot that stops answering does not.
Check Renew / Upgrade on the Account page for more allowance, and read the Usage Log first so you buy the one you are actually short of.
What to do next
Billing, Plan Limits & the Usage Log puts these two allowances in the context of the rest of your plan, including addons and expiry dates. If subscriber counts rather than messages are the constraint, Subscriber Metrics & Recounts explains which figure to trust.