Skip to content
Praion
AEO · STRATEGY

How to measure AEO ROI

Praion team · 14 min read

Published

QUICK ANSWER

AEO ROI isn't a single position in a list: since there's no stable ranking, you track different forms of presence across a fixed set of questions and connect them with identifiable AI traffic, conversions, economic value and total cost. In practice: (attributed gross profit − AEO cost) ÷ AEO cost × 100. You measure from the start with a baseline; if you can't attribute value and cost consistently, call it performance or contribution, not ROI.

You put time and money into AEO, but when you open your traffic and performance data, no single position is climbing.

There's no clear "you're third for this phrase," and no single line that shows you moving up.

And the basic question stays:

Is AEO paying off, or are you just spending time and money?

AEO can be measured. What it takes is more than one ranking or one visit count.

You need to connect the whole path:

  1. How often your business appears in relevant answers.
  2. How many people reach your website from AI tools, when the source of that visit can be identified.
  3. What they do once they arrive.
  4. What economic value they create.
  5. What the work that produced those results actually cost.

The first three show whether your presence is growing and whether interest is being created.

Real ROI begins when you add the economic value and the total cost.

Let's look at it in practice.

Why AEO isn't measured like SEO

In ChatGPT and Perplexity there's no stable ranking that works the way a traditional page of organic results does. This is where AEO and the way it differs from SEO start to diverge in how you measure them.

Presence isn't simply a matter of appearing or not appearing.

A business can:

  • be named directly,
  • be presented as one of several options,
  • appear first or last in a list,
  • have its site used as a source,
  • be mentioned through a third party with no link back to its own site,
  • appear in a positive, neutral or negative context.

So you don't track a single position.

You track different forms of presence across a fixed set of questions.

For Google's AI features there is now an official data source as well. As of 31 August 2026, Google says the Search Console Generative AI performance report has rolled out worldwide. It shows how often your site's pages appear in AI Overviews and AI Mode. If the report doesn't appear for a property, one possible reason is that the site doesn't have enough generative-AI impressions to show data.

This means measuring AEO doesn't replace your existing search and analytics tools.

It adds new signals to them:

  • mentions of the business,
  • recommendations of it as a possible option,
  • use of the site as a source,
  • links to pages that support the answer,
  • overall visibility inside the answers AI tools give.

How do you check whether AI tools mention your business?

You start with a presence audit.

It's a repeatable check of how your business shows up in real customer questions.

You build a fixed list of complete questions. For an initial audit, ten to fifteen questions is a practical starting point, not a required number. You don't rely on isolated keywords alone.

For example:

Which accounting firm handles startups in Thessaloniki?

Then you run the same question:

  • in ChatGPT Search,
  • in Perplexity,
  • in Google AI Mode,
  • in classic Google Search, noting whether an AI Overview appeared.

So you can compare results from one month to the next, you keep as stable as possible:

  • the exact question,
  • the language,
  • the country and location,
  • the platform and feature you use,
  • whether you're signed in or signed out,
  • the date of the test.

A single test isn't enough.

Answers can change. They can also be influenced by general location and, in some cases, by the user's history or settings.

For each key question it's better to run more than one test and record the trend, not one isolated answer.

In each check you record at least:

  1. Mention: does the business name appear?
  2. Recommendation: is it presented as a possible solution or option?
  3. Use as a source: is one of your site's pages used to support the answer?
  4. Source page: which page exactly was used?
  5. Competitors: which other businesses or sources appeared?
  6. Context: is the mention positive, neutral or negative?

Because each question is checked more than once, you need to separate two different measures.

Query coverage

Questions where the business appeared at least once
÷ Total distinct questions
× 100

Mention frequency

Tests where the business appeared
÷ Total tests
× 100

The first measure shows how many distinct questions your business appears for.

The second shows how often you appear when the same questions are repeated.

A presence audit isn't final proof of performance. It's a comparable visibility signal.

