What to Track Once a Client Starts Running Ads on a New AI Platform

A client agrees to test advertising on a new AI platform. The campaign goes live, traffic starts arriving, and then comes the question every agency knows is coming: “So, is it working?”

That is where tracking new AI ad platforms gets complicated. The new advertising channels don't always roll out with the reporting capabilities that marketers are accustomed to seeing from older platforms. You might have some delivery numbers that are inside the platform, but you don't have enough information in order to confidently match those numbers to what people are doing after clicking.

The solution isn't to wait for perfect reporting. Agencies need an interim measurement system that captures what can be measured now without pretending that the gaps do not exist.

Start With Two Lists: What You Know and What You Don't

Before building a dashboard, separate available data from missing data.

A new platform might provide basic campaign information such as:

• Impressions
• Clicks
• Click-through rate
• Campaign delivery
• Conversions reported by the platform
• Basic audience or placement information

Exactly what is available will depend on the platform.

Now make a second list.

What would you typically expect from a well-established ad platform that you can't see yet?

Perhaps it's a detailed attribution. Maybe reporting cannot be broken down sufficiently by creative, placement, audience, or campaign. Additionally, conversions can be reported through the platform and on the client's website may differ.

The second list is as important as the first.

It prevents an agency from creating a performance narrative out of incomplete data.

Build Measurement From the Destination Backward

When native reporting is limited, the client's website becomes particularly useful.

Instead of asking only, “What does the AI advertising platform report?”, work backward from the action the client actually cares about.

For example:

• Ad

• Click

• Landing Page

• Engagement

• Key Action

The key action may be a form submission, account registration, purchase, demo request, or other action for the business that the client is already measuring.

This changes the conversation. You're no longer relying on one platform to tell the complete story of the customer journey.

For instance, using Google Analytics, sessions, users, engagement, landing-page activity and configured key events can be displayed. Its Landing Page report is also available along with the source/medium information to determine which pages were visited from which traffic sources.

A Simple Interim Measurement Map

It is not a goal to make all the numbers match in all sources. It is to understand what each measurement layer can actually tell you.

Make UTM Discipline Non-Negotiable

This is one of the least exciting parts of campaign management. It is also one of the easiest places to create reporting problems.

UTM parameters can be used to manually tag campaigns in Google Analytics. These parameters can be used to report on details including source, medium, campaign, campaign ID and ad content. Google recommends consistently setting the relevant parameters because incomplete tagging can produce (not set) values.

Before launching, create a naming convention.

For example, decide how your agency will write:

• Platform/source name
• Paid traffic medium
• Client campaign
• Campaign ID
• Creative or content variation

Then document it.

Don't allow one account manager to use “ai-paid”, another to use “AI_ads”, and somebody else to call the same traffic “paid-social”.

When multiple clients start playing with the same new channel it's especially important to be consistent.

Watch the Landing Page, Not Just the Click

A new channel can generate a lot of clicks, but if that channel is low-quality traffic, that's a problem.

This is why tracking new AI ad platforms should extend beyond the platform dashboard.

Observe visitor behavior once they're at your site.

Do they engage with the landing page? Do they continue deeper into the site? Are they triggering the key events that matter to the client?

A useful review might compare:

1. Sessions arriving from the new channel
2. Landing pages receiving that traffic
3. Engagement after arrival
4. Relevant key events
5. Performance differences between campaigns or creatives where identifiable

GA4 Landing Page report consists of metrics like sessions, active users, average session duration, and key events.

This provides an agency with an additional piece of proof where only a portion of evidence is provided by an ad platform.

Don't Force the Numbers to Match

Let's say the AI platform is reporting 120 conversions, and your analytics system is logging 87 relevant actions.

Which is right?

There might not be a simple answer to that question.

Various rules, identifiers, measurement windows, and methods can be applied to the platforms and the analytics systems. The real solution is to examine the distinction and not simply opt for a more appealing number.

Keep an easy discrepancy log.

