SEO, AI citations, and acquired customers: how to connect the data without forcing attribution
Assess opportunities and leads by distinguishing what's measured from what's estimated.
SqualiOnline editorial team · 2026-09-07
The question always comes at the same moment: after a few months of work on visibility, someone asks how many customers it has brought in. The honest answer is that part of it can be measured, part of it can be estimated, and part of it will never be known. The problem isn't admitting that: it's telling the three parts apart, instead of pretending they're all the first one.
Anyone who promises a direct, complete link between a search and an invoice is oversimplifying something. Anyone who says nothing can be measured is giving up. In between there's workable ground, and it starts with defining the words.
Three words to define before counting
- Valid inquiry: not every inquiry received is one. Automated messages, incoming sales pitches, and questions outside your services need to be excluded. The definition needs to be written down, otherwise each person counts things differently.
- Opportunity: an inquiry that turns into a deal, meaning something someone follows up on and to which even an approximate value can be assigned.
- Sale: a contract or order, with its date, which is often far removed from the date of the initial inquiry.
Until these three words have a shared definition between whoever handles the site and whoever sells, every count remains questionable. It's the least technical step of all, and it's the one that changes the result the most.
What can be measured and what has to be estimated
It's worth putting in writing, once, what kind of answer you can get for each question. It reduces later arguments and stops an estimate from being treated as a fact.
| Question | Where the answer comes from | How reliable it is |
|---|---|---|
| How many people arrive from search, and on which pages | Site data | Good, if the tracking is set up correctly. |
| How many inquiries came from the site | The inquiries actually received, checked against the logged submissions | Good, if you count the messages that arrived and not the clicks. |
| Which page preceded the inquiry | Site data, within the limits of a single visit | Medium: it describes the visit, not the entire journey. |
| What made the customer aware of the company | A direct question to the customer, logged in the business management system | Medium: it depends on memory, but it's the only source for what happens outside the site. |
| Whether an answer generated by an assistant brought in a contact | Indirect signals and a direct question | Low: a recognizable reference rarely remains. |
| What an acquired customer is worth | Business management system | Good, and it's the data point most often missing from the reasoning. |
The most useful row is the last one. Without the value of a customer, all the other numbers remain counts: they say how many, not how much, and they don't let you decide where to put next year's work.
Connecting the data where they meet
The practical connection happens at one single point: the inquiry. It's the only moment where the anonymous world of visits and the named world of the business management system meet.
- At the moment of submission, record the available information along with the inquiry: where the visit came from, what the last page was, which form was used.
- Bring this information into the business management system as fields of the deal, not as an attachment nobody will open.
- Add the question for the customer, asked by whoever responds: how did they find you. A single question, in plain language, asked every time and not only when someone remembers.
- When the deal closes, feed the value back and attach it to that information.
From here on you can tackle questions that couldn't even be formulated before: which topics bring inquiries that turn into deals and which bring inquiries that turn into nothing. These are two different things, and they often come from different pages of the same site.
Multiple paths, direct traffic, and imprecise memories
A customer's actual path rarely looks like a straight line. Someone reads an answer that cites you, then searches for your name, then comes back two weeks later and calls. In the site's data, that person shows up as having arrived directly, or from a search on your name: the part that actually made the decision doesn't show up anywhere.
- The last source before the inquiry is the easiest to record and the least informative: it's almost always your own name.
- An increase in searches for the company's name is itself a signal: it means someone came to know you somewhere else.
- Traffic that shows up as direct lumps together very different things — a message, word of mouth, an address typed in by hand — and shouldn't be read as "they already knew us."
- Citations inside answers generated by assistants rarely leave a recognizable trace: for those, the direct question to the customer remains the main source.
Don't count twice
When you combine multiple sources, the same customer shows up more than once and the totals start to exceed reality. Three rules are enough to avoid this.
- An inquiry belongs to one channel only, according to a rule chosen in advance and applied the same way every time.
- If the customer says they found you one way and the data points to another, keep both pieces of information but with different roles: the customer's answer for how they came to know you, the data for how they reached the inquiry.
- The totals need to be checked against the number of new customers for the period: if the sum of the channels exceeds it, something is being counted twice.
The final check is always the same, and it needs no tools: the sum of the parts can't be bigger than the whole.
What this guide doesn't cover
This guide covers connecting visibility to business opportunities. How to build a periodic report on search results — what it contains, how often it's read, which variations are just noise — is covered separately, as is tracking your company's citations in assistants' answers, which has its own methods and limits. The choice between working on the site first or on advertising also has a dedicated guide.
Frequently asked questions
Why don't site data and business-management-system data ever match?
Because they count different things: the site counts submissions, the business management system counts the inquiries someone actually picked up. There's always a difference; the problem starts when nobody knows how big it is or what it depends on.
Is the question "how did you find us" reliable?
It's imprecise, but it's the only source for what happens outside the site. It becomes useful if whoever responds always asks it, in the same words, and if the answer ends up in a field of the business management system instead of a free-text note.
How much time is needed before you can read anything meaningful?
The time it takes to gather a sufficient number of inquiries and let deals mature through to closing. In industries with long sales cycles, it's worth looking at open opportunities first, and at sales only later on.
Let's connect visibility data to business opportunities.
If you’d like to talk it through, the service that handles this is GEO & SEO.