Every question where you don't appear is a point to investigate.

The cause might be:

  • a content gap,
  • a technical issue that stops search engines from finding or indexing the site properly,
  • low relevance,
  • inconsistent business information,
  • a lack of independent references,
  • a different intent behind the question.

The audit doesn't automatically tell you "write one more page."

It shows you where to look for the cause.

Can you see AI traffic in your analytics?

Part of it can be recorded.

Not all of it, and not always in the same way.

A useful measurement framework separates three categories.

1. Identifiable AI traffic in GA4

When a link carries source information, you can see in GA4 which source and channel the visitor came through.

ChatGPT automatically adds the parameter utm_source=chatgpt.com to the links in its search results, so that traffic can be tracked.

Other platforms may also pass information that shows where the visit came from.

GA4 now also includes a default AI Assistant channel in its default channel group. Google uses it for identifiable traffic from recognised sources such as ChatGPT, Gemini, DeepSeek, Copilot and Grok. Google's own AI Overviews and AI Mode are excluded from this channel and are measured in Search Console.

If you need a different grouping, or want to include additional sources that aren't classified the way you need, you can still create a custom channel group.

A custom channel group organises source information that has already been recorded. It can't discover the true origin of a visit when that information doesn't exist.

2. Visibility in Google's AI features

For AI Overviews and AI Mode you use the Search Console Generative AI performance report. Google says that, as of 31 August 2026, the report has rolled out worldwide, although a property may still not show it when there aren't enough relevant impressions.

The report can show:

  • impressions,
  • pages that appeared,
  • countries,
  • devices,
  • changes over time.

This is the official source for your site's visibility in Google's AI features.

You shouldn't try to infer that presence from GA4 alone.

3. Traffic with no identifiable source

GA4 classifies as (direct) / (none) any traffic for which there's no clear source information.

Some direct visits can come from environments that didn't pass referral or UTM data, such as:

  • private apps,
  • copied links,
  • apps that don't carry source information.

In other cases, a user may have discovered the business through AI and visited the site later by typing the address directly.

The later visit may leave no new identifiable signal showing AI's earlier influence. Depending on the report, scope and available history, GA4 may show it as direct or retain a previously recognised source.

GA4 can't separate or prove these cases after the fact.

Neither the default AI Assistant channel nor a custom channel group can take a genuinely direct visit and discover after the fact that it came from AI.

They can only organise visits that already carry identifiable source information.

So your measurement has to keep three things clearly apart:

  • identifiable AI traffic,
  • visibility in Google's AI features via Search Console,
  • traffic of unknown source.

Does AI traffic convert better?

There's no rule that holds for every business.

Some studies and vendor data have recorded a higher conversion rate or greater economic value for identifiable AI traffic.

The results, though, vary widely depending on:

  • the site,
  • the sector,
  • the period,
  • what counts as a conversion,
  • how the conversion is attributed across channels.

One likely explanation is that the user already has more context before clicking.

They may have:

  • read a comparison,
  • narrowed their options,
  • understood better what they're looking for.

That can raise the quality of some visits.

It shouldn't be treated as a given.

You check it against your own business data.

Compare identifiable AI traffic with other channels on:

  • conversion rate,
  • share of quality leads,
  • average order value,
  • revenue or gross profit per visit,
  • time to conversion,
  • returns and cancellations,
  • customer quality.

You also need care in how you judge which channels contributed to a conversion. That process is called attribution.

A user might discover the business through AI, leave without clicking, and later return by searching the business name or typing the address directly.

If the first touch left no identifiable signal, no attribution model can pin it down with certainty after the fact.

A multi-touch attribution model only helps when the different touches were recorded and can be linked to the same journey.

So you combine your analytics data with:

  • a CRM field where the source of each lead is recorded,
  • a "How did you find us?" question in the form or after purchase,
  • data on the user's first source and the source of the specific visit,
  • GA4's path and attribution reports, which help you see the different touchpoints before a key action and how value is distributed across channels,
  • trends in searches for the business name,
  • Search Console data,
  • sales and lead-quality figures.

