# What Data Does the AI Actually See From Google Ads?

Three parties could see your ads data — connector, AI vendor, model — under different retention rules. A map of who sees what, and how to shrink exposure.

Questions · Published 2026-08-10, updated 2026-11-11.

> Three parties see three different things. The connector sees what its tools return and keeps a bounded part of it: it caches audit reads for up to an hour and keeps a saved audit's findings for 180 days, beside your sign-in identity, one encrypted Google refresh token, the connection record, a ledger entry per call and your workspace's business profile. The AI vendor (OpenAI, Anthropic) sees whatever enters the conversation, kept under your plan's retention and training policy. The model sees only the conversation's context. Your Google password and payment details never enter the pipeline at all.

"Is my data safe" gets answered with reassurance too often. More useful is
a map: when an AI reads your Google Ads account, exactly three parties
could see something, they see different things, and different rules govern
each. Once the map is clear, the controls become obvious.

## The three-party map: connector, AI vendor, model

The pipeline runs: Google Ads → connector → your AI client → the model.

| Party | Sees | Retains |
| --- | --- | --- |
| Connector (AdCopilot) | What each tool call returns | Your name and email, one encrypted refresh token, the connection record, a ledger entry per call, the workspace's business profile; audit reads cached for up to an hour, a saved audit's findings for 180 days |
| AI vendor (OpenAI, Anthropic) | The whole conversation, tool results included | Per your plan's retention and training policy |
| The model | The conversation in its context window | Nothing between conversations, unless the vendor trains on your tier |

Each row deserves its own honest paragraph. One flow sits beside this
pipeline rather than in it: AdCopilot's own use of an AI vendor, Anthropic,
as its sub-processor, which the next section describes.

## What flows through the connector — and what it retains

When you ask "which search terms wasted money last month", the connector
calls Google's API with your OAuth grant, receives the rows, and hands
them to your AI client. What AdCopilot keeps is bounded, and the full
inventory, with how long each record is kept, is on the
[privacy policy](https://adcopilot.cloud/privacy):

- **Who you are**: the name and email Google confirms when you sign in,
  so a connector is tied to the account that authorised it.
- **One credential**: an encrypted Google
  [refresh token](https://adcopilot.cloud/glossary/refresh-token) — the thing that lets it act
  on your OAuth grant. Never your password; no party in this pipeline
  ever holds your password.
- **Your connection record**: which AI-client connection is yours.
- **A ledger entry per call** — an [audit trail](https://adcopilot.cloud/glossary/audit-trail):
  which tool, which account, success or refusal, the reason, the
  timestamp. The record that access happened — not the rows that came
  back. Kept for 180 days.
- **Your workspace's business profile**: the business details and budget
  you give it, the first campaign it built, a log of decisions and a note
  on the last session, so the next conversation starts where the last one
  stopped. Kept until the workspace owner deletes it.
- **A short-lived cache**: an audit's reads, held for up to an hour, and
  the account summary your assistant opens with, for 10 minutes, so a
  repeat does not ask Google twice.
- **Saved audit results**: when an audit's results are saved to your
  dashboard, each finding keeps its description, the changes it suggests
  and up to 10 example rows, such as search terms, keywords, ad group and
  campaign names, landing-page URLs, cost and conversions. Kept for 180
  days, then deleted.
- **The server's system log**: each Google Ads request's account ID and
  the text of the query that ran, and the full content of any request
  Google rejected, such as the keywords or ad text it tried to set. Kept
  until the server's routine log rotation removes it.

The other rows your assistant reads pass through live and are not kept
by the connector once the response completes. The [security page](https://adcopilot.cloud/security)
states this retention posture as a commitment, alongside scoping and
revocation.

AdCopilot also uses AI models from Anthropic on its own side, apart from
your conversation. Its staff analyse aggregated statistics across
workspaces with them to improve the audit's checks: the statistics name no
person, workspace, account, campaign, keyword or search term, hold no figure
about fewer than five accounts, and are all Anthropic receives for that
work. The same models are the staff's working tool, and they read Google
data only when someone investigates a bug, abuse or a problem you asked
about, such as the system log entries of a failed audit. Anthropic
processes that data for AdCopilot as a sub-processor, its commercial terms
do not permit it to train models on it, and it deletes it within 30 days,
keeping it longer only where those terms allow or the law requires it. A
workspace is left out of the statistics until its owner, and anyone who
connected a Google account to it, accepts the privacy policy, and from the
day its owner switches them off in Settings. The [privacy policy](https://adcopilot.cloud/privacy#google) states each of
these as a commitment.

