Plans, allowance and models

Octolift documentation

← Back to documentation

One number: product improvements

Each plan includes a number of product improvements per billing period. That is the only thing counted. One improvement is one run of Improve with AI on one product: the description, the titles, the category and the category field suggestions together. Scoring your products is included and does not use the allowance, however many you look at, because scoring is the cheap half of the work and it is what shows you which listings are worth improving in the first place.

The plans

PlanImprovements per month
Free10
Core150
Scale750
Enterprise3,000

Every plan has every feature and can use every suggestion model; the plans differ only in how many improvements they include. Prices are shown in Shopify's plan picker, which opens from Change plan (or See plans, before you have picked one) on the Plan and model page, and are billed by Shopify on your Shopify invoice.

What counts, and what does not

Uses the allowanceIncluded
  • Generating suggestions for a product (Improve with AI), including the rounds in which the draft is scored and polished before you see it.
  • Regenerating suggestions for the same product uses an improvement each time.
  • Loading category field suggestions after assigning a new category on the improve page.
  • Scoring a product, however often.
  • The category audit inside a scan.
  • How AI assistants see your product, the shopper questions and their rankings, Try your own search, and tracked questions.
  • Agentic sales.
  • Reviewing, editing, saving and rescanning.
  • A generation that failed or came back unusable is not charged.

Reopening a product you have already scored costs nothing extra: the text verdicts are kept and reused for 30 days while the listing has not changed.

Renewal, running out, changing plan

A separate, much larger ceiling on the included work (scans, the category audit, the shopper questions and their rankings) exists to stop runaway automation, not to limit you. The Plan and model page only mentions it once a store has used most of it, which normal use never does.

Choosing a suggestion model

Octolift runs on several language models from OpenAI, Anthropic and Google, and every plan can choose any of them on the Plan and model page. Your choice applies to writing suggestions and to nothing else. A model that costs more to run uses more of your allowance per product, so it costs you improvements rather than money; the plan page shows exactly how many each one leaves you before you switch, and marks the ones that can still cover your whole catalogue.

ModelProviderAllowance used per product
GPT-4o mini (default) OpenAI 1 improvement
Gemini 3.1 Flash Lite Google 2 improvements
Claude Haiku 4.5 Anthropic 12 improvements
GPT-5 mini OpenAI 15 improvements
Gemini 3.6 Flash Google 20 improvements
Claude Sonnet 5 Anthropic 40 improvements
GPT-5 OpenAI 88 improvements

Scoring always runs on the same model for every store, whichever one you pick here. A score is only comparable if one rater produced it: measured on the same listings, models disagree with each other by a wide margin, so a score that changed with your model choice would rewrite your own history and would not mean the same thing as anyone else's.

A more expensive model is not automatically a better one for this task. Switching is free, takes effect on the next generation, and does not change any score you already have. While support is looking into something for you, the model can be temporarily fixed by us; the plan page says so when that is the case.

Billing, trials and cancelling

Need a hand?

Open Support in the app menu to chat with us, or email support@craftshift.com. Tell us your store URL and the product you were looking at and we will pick it up from there.

Related guides