# Give your ad tools better conversion signals.

Ad platforms optimise toward whatever you report as a conversion. Label each session as human, bot or agent, and keep automated events out of the signals your campaigns learn from.

For performance marketing teams and agencies running their own site tags.

Status: Available with Expose

## Your bidding learns from every conversion you send.

When a bot fills a form or completes a checkout step, that event can reach Meta, Google Ads or TikTok as a conversion. The campaign then looks for more visitors like it.

Expose labels the session and holds likely bot conversions back from supported ad tools. Agent-assisted purchases are tagged, not dropped, because a shopping agent can bring a real customer.

## Where $50,000 of ad spend went

- Ad spend: $50,000
- Reported conversions: 1,000
- Conversions from automated sessions: 180 (Bot and unwanted automation; agent-assisted purchases are kept by policy.)
- Qualified conversions: 820

Illustrative example. $50,000 spend, 1,000 reported conversions, 18% from automated sessions, $200 average order. Hypothetical.

## Same campaign, two views.

| | Reported | With evidence |
|---|---|---|
| Conversions | 1,000 | 820 |
| Cost per conversion | $50.00 | $60.98 |
| Return on ad spend | 4.0× | 3.3× |

Illustrative example.

## Your numbers

Qualified = reported × (1 − share). Real cost = spend ÷ qualified. Real ROAS = qualified × order value ÷ spend.

## One conversion, before it reaches Meta.

- Event: Purchase, $200
- Session label: bot · confidence high
- Policy: Hold back bot conversions
- Sent to Meta: No

The order still exists in your shop. Your ad tools simply don’t learn from it.

## How it fits

1. **Install the script.** Paste the snippet or use the CLI. No account needed to start.
2. **Check your tags.** Confirm supported tools (Meta, Google Ads, TikTok, GA4) are on the page.
3. **Choose agent policy.** Keep agent-assisted purchases, or treat them separately.
4. **Compare outcomes.** Claim the domain for HQ and compare qualified conversions over time.

## What your team can do

- **Hold back bot conversions.** Supported ad tools stop learning from likely bots.
- **Report on people.** GA4 and Shopify carry the label, so conversion rate is measured on people.
- **Keep real agent buyers.** Agent-assisted purchases are tagged, not dropped.

## Before you start

### Will this refund my past ad spend?

No. Holding back an event changes what your campaigns learn from now on; it does not recover past charges.

### Are AI agents bad traffic?

Not necessarily. An agent can be shopping for a real person. You decide how agent conversions are treated.

### What about sessions you cannot classify?

They stay unknown. An absent bot signal does not prove a person was there.

### Which tools are supported?

See the integrations list in the docs; generic retargeting suppression is not automatic everywhere.

[Install Expose](https://doubleagent.so/docs/)

- [Double Agent Expose](https://doubleagent.so/)
- [Install guide](https://doubleagent.so/docs/)
- [Meta integration](https://doubleagent.so/docs/meta/)
- [All use cases](https://doubleagent.so/use-cases/)