01 · THE MODEL PROPOSES
Budget $34 → $38
+12%The model proposes, the backend decides.
02 · THE BACKEND DECIDES
- Target CPA · $25 maximum
- Budget cap · $40 per day
- Maximum change · +20% per action
- Protected campaigns
Applied, and logged
Author, date, rationale and before/after values. Reversible as long as the Google Ads API allows it.
- Tokens encrypted, never passed to the model
- Paused rather than deleted
- Audit log on every action
Your money and your data, under control
SaveAds acts on real advertising accounts, where mistakes can cost money. Its architecture is designed to greatly reduce that risk through checks independent from the AI model.
Your money's safety, first
The AI suggests. The backend verifies and blocks actions outside your limits.
Server-side safeguards
Target CPA, maximum budget, maximum variation and protected campaigns are checked after every AI proposal to greatly reduce the risk of mistakes.
We pause, we don't delete
By default, no deletion. Gradual changes and cooldown periods to avoid breaking the Ads platforms' learning.
Full audit log
Who, what, when, why, before/after values — with rollback when the API allows it.
Your secrets protected
Your ad-platform access is encrypted, never stored in clear text and isolated per user. Technical controls limit the risk of unauthorised access.
Exactly what SaveAds accesses
Connections use each compatible platform’s official authorization flow. You see the requested scope and can revoke it from that platform at any time.
Google profile and email address
Create your SaveAds account and identify you at sign-in. No other use, never sold.
Google Ads API (read)
Read campaigns, ad groups, keywords, search terms, ads, statistics and conversion settings, to produce analyses and recommendations.
Google Ads API (write)
Apply a change you approved — or, at levels 4 and 5, an action explicitly allowed by your safeguards. Never outside that frame.
Meta Ads
Read authorised accounts, campaigns and statistics; write actions depend on the Meta permissions granted and your approval level.
TikTok Ads and LinkedIn Ads
Read campaigns and performance for selected accounts through their official permissions. Access is separate and revocable platform by platform.
Safeguards the AI cannot bypass
Target CPA, maximum budget, maximum variation per action and the list of protected campaigns are checked by the backend, after the model proposes. An out-of-bounds action is rejected, even if the AI considers it relevant. The model suggests, the backend decides.
We pause, we do not delete
By default no deletion is possible: a campaign or keyword deemed useless is paused, hence reversible. Budget and bid changes are gradual, with cooldown periods, so Google's learning is not reset.
Encrypted platform access
OAuth tokens for Google, Meta, TikTok and LinkedIn are encrypted and kept in a secret manager. They are never stored in clear text, returned to the browser or sent to the AI model. AI context is filtered.
Full audit log
Every action records its author, date, rationale and before/after values. You can reconstruct your account history, and roll back whenever the Google Ads API allows it.
Strict account isolation
Data is partitioned per user at the database level (Row Level Security): a query cannot reach another account's data, even if the application has a bug.
Google API Services compliance
Use of data received from Google APIs complies with the Google API Services User Data Policy, including its Limited Use requirements: no resale, no advertising use, no human access outside your consent, support at your request, security or legal obligation.
Ready to let AI run your campaigns?
Connect your account in 2 minutes. No change without your approval.