AI content

AI Content Creator: bulk article drafting with a credit budget you set first

The problem with AI content is never generating it. It is generating it at a known cost, in a consistent structure, into a draft state where an editor can still do their job. GoGee's content creator is built around those three constraints.

You pick the writing model, the image model, the image quality and how many articles to run. The builder shows the credit cost before it starts, drafts each article into your blog as an unpublished draft, and reconciles your wallet afterwards.

GoGee feature series · 7 of 33

How it's actually built

Writing models: Gemini 3.1 Flash Lite, 3.6 Flash, 3.5 Flash, GPT-5 Mini, GPT-5.4, GPT-5.5
Image models: Economy, Standard, Alternative and Premium tiers
Article structure: 600–1,000 words, at least three real H2 headings
Output shape: Strict JSON: title, excerpt, body
Reference images: Up to 6, used to steer style and subject
Billing: Credit wallet with idempotent spend and refund on failure

Article pipeline

01Input
Brief and references
Up to 6 reference images
02AI
Draft
Selected Gemini or GPT model, strict JSON output
03Deterministic
Structure check
600–1,000 words, at least three real H2 headings
04Human
Editor approves
Drafts never publish unattended
05Output
Published article
Credit wallet spend recorded idempotently

Guardrails and fallbacks

Refund on failed generation
Image tiers from economy to premium

AI Content Creator, data flow, generated from the shared GoGee feature diagram template.

Cost is decided before generation, not discovered after

Each model carries a credit cost, and each image carries a cost per quality level. Total cost for a run is calculated as the writing cost plus the number of images multiplied by the image cost, and that number is shown against your wallet balance before anything runs.

The builder also enforces a per-run budget cap client-side with batch sizes of one, five, ten or twenty. If a run would exceed the budget, it does not start. If your balance is short, the error states exactly how many credits are required and how many you have.

Spending is idempotent and recorded in a ledger with the reason and the resulting balance, and partial failures are refunded. That is the difference between an AI feature you can hand to a marketing assistant and one you cannot.

Structure is enforced in the contract, not hoped for

The generation call requests strict JSON containing a title, an excerpt and a markdown body. The instructions demand a body of 600 to 1,000 words with at least three genuine H2 subheadings, no placeholder headings, and no repetition of the title or excerpt inside the body.

Because the response shape is fixed, drafting a hundred articles produces a hundred articles with the same skeleton, which is what makes them editable at volume and what keeps the on-page SEO audit green.

Imagery in the same pass

Cover and in-article images are generated in the same run using the image model and quality you selected, with an editorial photographic style applied by default. You can attach up to six reference images to steer subject and style toward your brand rather than generic stock aesthetics.

Image models are tiered on purpose. An economy model is right for a hundred internal articles; a premium model is right for the six pages that carry your campaign.

Cover image generated alongside the draft
Reference images steer style, colour and subject
Quality tier chosen per run, priced per image
Unused credits refunded when a step fails

Everything lands as a draft

Each generated article is saved as a draft post with plain-text content stored for search and the chosen cover image attached. Nothing is published automatically.

We are also direct about the limits: there is no automated fact-checking pass and no brand-voice verification step. Structure is enforced, truth is not. An editor still reads it. Anyone selling you "fully autonomous publishing" is describing a liability, not a feature.

Questions we get asked

Does it publish automatically?

No. Articles are created as drafts for editorial review, with structure enforced so review is fast.

How is it billed?

From a credit wallet. The cost of writing and imagery is shown before the run, spending is logged in a ledger, and failed steps are refunded.

Does the AI fact-check itself?

No, and we will not claim otherwise. There is no fact-checking or tone-verification step, structural rules are enforced by the prompt contract and an editor reviews the draft.