How to track AI usage and estimated cost

uplads tracks every AI generation it runs for you — ad copy, vision captions, brand voice and translations — and shows you exactly how many tokens you've used and roughly what it cost. Because you connect your own provider key, the provider bills you directly, so this panel is the easiest place to see your AI spend in one view. Here's how to read it.

Opening the AI usage panel

The usage panel lives inside your AI provider settings, right below where you choose your provider and model.

1. Go to Configuration and open the AI section.
2. Scroll down to the card titled Usage & estimated cost.
3. If you've run at least one generation, the stats load automatically. Use the refresh icon in the top-right of the card to pull the latest numbers.

If you haven't generated anything yet, the panel shows a short prompt instead of stats. Run a generation from the bulk launch wizard and it will start populating. First time here? Make sure you've set up your AI provider and API keys.

Lifetime vs. last-30-days

The top strip shows four lifetime totals: Total cost (estimated), Generations, Input tokens and Output tokens. These are running totals since you first used AI in uplads.

Inside the breakdown tables, each row also shows a smaller last 30d figure underneath the lifetime cost — but only when the recent number differs from the lifetime total. So if a model or feature was only used recently, the two will match and no separate 30-day line appears. This lets you spot at a glance whether your spend is mostly historical or happening right now.

Breakdown by provider, model and feature

The main table breaks usage down by provider and model — for example OpenAI · GPT-4o or Anthropic · Claude Sonnet 4.5. Each row lists:

• Cost — the estimated spend for that model (with the last-30-day figure when it differs)
• Tokens (in / out) — input and output tokens consumed
• Calls — how many generations ran on that model
• Last used — a relative timestamp like "3d ago"

To see the same spend grouped by what the AI was actually doing, click the By feature button at the top of the card. This opens the AI spend by feature dialog, which maps usage to friendly feature names:

• Ad copy generation — the standard copy generator in the launch wizard
• Per-creative copy (Vision) — copy written per creative using image analysis
• Brand voice — generations tied to your brand voice settings
• Website brief — copy drawn from a scraped website
• Multi-language translation — copy translated for multi-language ads
• Optimize — AI summary — summaries produced by the optimization audit

Each feature row shows its estimated cost, number of calls and when it was last used.

Reading recent generation requests

The Last used column in both tables is your quick view of recent activity. It shows a relative time — seconds, minutes, hours, days or months ago — for the most recent generation on each model or feature, so you can tell what's active versus dormant without digging through logs.

Why cost is an estimate

The dollar figures are estimates, not your provider's actual invoice. uplads calculates them from each provider's published per-million-token rates, multiplied by the input and output tokens recorded for each generation. As the card notes: the provider bills you directly — uplads never proxies your requests or charges you for AI.

A few things this means in practice:

• Published rates can change; uplads keeps its rate table current, but there can be a short lag.
• Your provider's real bill may include taxes, discounts, batching credits or other adjustments that this estimate doesn't model.
• For an exact figure, always check your billing page at OpenAI or Anthropic.

To put a ceiling on AI spend regardless of the estimate, see how to set a monthly AI spend cap. If generations are being blocked or failing, see why is AI copy generation blocked or failing.