How to Measure the Success of Generative Engine Optimization Campaigns

Since digital marketing became a thing, we’ve all been obsessed with measurement.

Unfortunately, measurement has been getting harder instead of easier, which is frustrating for marketers everywhere.

And now there’s this thing called Generative Engine Optimization, and measurement seems next to impossible.

So let’s break down the best ways to measure your GEO efforts.

Step 1: Set up custom channels and segments for LLMs

What you are measuring: Website traffic and leads/revenue coming from LLMs

Success measures: Consistent increases in website traffic and leads from LLMs

Bonus measurement: Look at landing pages from LLMs to predict the user queries

Generative Engine Optimization improves your visibility in AI-driven search engines like ChatGPT, Perplexity, Claude, Gemini, and other AI chatbots.

Sometimes, these tools will provide links, and fortunately, these links are attributed to the chatbots when they flow into your analytics tools.

For this tutorial, we’ll assume you are using Google Analytics 4.

There are two steps you’ll want to take to set up your tracking for these LLM tools.

First, let’s set up a Custom Channel for LLMs.

How to set up a custom channel for AI-driven search engines

If you haven’t already set up a Custom Channel Grouping in GA4, you’ll need to create a new one. You can do this by going to Admin > Channel groups (under Data Display in your Admin panel). Click on the big blue button for “Create new channel group.”

If you already have a custom channel grouping, it’s up to you if you want to create a new one or just add a new channel within your existing grouping.

Whichever you decide, your next step is to click “Add new channel” (again, it’s the big blue button).

Name your channel something like LLMs or AI Chatbots.

Now it’s time to set the conditions.

• Click on +Add condition group.
• Select “Source” for your condition.
• Click on “Add condition’ and choose “matches regex” and then paste this regex string:

.*chatgpt.*|.*perplexity.*|.*\.openai.*|.*copilot.*|.*gemini.*|.*gpt.*|.*claude.*|.*anthropic.*|.*bard.*|.*llama.*|.*alpaca.*|.*bing&s?chat.*|.*cohere.*|.*davinci.*|.*youchat.*|.*chatglm.*|.*falcon.*|.*stablelm.*|.*mpt.*|.*bloom.*|.*ernie.*|.*characterai.*|.*baichuan.*|.*wizardlm.*|.*pangu.*|.*chatsonic.*|.*phind.*|.*palm.*|.*flan.*|.*blenderbot.*|.*wizardcoder.*| ^.*ai|.*gemini.*google.*$|.*google.*bard.*|.*bard.*google.*|.*bard.*

This should cover every known LLM that might be sending you traffic (and some that probably don’t).

Obviously you’ll have to come back and update this channel group if there are new LLMs introduced in the future (there will be).

After you’ve done that, click the blue button to save your channel.

Now you have to reorder your channels. If you don’t, you’ll never get any traffic reporting in your new channel because it will all go to referral (GA4 processes channels in order).

So click the white Reorder button, and then drag your new LLM channel right above your Referral channel. This will make sure Google processes your LLM channel before your referral channel.

After you’ve done this, you need to go all the way to the top of the page and click the “Save group” button.

And now you have your custom channel all set up.

Thankfully, custom channel groupings are retroactive, so all your LLM traffic should now be assigned to this channel.

Let’s check it by going to your Traffic Acquisition report and selecting your new Custom Channel Grouping from the dropdown.

You should see your LLM channel listed in your report.

If you don’t see it, then there are two possibilities:

1. You set it up incorrectly
2. You haven’t received any traffic from LLMs yet

This gives you access to your AI chatbot traffic in a standard report, but let’s also get a custom exploration set up with an LLM segment so you can do more with your data analysis.

How to set up a custom exploration and segment for AI chatbot traffic

If you want to get even more out of your AI search reports, then you need to set up an exploration with a custom segment.

To do this, simply go to the Explorations tab and then create a blank free-form exploration.

Create a new segment for your AI Traffic (use the same rules and regex from the channel group you created before).

Add the dimensions and metrics you want to explore, and then populate your explorations.

I recommend two separate visualizations in your exploration:

1. A table that shows sessions and key events broken down by LLM source and landing page
2. A time series that shows weekly growth (or decline) of LLM traffic

Look at these reports at least once a week to see where your LLM traffic is coming from, what the main landing pages are, and which LLMs are driving conversions. This will help inform your ongoing GEO strategy so you can make sure you are optimizing for the right chatbots and creating the best content.

Step 2: Establish benchmarks for your direct traffic and branded search volume

What you are measuring: The amount of direct traffic to your website and the number of times people search for your brand each month on traditional search engines like Google

Success measures: Increases in direct traffic and branded search volume can both suggest GEO success

Unfortunately, AI-driven search engines don’t always provide links. And when they do, users don’t always click them.

Here’s what happens most of the time.

ChatGPT will mention your brand as an option for whatever your target customer is searching.

But ChatGPT either won’t link to you or the user will choose not to click the link.

Instead, the user will open a new tab and either go directly to your site by typing in your URL, or they’ll go to Google and search for your brand.

In both cases, ChatGPT obviously won’t get the credit because how would GA4 or any other analytics tool know how they really found you?