You don't need a perfect attribution model.

You need enough independent signals so you're not relying on the last click alone.

When does it make sense to draw a conclusion?

Measurement starts before the changes, not two or three months later.

You need a baseline so you know:

  • how many questions you appeared in before,
  • which pages were used as sources,
  • how many impressions there were in Google's AI features,
  • how much identifiable AI traffic was coming in,
  • how many conversions and what economic value were created,
  • what the starting cost was.

From day one, you record data.

That doesn't mean you draw a final conclusion from the first month.

There's no universal rule of:

  • 90 days,
  • 3–6 months,
  • 6–12 months.

The time depends on:

  • your existing organic and commercial presence,
  • how the site is crawled and indexed,
  • the language and market,
  • how often the relevant questions are asked,
  • the number of conversions,
  • the type and scope of the changes,
  • the third-party sources that mention the business.

The practical rule is different:

You measure from the start, watch the trend each month, and wait for a large enough sample before you connect a financial result to AEO.

On a site with a small volume of data, it can be more useful to look at the last three months each time, rather than compare two single months.

On a high-volume site, you can spot changes earlier.

The point is not to confuse insufficient data with an absence of results.

How you actually calculate AEO ROI

Presence, impressions, visits and conversions are not ROI on their own.

To calculate return on investment, you need:

  1. the economic value that can, based on the available data, be connected to AEO,
  2. the total cost of the investment.

A practical formula is:

AEO ROI =
(Gross profit attributed to AEO − Total AEO cost)
÷ Total AEO cost
× 100

By "gross profit attributed to AEO" we mean the part of gross profit that can, based on the available data and the attribution method you use, be defensibly connected to your AEO work.

For financial evaluation, gross profit is generally more useful than revenue alone.

Revenue isn't the same as profit. Production, purchasing and service costs can materially reduce the value that remains with the business, so using revenue alone can overstate the economic return.

What AEO cost includes

  • initial audit and research,
  • content production or revision,
  • technical changes,
  • tracking tools,
  • internal team hours,
  • external partners,
  • ongoing maintenance,
  • performance reporting.

How you value the economic outcome

For e-commerce you can use the gross profit from purchases connected to identifiable AI traffic, or to a recorded customer journey that an AI tool contributed to.

For lead-based businesses, you can estimate:

Lead value =
Quality leads
× Close rate
× Average gross profit per new customer

By "quality leads" we mean the contacts that fit the target audience and have a real chance of becoming customers.

Example:

  • 20 quality leads connected to customer journeys an AI tool contributed to,
  • a 20% close rate,
  • 1,000 euros average gross profit per customer.
20 × 20% × 1,000 € = 4,000 € estimated value

If the total AEO cost was 2,500 euros:

Estimated ROI connected to AEO =
(4,000 € − 2,500 €)
÷ 2,500 €
× 100
= 60%

The 4,000-euro value rests on an expected close rate, not solely on gross profit that has already been realised.

That's why the 60% is an estimated ROI, not a final financial result.

You can calculate ROI based on realised financial outcomes once you know:

  • how many leads became customers,
  • what gross profit was realised,
  • what the total AEO cost was.

Even then, the result doesn't prove that all the value was caused by AEO alone.

For a stronger estimate you need:

  • a comparison with the baseline,
  • historical data,
  • controlled changes,
  • another method that helps separate the added effect of AEO from the natural course of the business.

If you can't yet attribute value and cost consistently, don't call the result ROI.

Present it as AEO's performance or contribution.

What you track in the end

You need a small scoreboard on four levels.

1. Presence

  • query coverage for mention, recommendation and use of the site as a source,
  • mention, recommendation and source-use frequency across all tests,
  • pages used as sources,
  • competitors that appear.

