INTERNATIONAL RESEARCH / PREPRINT

Can we measure the sales impact of AI visibility?

An AI answer recommending your brand may bring you a new customer. But how can you tell? This paper proposes a way to measure the connection between appearing in an answer and making a sale, and explores it through computer simulations. It does not prove a sales increase in a real campaign. Our marketing recommendation at DijitalPi is to track brand mentions, website visits, enquiries and sales separately. This helps distinguish greater visibility from greater revenue when deciding where to spend your budget.

ACADEMIC WORK REVIEWED

Generative Marketing Mix Modeling: A Causal Inference Framework Linking GEO and GEM to Business Impact

ResearchersMasahiro Kato, Daiki Honma, Taka Kato.

arXiv · 10 September 2026 · v1 · Method and simulation preprint · Source language: English

Open the original publication Find on Google Scholar

ENGLISH REVIEW AND COMMENTARY: DIJITALPI

We explain the paper in plain English, discuss what it may mean for organisations and identify our own examples separately.

Read the explanation Go to 3 business examples

English review: 21 September 2026 · The source date appears in the citation above.

Editorial review draft

The source has been checked; the review still carries its original editorial status. DijitalPi did not conduct a customer experiment for this paper.

LET'S READ THE RESEARCH TOGETHER

We first explain the researchers' question, method and findings. We then discuss how to interpret the results, clearly separating DijitalPi's commentary from the source.

01 / WHAT DID THE RESEARCHERS WANT TO UNDERSTAND?

How can visibility be connected to business outcomes?

The paper proposes a framework for measuring business effects of GEO and sponsored AI visibility.

Explaining the context · DijitalPi commentary

DijitalPi commentary: Recognising a brand, visiting its site, requesting a quote and purchasing are separate events. Combining them into one success metric makes investment decisions harder to assess. Start by stating which stages you can actually observe.

02 / HOW WAS THE RESEARCH CONDUCTED?

How did they test the question?

Repeated model answers are combined with usage and notice assumptions; recommendation simulations use English and Japanese.

Understanding the method · DijitalPi commentary

Separate observed inputs from assumed ones. A recorded form submission is an observation; how many people noticed an AI answer needs additional measurement. Treating an unknown input as an exact number creates false precision.

03 / RESEARCH FINDINGS

A framework is not a revenue guarantee.

The method sets identification conditions for causal effects and is examined through simulations.

Source: GMMM v1: framework, assumptions and simulations ↗

04 / DIJITALPI'S EXPLANATION

How should we interpret these findings?

DijitalPi commentary: Add data sources and uncertainty to a visibility report before adding a larger success percentage. Record concurrent changes in pricing and campaigns so that the attribution claim can be questioned.

Budget decisions should consider lead quality and measurement coverage together. An unmeasured channel contribution is neither automatically zero nor a proven gain.

05 / CONCLUSION AND OPEN QUESTIONS

What did we learn, and what do we still not know?

This preprint does not report measured campaign sales uplift in Türkiye.

Our open question: which stages from visibility to sale can your business trace? The applications below prepare data; they are not results from an implemented GMMM model.

Return to the researchers' original publication

GMMM v1: framework, assumptions and simulations

DIJITALPI'S APPLICATION COMMENTARY

How can we use this in marketing?

We created these scenarios to make the topic concrete. They are not cases from the paper or measured client results.

EXAMPLE 01

Agency reporting: more mentions

A report shows more mentions, but the questions and model changed this month.

Are those months directly comparable?

Open the recommendation for example 01

Keep measurement conditions consistent.

Record question set, language, model version, search settings and repetitions. Report old and new question sets separately. Distinguish brand mentions from linked citations and explain which changes might reflect the measurement process itself.

EXAMPLE 02

Ecommerce: traffic versus orders

Visits classified as AI referrals rise, but visitor needs are unknown.

Does more traffic necessarily mean more customers?

Open the recommendation for example 02

Examine the steps from visit to order.

Track landing pages, product views, carts and completed orders separately. Do not force visits with unknown origins into the AI category. Include cancellations and returns, and avoid precise gain claims with a small sample.

EXAMPLE 03

B2B: increasing the visibility budget

Sales reports more enquiries while advertising spend and the offer also changed.

Which activity generated the increase?

Open the recommendation for example 03

Document concurrent changes.

Follow lead quality, meetings and the sales cycle. A staged test across comparable markets or products may be worth designing, but comparability needs checking. Do not attribute a simple before-and-after difference to a single channel.

TRY IT WITH YOUR TEAM

Separate visibility, visits, leads and sales.

Define each metric, its data source and missing observations. Keep brand mentions, traceable visits, qualified leads and completed sales on separate report lines. This starter sheet is not an implementation of the paper’s statistical model.

Explore the DijitalPi visibility research method

Source and ownership

Masahiro Kato, Daiki Honma, Taka Kato.
The University of Tokyo; Mizuho-DL Financial Technology Co., Ltd.; NP-hard.

arXiv · 10 September 2026 · v1 · Method and simulation preprint. Open the original publication · Find on Google Scholar

The academic work belongs to the researchers named above. This page contains DijitalPi's explanatory review and original business examples; it is not a full translation of the paper.

← Return to all research articles