Credits

Credits are the unit this platform bills in. Each shop has its own balance, and you always see what something costs before you agree to it.

There are no subscriptions, no per-word pricing and no surprise invoices. A credit is spent when something you asked for actually succeeded — a piece of content that came out of the AI, or a piece of content that arrived in your shop.

What costs credits

Event	Charge
A generated item is completed	One charge per item. An item is one subject in one language — one product description in German is one item, the same product in English is a second. The price depends on the AI model and the task.
An item is successfully sent to your shop	One flat charge per item, the same no matter what the item cost to generate.
An FAQ suggestions item is completed	One charge per category — not per suggestion. An item costs the task’s price whether it produces two suggestions or ten, and it writes nothing into your shop yet. Sending the accepted suggestions is charged per item, like any other push.
Web research for an item	An optional surcharge, added to that item’s price — only if the task has web research switched on and a lookup actually ran.
Keyword research for an item	An optional surcharge, likewise only charged when a live keyword lookup really happens. A result taken from the cache is free.

Three rules follow from this, and they matter more than the numbers:

Failures are never charged. An item that ends as Failed, and a push that ends as Push failed, cost you nothing. Retrying is free of any penalty.
Regenerating charges again, in full. The AI does the whole job a second time, so it is billed a second time. That is why the app asks you to confirm before regenerating.
Skipped items are not charged. If a task is configured to skip products that already have a description, the skipped ones cost nothing.

If your platform has not set a price for a particular model and task, that combination is simply free. The model then shows no credit cost when you choose it, and the task can show “This task currently costs no credits.”

Balance, reserved and spendable

You will see two numbers instead of one whenever a job is running. They mean different things:

Balance: Everything the shop has. This is what the Current balance figure shows.
Reserved: What running jobs still have to spend before they finish. The credits are not gone yet, but they are spoken for.
Spendable: Balance minus reserved — what a new job or a push may use right now. The wizard puts it plainly: “N are already reserved for runs that have not finished, so M can be spent right now.”

Why a run can be refused while you still have credits

Without the reservation, a big job could start, a second job could spend its budget, and the first job would then die of poverty half-way through — with your content generated but only partly paid for. The reservation is what makes a started run safe to finish.

If a run is blocked for this reason you are told exactly that: “Starting this job needs about N credits and only M are free right now (other runs still have credits reserved). Add credits, or wait for the running jobs to finish.”

You always see the total first

There are three points at which the app states the cost, and none of them spends anything:

In the job wizard, the What this will cost panel names your balance, the per-item price of the chosen model, and closes with “You will see the exact total for this run on the next page, before anything is generated.”
On the draft page (step 2 of 2), the What this run costs card gives you “about N credits”, the spendable balance and the note “This is an upper estimate — items that are skipped are not charged.” If the balance is short, it says by how much. Only Start generation spends anything.
On the push confirmation page, before content reaches your shop: “It costs N credits (M per item). Your shop can spend X right now.”

A test generation in the wizard is free — nothing is saved and nothing is charged. Use it before committing to a large run.

The Credits page

In the shop’s left sidebar, open Credits. Only Owners and Managers see it — an Editor or Auditor has no entry there at all.

The page opens with “Credit ledger for this shop — every generation, push, and top-up is recorded here.” and has three sections.

Current balance

The big number at the top. If part of it is reserved, the line underneath says “N of these can be spent right now — the rest is reserved for jobs that are still running.” Next to it sits a Get more credits button that opens a short explanation rather than a checkout.

At zero you get a warning box: This shop has no credits left. — “Nothing can be generated or sent to your shop until credits are added. Ask whoever runs this platform for a top-up — everything you already generated stays where it is.”

What each task costs

A read-only price list: one row per task, and within it one price per AI model, written as “<model>: N credits per item”. Underneath come the three extras:

Sending content to your shop — “Charged once per item you send, whatever it cost to generate.”
Looking the product up on the web — “Only when a task has web research switched on. Added to that task’s price, per item.”
Looking up search keywords — same principle, for keyword research.

Only tasks your shop is actually entitled to appear here. If nothing is switched on you see “No tasks are switched on for this shop yet, so there is nothing to price.” Prices can be negotiated per shop, so another shop on the same platform may see different numbers.

Recent activity — the ledger

Every movement of credits, newest first, in seven columns:

Column	What it tells you
Date	When it happened.
Action	What happened — see the list below.
Model	Which AI model was used, where one was involved.
Amount	Negative when credits were spent, positive when they were added.
Balance after	The balance immediately after this entry, so you can follow the arithmetic.
Actor	Who triggered it.
Job / comment	A link to the job it belongs to, and any note left with a manual adjustment.

The action types are:

Generation: <task> — one generated item.
Push — one item written into your shop.
Research — a research surcharge. Where it is folded into a generation entry it appears as “(includes N for research)”.
Top-up — credits added to your shop.
Adjustment — a manual correction, up or down, usually with a comment.

Two views share those columns, switchable with the buttons in the table header:

Grouped (the default) — “Grouped by job and day”. One line per job and day with the totals, and a ×N badge for the number of entries behind it. This is the view for “what did last week cost me”.
Single transactions — “One row per transaction”. Every individual charge. This is the view for “why exactly was I charged that”.

Until anything has happened the table reads “No credit activity yet.”

The shop Overview also carries a Credit balance tile — again only for Owners and Managers — showing how much was spent in the recent period and linking straight into this Ledger.

Getting more credits

There is no self-service top-up

You cannot buy credits inside the app, and no page will ask you for a card. Credits are added for you by whoever runs this platform. Ask them for a top-up and say how many credits you need — the price list on the Credits page tells you what your planned work will cost.

A useful way to phrase the request: count the products you want to cover, multiply by the number of languages, and look up the per-item price of the model you intend to use. Add one push credit per item on top. Round up — an over-estimate does not expire.

What happens at zero

Nothing is deleted and nothing is lost. Only new spending stops:

New jobs will not start. The draft page tells you how many credits you are short.
Content cannot be sent to your shop. The confirmation page hides its confirm button.
A run that hits zero part-way through stops there. Items already generated are kept and were charged; the rest is not. You are told: “Not enough credits left: this item needed N and M were available. Top up your credits, then run the job again.”
Everything you already generated stays exactly where it is, and everything already published in your shop stays published.

Reading, reviewing, syncing your catalog, editing templates and changing the configuration all keep working at a balance of zero.

Keeping the bill down

Use the free Run test in the wizard until the writing style is right — then start the real run.
Start with one or two products, look at the result, and only then queue the whole catalog. That is much cheaper than regenerating five hundred items.
Switch on the “skip what already has content” options so you are not paying to rewrite what is already good.
Only switch on web or keyword research where it genuinely improves the result — both add a surcharge to every single item.

See also Generating content for the cost preview inside the wizard, and Sending content to your shop for the publishing charge.