## What the AI vendor sees: conversations, under plan rules

The larger share of the privacy question lives one hop downstream, and it
is the hop people forget. Everything the agent read into your
conversation — the tables, the search terms, the client name you typed —
sits in your chat history with OpenAI or Anthropic, governed by the plan
you hold, not by the connector.

Both vendors publish the rules. OpenAI describes consumer-side usage in
[how your data is used](https://openai.com/policies/how-your-data-is-used-to-improve-model-performance/)
and the business-side defaults in its
[enterprise privacy commitments](https://openai.com/enterprise-privacy/).
Anthropic's are in the
[consumer terms update](https://www.anthropic.com/news/updates-to-our-consumer-terms)
and its [training policy article](https://privacy.claude.com/en/articles/10023580-is-my-data-used-for-model-training).
Policies move; the links above are the current word, and this page was
checked against them in August 2026.

## Training: what the majors say, by plan type

The pattern, as of this writing, is consistent across both vendors:

- **Consumer plans** (ChatGPT Free/Plus, Claude Free/Pro/Max): whether
  conversations train models **rides on a toggle** in data or privacy
  settings — offered at sign-up, changeable after, and worth checking
  before the first connected session.
- **Work and API tiers** (ChatGPT Business/Enterprise, Claude Team/
  Enterprise, both APIs): **excluded from training by default**, with
  contractual retention terms.

The practical rule falls out directly: if your ads data is commercially
sensitive or belongs to clients, run agent work on a work-tier plan, or
set the consumer opt-out before connecting anything.

## What never leaves Google at all

Some worries can be closed completely:

- **Your Google password.** OAuth exists so that it never transfers —
  you authenticate with Google, and Google issues the connector a scoped,
  revocable token.
- **Payment details.** AdCopilot exposes no billing tools, so cards and
  payment profiles are simply outside what any conversation can reach.
- **Anything outside the scopes you granted.** Each Google product is
  connected on its own grant — Google Ads, and Analytics, Search Console or
  Tag Manager only if you connect them — not Gmail, not Drive. Scope is
  enforced by Google, not promised by the vendor.
- **Accounts you exclude.** A connector can be
  [scoped to specific customer IDs](https://adcopilot.cloud/questions/can-claude-access-google-ads),
  making off-limits accounts unreachable rather than merely unmentioned.

## Reducing exposure: the four controls worth using

1. **Choose the plan deliberately.** The AI vendor's tier is the biggest
   single variable in this whole map — set training and retention there.
2. **Scope the connector.** Limit it to the accounts the work needs;
   agencies can have support hold a member to named accounts, with each
   seat individually revocable.
3. **Keep identifiers out of prompts.** The agent reads account data by
   itself — you rarely need to paste customer lists or personal data into
   the chat, and the habit of not doing so costs nothing.
4. **Revoke when idle.** Access withdraws in one click from your Google
   account's security page, and a reconnect restores the same seat later.
   A connector that is off sees nothing.

Mapped this way, the question stops being "is it safe" and becomes three
smaller, answerable ones — what does each party hold, under what rule,
controlled by which switch. The connector's own answers, stated as
commitments rather than defaults, are on the
[security and data page](https://adcopilot.cloud/security).

## Questions

### Will my ads data be used to train ChatGPT or Claude?

Plan-dependent, and both vendors publish the rules. OpenAI may use consumer ChatGPT conversations for training unless you opt out in data controls; business, Enterprise and API traffic is excluded by default. Anthropic, since its August 2025 consumer-terms update, trains on Free, Pro and Max chats only when the model-improvement setting is on — you choose at sign-up and can change it later; Team, Enterprise and API are excluded. On work plans with default settings, the practical answer is no.

### Does the connector store my performance data?

Some of it, for a bounded time. AdCopilot reads your account live through Google's API and passes results to your AI client. An audit's reads are cached for up to an hour, and a saved audit's findings — up to 10 example rows each, such as search terms, cost and conversions — are kept for 180 days, then deleted. It also keeps one encrypted Google refresh token, an audit trail of each call's tool, account, outcome and time, and your workspace's business profile. The privacy policy lists every record and its period.

### Can I use this under an NDA or client confidentiality agreement?

Treat the AI vendor's plan as the deciding variable, because the conversation is where client data lives. A work-tier plan with training excluded and short retention typically fits confidentiality obligations the way any SaaS processor does — but that is your call and possibly your client's. Scoping the connector to only the permitted accounts keeps the boundary mechanical rather than behavioural.

---

This is the markdown copy of https://adcopilot.cloud/questions/what-data-does-the-ai-see-google-ads. The index of the site's pages is https://adcopilot.cloud/llms.txt.
