What the service does
It reads performance reporting for the Meta ad accounts you authorize and presents it in dashboards and tables. If you turn on the optional AI assistant, it also answers questions about that reporting and about your ads, and can create new images and videos for you to use. It does not run your advertising.
What it cannot do
The app requests read-only access. It cannot create, edit, pause, delete or re-budget campaigns, ad sets or ads, and there is no code in it that could. Any recommendation the assistant makes is something for you to carry out yourself in Meta, and any image or video it creates is yours to download: nothing is uploaded to Meta or published for you.
Your account
- You create an account with your own details, which have to be true, and with an email address you can read: it has to be verified before the account can be used. You have to be at least 18 years old.
- Keep your password to yourself, and do not use one you use elsewhere. You are responsible for what is done in your account. Anyone who can read your inbox can choose a new password for it, so the inbox needs looking after too.
- You may only connect Meta ad accounts you are authorized to access. The app verifies this with Meta, but the authorization itself is yours to hold.
- You may stop using the service and delete your account at any time.
Accuracy of the figures
Figures come from Meta’s Marketing API and are shown as Meta reports them. In particular, conversions and conversion value are attributed by Meta under the attribution settings shown alongside the data — they are not verified sales, not store revenue and not profit.
The app tries hard to be correct about what it does show: ratios are computed from period totals, unavailable values are marked unavailable rather than shown as zero, and incomplete periods are labelled. Even so, this is reporting, not accounting, and you should not treat it as a financial record. Meta can also restate figures after the fact.
The AI assistant
- It is optional and off until you add your own API key and accept the notice shown in Settings.
- You pay the providers, not us. Questions are charged to your own OpenAI or Anthropic account, and so is turning the sound of your video ads into text, which uses your OpenAI key. Images and videos are charged to your own Higgsfield credits; a video is charged by the second and costs far more than an image. An API key is billed separately from any consumer subscription such as ChatGPT Plus or Claude Pro. Where the app shows what something costs, that is an estimate: what you are charged is decided by the provider.
- What is sent. To answer you, the assistant sends the provider you chose your question and what it needs to answer it: figures, the names of the campaigns, ad sets and ads involved, and, when you ask about creatives, their text, images and frames of their videos, as well as any image you attach. To create an image or a video it sends Higgsfield the instruction and the pictures it is based on. The privacy notice says exactly what goes where. Each provider handles it under its own terms.
- It makes mistakes. Language models do. Check anything that would inform a spending decision against the dashboard figures it cites. The assistant sees a video as a handful of frames and a transcript, not as you would watch it.
- Limits apply to how many questions, images and videos an account may ask for in a given time. They are there to keep a mistake from spending your credits without end.
Images you attach and what the assistant creates
- Attach only what you may use. An image you attach is sent to your AI provider and, if you ask for something to be made from it, to Higgsfield. Attaching someone else’s ad to learn from it is one thing; publishing something built from their brand, their product or their people is another. The assistant is instructed not to copy another brand’s name, logo, product or people, but it can get that wrong, and what you publish is yours to answer for.
- Check every image and video before you use it. They are made by a model. Text can come out misspelled or changed, a face or a product can differ from the original, a spoken line can be mispronounced, and a price, a claim or a guarantee can appear that is not true of your business. Nothing the assistant creates has been checked by a person.
- Publishing is your decision and your responsibility. That includes Meta’s advertising policies, the rules on advertising where your ads run, the rights of anyone who appears in a picture or whose likeness a video is based on, and any duty to say that an image or a video was made with AI.
- Your rights in what is created come from Higgsfield’s terms for your account and from the law where you are. This service claims no rights in it and gives no assurance about them.
- Download what you want to keep. Images, attached or created, and videos are removed after 7 days. Each one shows how long it has left, and once removed it cannot be brought back.
- A provider may refuse. Higgsfield and the AI providers apply their own content rules and may decline a request. The app tells you when that happens.
Availability
The service is free and provided as-is. There is no uptime commitment, no support guarantee, and no promise that any particular feature will continue to exist. Meta, OpenAI, Anthropic and Higgsfield may each change or restrict what they offer in ways that affect what can be shown or made. Reasonable notice will be given before the service is discontinued.
Acceptable use
Do not attempt to access another user’s data, disrupt the service, or use it to circumvent Meta’s own terms or those of the AI providers. Do not use the assistant to make content that is unlawful, that deceives people about who or what they are looking at, or that uses someone’s likeness, work or brand without the right to. Accounts doing so may be suspended.