And this is one of the biggest holes in measurement today. Analytics tools only capture the exact click the user made to get to your site, not the complex journey they took to learn about you in the first place.

Which leads to all kinds of bad marketing decisions because we couldn’t attribute the exact revenue to the exact channel, and so we stopped doing stuff that was actually working and invested more into channels that were swooping in and stealing credit at the very end.

But that’s not really what this post is about, so let’s get back to setting those benchmarks.

There are two main things you want to benchmark to help with understanding the impact of GEO:

1. Your branded search volume
2. Your direct traffic

Why branded search volume? Well, if you see an uptick in branded searches, then it’s a good sign some other marketing effort is working. Measure that volume against when you started GEO and you’ll get a good sense of the impact. Just keep in mind any other efforts you started around that time.

Why direct traffic? Because many users who hear about you on ChatGPT will go directly to your website. If you start to see an increase in direct traffic, this could be a sign your GEO is working. Of course, the same caveat as before applies here. Other efforts will influence this as well.

Finally, you should be looking at overall business metrics, like revenue, sales, qualified leads, etc. If all of those are going up, then that’s another good sign GEO is working positively for you.

Step 3: Collect self-reported and sales-collected attribution and log it in your CRM

What you are measuring: The number of times people mention an LLM as the source of how they found out about you

Success measures: Increases in the number of people mentioning LLMs as the way they heard about you

Not enough companies are collecting self-reported attribution. Not enough companies ask “how did you hear about us” during the sales process.

These should not be optional questions.

If you have a lead gen form, you should have an open-ended response field for “how did you hear about us.”

I strongly recommend keeping this open-ended as opposed to providing a list of options. People will always have the tendency to select the first option, especially if the real way they found you isn’t listed.

Immediate action items for you:

1. Add a “How did you hear about us” field to all your lead gen forms
2. Add a field in your CRM for how a prospect heard about you and require your sales team to fill this out for every single prospect that comes in the door

You’ll be amazed at the detailed responses you’ll get from people. Seriously. Many of them will be excited to tell you they first heard about you in a Facebook group and then asked ChatGPT about your brand.

But you have to ask! If you don’t, the analytics report and the CRM will give way too much credit to direct and organic.

Once you collect this information, then you can run reports from your CRM that show how many people are coming from LLMs.

Those reports will be beautiful and they will show your GEO efforts are crushing it.

Step 4: Monitor appearance in LLMs

What you are measuring: Whether or not you appear in LLMs and if the information is accurate

Success measures: You appear frequently in test searches and the information is all accurate

This is the least “measurable” part of your GEO measurement strategy, but it’s still an essential process.

You need to implement a regular cadence for monitoring how your brand appears in your preferred AI-driven engines. And when I say your preferred, I mean the ones your target audience is using the most.

I recommend doing this in two different ways:

1. On a monthly basis, ask ChatGPT (and other LLMs) to summarize your brand’s online presence, including an analysis of the pros and cons and any other sentiments that exist. We built a Brand Recommendation Analyzer Custom GPT that provides us with the exact information we want to see returned, which makes this process much easier. If you see any issues with the summary you receive, then you’ll want to find where that information exists and take corrective action. Our Custom GPT is programmed to return all the source information, so this really speeds up the process of finding any negative sentiment or incorrect information.
2. On a weekly basis, ask ChatGPT (and other AI chatbots) if it would recommend your brand for your service in your location (location is optional). Again, we have a Custom GPT that does this for us, our Brand Snapshot Analyzer GPT. The prompt is simply “Would you recommend [company name] for [service] in [location].” An example of this prompt in action is below:

Our Custom GPT is programmed to return the top three results for that service in that location. If the selected brand isn’t on there, then ChatGPT explains why it didn’t list it. Here’s an example of what that looks like:

There aren’t really metrics you’ll be tracking here, so this isn’t necessarily something you’ll put into an analytics report in a table or graph, but it’s still an important process to help you track if your GEO efforts are actually working.

Limitations with GEO measurement

Measuring your generative engine optimization efforts will not be perfect. If you are investing in GEO (which you should be), you do have to deal with some unknowns.

You won’t be able to get data for any of these things you’re used to seeing with Google:

• Search volume (how often keywords are searched on LLMs)
• Impressions (number of times you show up in LLM searches)
• Tracked rankings (where you appear in LLM responses)

Although you can track clicks, this will leave major gaps in the story simply because of how AI-driven search works.

Measure what you can. If you have the right measurements in place, you’ll know if it’s working for your business.

Final thoughts on GEO measurement

If you aren’t already planning your GEO strategy, now’s the time to start. Yes, Google still dominates, and GEO is a largely an unknown field. But ChatGPT and other AI-driven search engines are only going to increase in popularity, and the brands that adapt now are the ones who are going to win big.

You need to go into GEO with proper expectations. It won’t yield immediate results (although we’ve seen brands start appearing in ChatGPT recommendations within weeks). And it will be difficult to fully track the impact. That said, if you follow all the measurement protocols outlined here, you’ll be able to see the impact. Just know the actual impact will be larger than what the reports tell you.

Free GEO consultation

As a thank you for reading this post, I’m offering you a free GEO consultation. Schedule time with me to talk about how generative engine optimization might work for you.