AI usage
See the AI tokens and the searches, lookups and screenshots the agents made for each site, per job, ticket and day.
The AI usage page shows what the agents used for each site: AI calls, input and output tokens, and the searches, keyword lookups and screenshots they made. Use this page to see how much of your plan each site has used and to find expensive jobs.
Open the page
Click AI usage in the site menu. Choose a period at the top: 7 days, 30 days or 90 days. The default is 30 days.
The Overview page also shows AI tokens, 7 days (in / out). Click it to open the AI usage page for the last 7 days.
What the page shows
Totals
| Number | Meaning |
|---|---|
| Input tokens | Tokens sent to the model. A note shows how many more were read from cache |
| Output tokens | Tokens the model wrote. Includes the model's thinking |
| AI calls | Number of requests to the model |
| Searches and lookups | Web searches, keyword lookups, SERP lookups and screenshots |
By job type
Tokens and calls for each kind of job, such as keyword research, AI visibility checks, briefs, writing, QA and social posts. This shows where your tokens go.
By model
Tokens per model. Useful when you use different models per job.
Searches, lookups and screenshots
Counts per service: Search Console reads, Google Autocomplete, Bing, DataForSEO, SERP services, AI web searches in ChatGPT and Gemini checks, and screenshots.
By day
A table with Day, AI calls, Input tokens, Output tokens, Searches and Screenshots.
Usage per ticket and per run
- Per ticket: open a ticket and find AI usage for this ticket. It adds up the brief, every draft and every QA review.
- Per run: open Agent activity, choose a run, and read AI usage for this run.
Tip: Look at a ticket that needed 3 rewrites. If most tokens went into rewrites, the cause is often a missing product fact or a strict rule. Fix the cause with a lesson instead of paying for more rewrites.
Turn tokens into money
AgenticSEO shows tokens, not prices, because your provider sets the price and it can change. To estimate cost:
- Note the input and output tokens for a period, per model.
- Open your provider's price list.
- Multiply each by the provider's price per million tokens.
cost = (input tokens / 1,000,000 x input price)
+ (output tokens / 1,000,000 x output price)
Cached input is usually cheaper. Check your provider's pricing for cache reads.
Note: Your provider's dashboard is the final word on your bill. Small differences can happen, for example from requests that failed or from time zones.
Lower your AI cost
| Change | Where | Effect |
|---|---|---|
| Plan usage for a site | Settings > AI prompts | Check runs per prompt and tracked prompts against your plan limits |
| Fewer runs per prompt or fewer prompts | Settings > AI prompts | Fewer ChatGPT and Gemini calls |
| Fewer AI checks | Settings > AI prompts | Lower the tracked prompts or runs per prompt |
| Fewer articles per week | Settings > Weekly schedule | Fewer drafts and QA reviews |
| Better approved facts and lessons | Settings > Business profile, Lessons | Fewer QA rewrites |
| Fewer research tickets | Settings > Weekly schedule | Fewer briefs |
AI visibility checks cost about prompts x runs per prompt x engines AI calls a week. See Mention rate.
What is counted
| Counted | Not counted |
|---|---|
| Every AI call by the agents: input, output and cached tokens | Your use of the same key outside AgenticSEO |
| Every web search, keyword lookup, SERP lookup and screenshot | Your CMS or analytics provider's own usage |
| ChatGPT and Gemini visibility checks | X API charges (see your X Developer Console) |
Related
Did this page help? Tell us at support@agenticseo.com.