An AI SEO agency should measure whether a business appears in relevant answers, how it is represented, and whether that visibility contributes to qualified demand and commercial results. Mentions and citations are useful early signals. They become meaningful when connected to the customers the business actually wants to reach.
Imagine a client’s AI mentions increase for three months, but enquiries remain flat. We would call that promising but unproven. The next job is to understand where the connection breaks.
Is the business appearing in buying conversations? Is it being recommended, or simply cited as a source? Are potential customers reaching the website and finding a reason to enquire? Does the buying cycle mean a result will take longer to emerge?
Those questions belong in the report, alongside the visibility chart.
Start with the business decision
Before choosing metrics, establish what success means for the client. A nursery needs suitable families to book visits and enrol. An industrial supplier may need fewer enquiries, provided they come from qualified buyers with substantial projects. A local service business may care most about booked appointments and repeat customers.
The baseline must reflect that difference. The competition matters too. A specialist academy with a handful of relevant competitors faces a different challenge from a coffee shop competing with hundreds of nearby alternatives.
We would set targets after researching the market, current visibility, website and competitive gaps. A universal promise made before that research is difficult to defend.
Measure the journey in layers
| Layer | What to measure | What it helps establish |
|---|---|---|
| Relevant visibility | Mention and recommendation frequency across agreed commercial questions, by engine and market | Whether the business reaches useful conversations |
| Representation | Accuracy, favourability, prominence and suitability of recommendations | Whether the answer gives buyers the right understanding |
| Source presence | Owned and third-party citations, including persistence across repeated checks | Which evidence repeatedly contributes to answers |
| Website response | Identifiable AI referrals, landing pages, qualified enquiries and booked appointments | What happens when a measurable visit occurs |
| Business results | Sales, revenue, acquisition cost, retention and customer lifetime value where reliable | Whether demand becomes commercially valuable |
These layers should remain distinguishable. A citation to a technical explanation is different from a recommendation to hire the company. A recommendation is different from an enquiry, and an enquiry is different from a customer.
Define the calculations too. For example, recommendation frequency can mean the proportion of eligible test answers that explicitly recommend the client. “Share of voice” needs a stated competitor set and denominator; otherwise two apparently similar percentages may measure different things.
A single screenshot is weak evidence
TJ Digital’s October 2026 citation study tracked 193,507 answers across ChatGPT, Google AI Mode and AI Overviews. Only 24–28% of pages cited for a question were cited again the following week.
It is an agency-client sample, not a universal benchmark, but it makes a useful point: one successful answer does not establish stable visibility. Repeated checks using a consistent question set are more informative. Track the model, interface, market and collection date so a change in measurement does not masquerade as a change in performance.
Our companion article explains why Dubai AI visibility needs its own measurement settings.
Follow the lead beyond the form submission
With a nursery client in Saudi Arabia, marketing was already generating enquiries and enrolments. We still believed conversion from enquiry to visit, and from visit to enrolment, could improve.
After reviewing the marketing, we examined WhatsApp and direct-message conversations. Replies were generic, and follow-up was inconsistent. Parents making a considered decision needed personal conversations and continued contact, rather than being treated as a stream of new enquiries.
We recommended a more personal approach and helped introduce follow-up processes. Both booked visits and enrolments subsequently improved. This was a broader marketing engagement, not an AI SEO experiment, and we cannot isolate the contribution of every change. It illustrates how we believe an agency should work: investigate the business result even when the next improvement involves the sales process.
“A recommendation is different from an enquiry, and an enquiry is different from a customer.
Barish Elrekabi, DGTL Force
Treat attribution honestly
Someone may discover a business in an AI answer and later search its name, call or visit directly. Referral analytics will miss some of that influence.
Use identifiable AI referrals alongside CRM records and a simple “How did you first hear about us?” question. Monitor branded searches and direct traffic as supporting context. None of these, alone, proves that AI SEO caused a sale.
Google’s generative AI performance report documentation describes impressions across AI Overviews and AI Mode, with page, country and device views. That helps measure exposure; it is not a record of qualified leads or revenue.
For longer buying cycles, compare enquiry cohorts over a suitable period. Report revenue and lifetime value only where the underlying data supports them, and label forecasts separately from realised results.
Expect accountability and a clear next step
An agency can commit to research, implementation, testing and an honest response to the evidence. Rankings, citations, recommendation share and referrals are outcomes it can influence, rather than fully control.
The report should explain what improved, what remains uncertain, where the business is losing opportunities and what will change next. Sometimes that means fixing the core website before publishing more content. Sometimes it means improving follow-up or revising the offer.
That is the thinking behind our AI SEO services. If your current reporting leaves those questions unanswered, talk to DGTL Force about connecting visibility to the results your business needs.
Sources
- TJ Digital: AI citation study 193,507 AI answers across ChatGPT, Google AI Mode and AI Overviews
- Search Engine Journal: AI-cited pages more often got straight to the answer Independent coverage of the TJ Digital study and its method
- Google Search Console Help: Generative AI performance report Impressions in AI Overviews and AI Mode by page, country and device
- Google Search Central: AI features and your website How AI Overviews and AI Mode select supporting links
Research checked November 2026. Figures belong to their publishers and reflect their own samples and methods.