Record:

• What the ad platform reports
• What website analytics reports
• The period being compared
• Any known measurement limitations
• Whether the difference is increasing or decreasing

This makes reporting more defensible.

A client can understand, “These systems currently measure the journey differently.” This is a far cry from offering one number as an irrefutable fact.

Set a Baseline Before Calling the Test a Success

You shouldn't look at a new platform in isolation.

Make comparisons to something that the agency is familiar with in terms of traffic and results.

That could mean comparing:

• Landing-page engagement
• Conversion rate
• Lead quality
• Repeat visits
• Device behavior
• Performance by campaign period

However, care should be taken with the direct comparisons. Users might be in very different stages of intent when they are attracted by a new AI environment or by a mature search advertising channel.

Do not use the baseline as a "winner vs. loser" test.

Our Google Ads Monitoring Tool offers campaign-level insights into a range of metrics including conversions, CPC, clicks, impressions and CTR. This level of reporting in mature channels can provide an agency with the context on how much information is still missing from a newer advertising environment.

Give Clients Three Levels of Confidence

Don't show all metrics, label what you know.

Confirmed

Data that is measurable and you can describe and explain, like tagged sessions landing on a landing page.

Indicative

Signals that indicate good performance and require corroborating evidence like high engagement from a relatively small test.

Unknown

Information the current setup or platform simply cannot establish reliably.

This approach makes tracking new AI ad platforms much easier to communicate because uncertainty becomes part of the report instead of something hidden in the footnotes.

It also safeguards the agency from claims they can't prove based on available data.

Keep the Reporting Setup Ready to Evolve

Today's workaround should not become next year's permanent reporting process.

As an AI advertising platform matures, revisit your measurement setup. Native reporting may improve. New integrations may become available. Attribution options may change.

The interim framework should therefore be documented well enough to modify.

Keep records of:

• UTM conventions
• Landing pages used
• Key events being measured
• Reporting gaps
• Client KPIs
• Data discrepancies
• Changes made during the test

That history becomes valuable when the platform gives agencies more measurement options later.

Report What You Can Defend

Being the first to a new advertising channel can also give you opportunities, but being first also means working imperfect information.

Agencies do not have to address all attribution issues prior to testing. They do require a measurement framework that distinguishes between reported performance on the platform and observed behavior on a website.

At DM Cockpit, we help agencies bring established campaign and analytics data into a clearer reporting environment. Our Google Analytics reporting and Google Ads monitoring capabilities can help teams keep familiar performance data visible while they evaluate newer channels. The simple objective is to make decisions based on what information you have and acknowledging that you don't have all the information.

Frequently Asked Questions

1. What should an agency track first on a new AI advertising platform?

Start with the reliable native metrics the platform actually provides, then track incoming website traffic, landing-page engagement, and relevant conversions or key events separately. This gives you both platform-side and website-side evidence.

2. Why are UTMs important when testing a new advertising channel?

UTM parameters help analytics tools identify where campaign traffic originated. They are particularly useful when a new advertising platform does not yet have the integrations or reporting depth available from more established channels.

3. What UTM parameters should agencies use?

Google recommends using relevant parameters consistently, particularly source, medium, campaign, campaign ID, and source platform. Agencies can also use content-related parameters where they help distinguish ads or creative variations.

4. Should platform conversions and GA4 conversions always match?

No. Different systems can measure and attribute activity differently. A mismatch should be investigated and explained rather than automatically treated as a tracking failure.

5. How long should an agency test a new AI advertising platform?

There is no universal test period. It depends on campaign volume, the client's objective, available audience, and how much evidence is needed to make a useful decision. Define the evaluation criteria before the test begins rather than choosing them after seeing the results.

6. What should agencies tell clients when reporting is incomplete?

Be specific about what is confirmed, what is only indicative, and what cannot currently be measured reliably. Clear limitations make an experimental-channel report more useful than filling reporting gaps with assumptions.