Understand credit estimates

An estimate answers one question: what would this operation cost if I ran it now? This page explains how reliable that answer is and when it can differ from the charge.

Where the number comes from

When you choose the estimate path, the tool sends the account service exactly the parameters it would send to generate — your prompt, the attached business context, the requested length, the selected items, the chosen profile — but asks it to price the work instead of doing it.

The account applies its pricing calculation to the size of the input, the requested output and the selected operation, rounds up to a whole number of credits, and returns that.

For long posts, full-website copy and WooCommerce descriptions, the estimate is a conservative allowance for the next step, not a fixed charge per word. It includes room for reasoning, applicable web research and processing context that has not been cached. A successful step uses measured usage when available and never charges more than its authorized estimate. If the service does not receive sufficient usage details, it uses that estimate.

Writing profiles and actual usage

The writing profiles use the same underlying model with different reasoning and research settings. Basic (called Standard in the post and website tools) does not request reasoning. Pro adds reasoning; Pro + internet also performs a separate web-research step. The underlying model is managed by the service and may change without changing these labels.

Longer business context and briefs increase input usage. Pro can consume more credits than Basic because its internal reasoning counts toward output usage, even when the visible text is equally long. Pro + internet adds research usage and search calls; subsequent writing uses that research without repeating the search on every section.

Creating reusable context in the provider cache can cost more than ordinary input, while reading cached context costs less. Reuse is not guaranteed, so estimates do not assume a cache discount. The final measured charge accounts for ordinary input, cache writes, cache reads and output separately. Request a fresh estimate after a change in service pricing or profile behavior.

Why the charge can still differ

The final charge can differ from an earlier estimate for several reasons.

Measured usage was lower than the allowance. Writing steps can use less output or reasoning than reserved, or receive a cache discount. The lower measured cost is used when complete usage is available.

You changed the form after estimating. Editing the prompt, the word count or the selection changes the input. Estimate again.

The operation has more than one stage. A tool that generates a structure and then generates content prices each stage separately, because the second stage's input does not exist until the first has run. The outline estimate does not include the article; the article estimate arrives afterwards, once there is an outline to measure.

The work expands beyond what you counted. Translating a post also translates the taxonomy terms attached to it when they have no translation yet. Those are additional charges the post count does not show.

The run was interrupted. A multi-stage run that stops halfway is charged for the stages that completed.

Compare estimates and charges for the same step and settings. A whole-project estimate and one completed step are not comparable amounts.

Reading an estimate sensibly

• Compare it with the balance you have, not with what feels reasonable. A number is only large relative to what you can afford to spend.
• Divide by the unit of work. For a batch of twelve images or forty translated posts, the per-item figure tells you more than the total.
• Estimate twice with different settings when you are unsure. Estimates are free, so pricing a 600-word draft against a 1,500-word draft before choosing costs nothing.
• Treat a surprising number as information. An estimate far above expectation usually means the scope is larger than intended — an outline with more sections than you wanted, or a selection that includes more items than you meant to tick.

What an estimate does not tell you

• Whether the result will be good. Cost tracks size, not quality.
• Whether you have enough credits. The estimate prices the work; the balance check happens when the work is submitted. A large estimate against a small balance will fail at that point. See Resolve an insufficient-credit error.
• Whether a domain limit will allow it. A per-domain spending limit is enforced separately from the account balance.
• How long it will take. Reasoning, research and longer output may take more time, but a credit estimate is not a duration estimate.

When there is no estimate

Some tools let you start without requesting a separate estimate. The account service still checks the available balance before each paid step. Skipping a separate estimate can be reasonable for a small, familiar operation — a single short paragraph in the editor.

It is not reasonable for anything batched, anything long, or anything on a site whose balance is shared with other sites. The cost of an estimate is a few seconds; the cost of a surprise is real credits.

After the operation

The service returns the amount actually used together with the new balance, and the tool stores that balance locally. Comparing the estimate with the balance movement is the quickest way to confirm that a run behaved as expected.

If they disagree in a way none of the causes above explains, collect the details listed in Information needed to review a charge before contacting support.