Provero AI Policy

How Provero uses AI

Provero uses Google Gemini for exactly two things, and **neither of them produces a number.**

The first happens before you ever open the app. US chains with 20 or more locations are required to publish full nutrition information, and those disclosures arrive as PDF and HTML documents in whatever shape each chain chose. The model puts those rows into one consistent schema: units, portion labels, dish naming. It does not change a value, and the figure that reaches your screen is the figure the chain published.

The second happens while you are reading a menu. On the top three dish rows only, the model writes one sentence explaining why that dish is near the top. That sentence is the whole of what AI produces inside the app.

The model runs on our servers, never on your device, and our access key is never included in the app.

The brief the model is given is a list of things it may not do. **It may not produce, adjust, round, scale, average or fill any calorie or protein figure**, and where a chain has not published a figure the app says so rather than letting a model supply one. It may not decide the order of the menu, which is arithmetic we do ourselves against the two targets you entered. It may not use the words healthy, good, bad, clean, guilt or cheat, or any word that turns a number into a verdict about a person. And it is instructed not to give medical, dietary, psychological, legal or financial advice, regardless of what it is asked.

If the sentence cannot be written, because you are offline or the call failed, the dish appears without it and nothing else on the row changes. The gap is never filled with a guess.

There is no chat surface in Provero, at any tier.

Informational only

The sentence the AI produces is informational. It is a starting point for your own judgement, not a verdict and not professional advice. If a result matters to a decision with real consequences, check it against a person who knows your situation.

The figures beside it are a different kind of thing: they are the chain's own published disclosure, and every one of them carries a link to the document it came from, so you never have to take our word for a number.

You decide

You stay in charge. Provero shows you what a chain published, ranked against numbers you chose yourself, and you decide what to order. We do not use AI to take actions on your behalf, and there is no step in the app where a model's output is applied without you seeing it.

Automated decision making

Nothing Provero generates produces legal effects or similarly significantly affects you, and we do not use it to profile you beyond producing what you see on the screen. The ranking is arithmetic against your own targets, not a judgement about you, and nothing in the app scores you for a purpose outside the app.

Limitations

AI-written text can be wrong, incomplete, or confidently mistaken. It can miss context that is obvious to you. It reflects patterns in the data the model was trained on, which do not represent everyone equally. Two runs on the same input can differ.

That applies to the one sentence Provero writes. It does not apply to the figures, which are not generated: they are read from a chain's published disclosure, and they carry the date the chain published them and the date we last read them, both visible on the screen. A chain can change a recipe or revise a disclosure after we last read it, and the dates are there so you can see how recent the figure is rather than having to assume.

We would rather say this plainly than let a confident tone imply a reliability the technology does not have.

Privacy

To write the sentence, we send the dish, the chain's published figures for it, and the calorie and protein numbers you have left today. That is sent to the model provider only to generate that sentence. It is not used to train their systems, and we do not send your email address or your account identifier with it. Your logged history, your usual orders and your week are never sent. The normalisation work described above involves nothing of yours at all. Our Privacy Policy describes what we store afterwards and for how long.

What we do not claim

We do not claim that Provero is an expert, that it replaces a professional, or that its results are guaranteed. It is a reference tool that re-sorts published figures against two numbers you gave it, and it is useful in proportion to how honestly it is described.

We do not certify or endorse any chain's nutrition information. We republish what the chain published, with a link, and the link is there because you should be able to check us.

Changes over time

We change the prompts and the models behind Provero as better ones become available, which means the one written sentence can shift over time. Where a change materially alters what the app tells you, we will say so in the app rather than let it change silently. A change to a model never changes a figure, because a model does not produce one.