−35% cost per click. −83% cost per order. ROAS more than 4×.
A direct-to-consumer brand in a strictly regulated product category, advertising across the DACH region and the wider EU, switched its Google Ads account to AI-driven management with AdPilot. Every change was suggested by the AI and signed off by a person. Here is what happened in the first 8 days.
This was measured in July 2026, when the agent still ran inside AdPilot. AdPilot now works as a connector for your own Claude or ChatGPT, and gives that AI the same playbook and Google Ads tools. The figures below are a past result of that approach on one account, not a measurement of the current setup.
Live shop: tongkatali1.com. The advertised business is publicly viewable, shared with the owner’s consent.
−35%
Cost per click
average CPC
−83%
Cost per order
CPA
+289%
Conversion rate
click → purchase
4×+
Return on ad spend
more than quadrupled (7.3× vs. the prior quarter)
+23%
Average order value
per order
More orders in 8 days than in the entire quarter before.
The starting point
A manually managed account with structural problems: keywords spread too broadly, the wrong conversion signal (optimizing for add-to-cart instead of actual purchases), unchecked click prices, and the policy constraints of a tightly regulated product category. The result had been months of unprofitable advertising.
What the AI agent did
From day one, an AI agent did the full campaign management through the Google Ads API, with a person signing off each change:
- Rebuilt and ran 13 campaigns across 5 country markets via the Google Ads API
- Fixed the conversion tracking (optimizing for purchases, not add-to-cart)
- ~150 purchase-intent keywords from data-driven research in 5 languages
- ~190 negative keywords to stop wasted spend
- Rolled out ad assets (sitelinks, callouts, snippets, promotions, prices) across every campaign
- Optimized search terms, devices and countries every day on real purchase data
How this was measured
Every figure above comes from the account’s own Google Ads reporting, not from our dashboard and not from an estimate. The comparison is the 12 weeks before the switch against the first 8 days after it, and the conversion counted is a completed purchase. Not an add-to-cart, not a page view, not a “conversion” Google modelled. That distinction matters more than it sounds: the account had been optimizing toward add-to-cart, which is exactly why it could look busy and still lose money.
Eight days is a short window, and short windows flatter whoever picks them. It is the window because it is when the account was measured, not because it was the best one available. You can run the same before-and-after arithmetic on your own account: AdPilot logs every change with its date and the old and the new value, and your AI can compare the same window before and after it.
What this case study does not prove
- The landing page and the pricing were changed during the same period. Some of the lift belongs to them, and this account cannot separate the two.
- It is one account, in one category, in one set of markets. One account is an existence proof, not a rate.
- The starting point was a badly broken account. There is more room in a broken account than a healthy one, and the percentages reflect that.
- Eight days is long enough to see a direction and too short to call it a trend.
What it does show is what the work consists of, and that it can be done without an agency retainer. Whether your own account has this much room in it depends on what is wrong with it today. The industry guides go through the usual answers trade by trade, and the audit checklist lets you check yours before you pay anyone, us included.
Questions about this case study
Is this a typical result?
No, and it is not offered as one. It is one account that started badly broken, measured over eight days, in a category with unusual policy constraints. A healthy account has far less room in it, and the percentages would be correspondingly smaller. Advertising outcomes depend on the product, the market, the starting point and the budget.
Who is the advertiser, and can I check it?
The shop is tongkatali1.com, published with the owner's consent, and it is publicly viewable. The Google Ads data behind the figures is the account's own reporting rather than ours.
Did the AI make these changes on its own?
No. Every change was suggested by the agent and signed off by a person before it was written to the account. Nothing in this account was applied automatically. At the time, in July 2026, the agent ran inside AdPilot. Today AdPilot gives the same playbook and Google Ads tools to your own Claude or ChatGPT, and that app asks you before every change.
Why compare eight days against a whole quarter?
Because that is when the account was measured: the switch happened, and eight days later the numbers were read against the period immediately before it. A longer post-period would be a stronger comparison, and this page will say so when there is one.
The landing page changed too. So how much of this is the ads?
This account cannot separate them, and we will not pretend otherwise. The pricing and the landing page were worked on during the same window, so some of the lift belongs to them. The campaign work is described above in full so you can judge it on its own terms.
Do you guarantee results like this?
No. Nothing on this page is a promise about your account. The one guarantee AdPilot offers is attached to the done-for-you Takeoff plan and is written into the terms with its own measurement definition. It is not this case study.
Individual result for the Google Ads account of tongkatali1.com, published with the account owner’s consent. Comparison period: the 12 weeks before the switch vs. the first 8 days of AI management (July 2026), with the agent then built into AdPilot. Basis: the account’s own Google Ads data, purchase conversions only. During the switch, the landing page and pricing were optimized alongside campaign management. Results are not guaranteed and vary by account, industry and starting point.
Want to see what your own account has room for?
Connect your Google Ads account, add AdPilot to Claude or ChatGPT, and ask your AI where your budget is going and how to do better. It asks before every change.
Get started free7 days free · then $79 per month · no change without your yes