2. Visibility and traffic

  • impressions in Google's AI features via Search Console,
  • pages and countries where you appeared,
  • identifiable AI visits in GA4,
  • new users,
  • engagement of visitors from these sources.

3. Business outcome

  • conversions,
  • quality leads,
  • conversion rate,
  • purchases or bookings,
  • economic value of conversions,
  • average gross profit,
  • time to conversion.

4. Financial return

  • total AEO cost,
  • revenue that can be connected to AEO,
  • gross profit that can be connected to AEO,
  • cost per quality lead,
  • customer acquisition cost,
  • ROI that can be connected to AEO.

You track them each month, but you don't read them in isolation.

You don't assume that more presence automatically brings more work.

You look at whether:

  1. presence is growing,
  2. more relevant visits follow,
  3. those visits connect to more or higher-value conversions,
  4. the financial outcome exceeds the total cost.

If in March you appeared in four of your fifteen key questions and in June in nine, that's an improvement in visibility.

If identifiable AI visits and quality leads are rising alongside it, you have a sign of commercial contribution.

Only when you can defensibly connect the economic value to AEO and subtract the relevant cost can you talk about ROI.

That's what AEO ROI looks like in practice:

not a position in a list, but a documented path from presence to profit.


Closing

You keep a fixed set of questions, separate identifiable traffic from visits of unknown source, and combine data from Search Console, GA4 and your CRM.

The presence audit can start with a simple logging sheet.

GA4's default AI Assistant channel organises part of the identifiable traffic from AI assistants; where you need a different grouping, you can also use a custom channel group.

Real ROI, though, needs:

  • the economic value of conversions,
  • the total cost,
  • a careful read of which channels contributed to each conversion.

This is where Praion can help, organising both the work and its measurement from the start.

We set up measurement alongside the content from the beginning, so you know which data you need before you start judging results.

If you want to see where your business stands today, an AEO Audit can show:

  • which questions you appear in,
  • which sources are used,
  • where the gaps that need investigation are.

Let's talk. · See pricing

AEO describes work aimed at improving a business's presence in the answers AI tools give. That work can create better conditions for visibility, but it doesn't guarantee a mention, a ranking, a conversion or a financial result.

FREQUENTLY ASKED

Frequently asked questions

You track different forms of presence across a fixed set of questions: mentions, recommendations, use of your site as a source, and links to pages that support the answer. Then you connect that presence with identifiable AI traffic, conversions, economic value and total cost. One ranking or one visit count doesn't tell you whether AEO is paying off.

It's a repeatable check of how your business shows up in real customer questions. You use fixed questions and conditions, run more than one test, and record separately the mention, the recommendation, the use of your site as a source, the source page, the competitors and the context of the mention. You watch the trend, not one answer.

No. GA4 can record and group visits when there is information showing their source, such as a source, referral or UTM parameter. The default AI Assistant channel and custom channel groups can only organise source data that has already been recorded; they cannot discover the true origin of a visit logged as direct when no other signal is available. Part of AI's influence can therefore remain invisible.

Some studies have recorded a higher conversion rate for identifiable AI traffic, but this isn't a general rule. The result varies by site, sector, how you define a conversion and how you attribute it across channels. The right comparison uses your own business data, set against your other channels, rather than an industry average.

Before the changes, so you create a baseline. From day one you record presence, visibility, traffic, conversions, economic value and cost. There's no universal deadline for results; you watch the monthly trend and wait for a large enough sample before you draw a financial conclusion. Don't confuse too little data with no result.

You take the economic value that can, based on the available data, be connected to AEO, subtract the total cost of the investment, and divide the difference by that cost. Ideally you use gross profit, not revenue. If you can't attribute value and cost consistently, present it as AEO's performance or contribution, not ROI.

Want to see whether AEO isdelivering for your business?

We start by identifying where your business appears today, which data you can reliably measure, and where the gaps are. That way, your AEO efforts are tied from the start to visibility, conversions, and real business value.