This policy explains how Coretas uses artificial intelligence, what its output is and is not, where the human decision sits, and who is responsible for what.
1. Where AI is used in Coretas
Coretas uses AI models, including large language models and statistical models, to:
- analyse historical performance across your connected advertising accounts,
- produce insights, diagnoses and explanations of what happened and why,
- generate recommendations, including budget and structural recommendations,
- generate media plans, campaign structures and advertising components including copy and configuration,
- generate reports, and
- answer your questions through Coretas Copilot.
2. You approve, we do not act alone
Coretas does not make changes to your advertising accounts on its own. Every change Coretas prepares is shown to you for review, and reaches your advertising account only after you approve it.
Approval is a real decision, not a formality. Before you approve, review the budgets, bids, targeting, copy and account destination that Coretas has proposed. Once you approve, the change is yours.
We record what was shown to you, what you approved, and when.
3. What the output is
AI generated output is an estimate produced from the data available to Coretas at the time. It reflects patterns in historical performance, not knowledge of the future.
Output from Coretas:
- is not professional advice of any kind, and not legal, tax, financial or accounting advice,
- is not a prediction, forecast commitment or guarantee of performance, revenue, return on ad spend, conversion volume or cost,
- may be incomplete, may contain errors, and may rest on platform data that is itself incomplete, delayed, modelled or later restated by the platform,
- may reflect correlation rather than causation. An insight that a channel is associated with revenue does not establish that the spend caused the revenue, and
- can be wrong in ways that look plausible. Language models can produce confident, well-formed output that is factually incorrect.
Where Coretas explains the reasoning behind a recommendation, that explanation is itself generated output and is subject to the same limits.
4. What you are responsible for
- Reviewing output before you act on it. Especially anything that changes spend.
- Advertising claims. Copy and creative that Coretas generates may state or imply claims about your products. You are responsible for their truth, their substantiation, and their compliance with advertising law and platform policy. Do not publish a generated claim you cannot support.
- Regulated categories. If you advertise in a category with specific rules, including health, supplements, finance, alcohol, gambling, children's products, or anything else your regulator treats specially, you are responsible for meeting those rules. Coretas does not screen output for category compliance.
- Intellectual property. Check that generated copy and assets do not use a third party's trademark, protected work or likeness without permission.
- Your advertising spend. All of it. See section 4 of the Terms of Use.
5. Rights in generated output
You may use output Coretas generates for you in your own business, including in live advertising campaigns, without further permission from us.
Generated output may not be unique. Similar prompts and similar data can produce similar output for other customers. We do not guarantee that output is original or that you can claim exclusive rights in it.
You may not use Coretas output to build, train or evaluate a competing advertising intelligence product, or resell it as a standalone data or analysis product.
6. How your data is used with AI models
To produce output, we send relevant parts of your data to OpenAI, acting as our subprocessor. It is listed under Subprocessors.
Your data is context, not training material. Your advertising data is given to the model as context for your own request, which is what makes the output specific to your account instead of a generic benchmark. Under OpenAI's published API data policy, data sent through its API is not used to train or improve OpenAI's models unless the customer opts in, and we do not opt in. OpenAI may retain API inputs and outputs for up to 30 days for abuse monitoring before deleting them, unless it is legally required to keep them for longer.
Long-term memory. The more you use Coretas, the better the output gets, because Coretas keeps a working context for your account built from your data, your previous questions, and the plans and campaigns you have worked on. That context is held inside Coretas, belongs to your account, is never used to produce output for another customer, and is deleted with the rest of your data when your account ends. You can ask us to reset it at any time by emailing info@coretas.ai.
We do not use one customer's identifiable data to produce output for another customer. We may use aggregated and de-identified insights to improve the platform, as described in section 8 of the Terms of Use.
We may change the model providers we use, the models themselves, and how context and memory work, at any time and without prior notice. The current provider list is always the one published under Subprocessors.
7. Limits and known failure modes
We would rather tell you where the platform is weak than let you find out on a live campaign.
- Attribution. Coretas reports what platforms report. Platform-reported conversions overlap across channels, and no cross-platform view resolves that fully. Treat platform ROAS as a signal, not as a measure of incrementality.
- Incrementality. A recommendation cannot tell you whether a sale would have happened without the ad. Branded search is the clearest example. Where incrementality matters, test it.
- Offline and unmeasured conversions. If a conversion happens somewhere the platform cannot see, including a retail point of sale, a marketplace or a showroom, that data does not exist in the account and the analysis cannot account for it.
- Short histories and low volume. Accounts with little history or thin conversion volume produce weaker analysis, and recommendations from them carry more uncertainty.
- Recent changes. Structural changes made outside Coretas may not be reflected until the next data read.
8. Human oversight and automated decision-making
Coretas is designed so that decisions affecting your advertising accounts are made by a person at your company. We do not use AI to make automated decisions about individuals that produce legal effects for them or similarly significantly affect them.
If we introduce a feature that acts on your accounts without a human approval step, it will be optional, clearly identified, off by default, and governed by terms we present to you before you enable it.
9. Reporting a problem with output
If Coretas produces output that is wrong, unsafe or inappropriate, tell us at info@coretas.ai. We read every report and it is the fastest route to a fix.
Changes to this policy
We may update this policy at any time. Unless the law, or an agreement we have signed with you, requires otherwise, a change takes effect as soon as we post it here, without prior notice to you, and the version on this page is the version that applies. We recommend checking this page from time to time. Prior versions are available on request from info@coretas.ai.
Material changes to the Terms of Use themselves follow the notice process in section 19 of those Terms.