Is the record sent to the model before asking it?

Never, and that is a foundational decision. If we told the model who you are before asking about your category, it would mention what we just told it and the measurement would be worthless. The record is used afterwards, when deciding whether an ambiguous mention is you.

What happens to the discarded mentions?

They are counted apart and the dashboard tells you how many. A denominator that silently shrinks is worse than no denominator: if forty of three hundred answers were discarded, you have a right to know before reading the percentage.

What if the system gets the separation wrong?

You can correct any answer by hand, and the correction is marked as human review rather than hidden. The recomputed metric says how many answers carry review, because a figure corrected by a person is not the same as an automatic one.

Does publishing the markup actually help?

It helps crawlers read your identity without having to infer it, and that reduces homonym confusion at the source. It is not magic and it does not fix a site with no content, but it is one of the few things you can declare yourself rather than waiting for somebody to work it out.

Content and brand

Let AI know who you are before it talks about you

Official name, aliases, domains, profiles and exactly what your company does. It is what makes it possible to separate your mentions from those of the other company with the same name.

Aliases and spelling variants, including the wrong ones your customers use
Explicit exclusions: what you are not, said by name
Markup ready for your site, so the signal comes from your side too

Free scanner See the full pricing

The problem

There is another company with your name in another sector and half your mentions are theirs.
Your brand is written three ways and you only count one.
A model describes you with the other company's activity and it sounds odd but not false.

A miscounted figure can be fixed. Being another company cannot.

How it works

The record is not sent to the model when measuring: that would be handing over the answers before the exam. It is used afterwards, to attribute correctly what came out.

1
You declare who you are
Legal name, trading name, aliases people use, official domains and verified profiles. The more complete, the less room for confusion.
2
You declare who you are not
It is the part that pays off most and the one almost nobody does. Naming the other same-named company explicitly lets ambiguous mentions be separated and keeps your metrics' denominator correct.
3
You publish the signal
The record generates markup you can paste into your site so crawlers read the same identity you declared. Declaring it only here tidies your data; publishing it changes what the rest of the world understands.

What you get

A record with its completeness level and the next concrete action.

What is left to declare, ordered by what prevents most confusion.
The markup ready to copy into your site.
The count of mentions separated thanks to the exclusions.

Measure before you pay anything

The scanner runs 10 real responses in about 60 seconds. No signup, no card. If the number it returns tells you nothing, do not buy.

Measure my brand

Who it is for

Brands with a common name, far more numerous than they think.
Companies with several legal names or that have changed name.
Anyone operating in several countries under slightly different names.

It is not for you if...

It is not for you if your name is unique and shared by nobody. There the record tidies your data and little more: the useful work is measuring and auditing pages.

Then look at GEO audit

An honest comparison

	By hand	With Citenza
What it costs you	Correcting by hand every mention that was not yours.	Exclusions apply themselves on every sweep.
The method	With no written criterion, so everyone corrects differently.	The criterion is written down and can be argued with.
Repeating it next month	On the next sweep, back to the start.	And the number of separated mentions is counted, so the denominator stays honest.

Availability by plan

It is on every plan, trial included, because without a record the metrics of a brand with homonyms are worthless and charging separately for the data being correct would make no sense. What grows by plan is the number of brands with their own record.

See the full pricing

Honest questions

Is the record sent to the model before asking it?: Never, and that is a foundational decision. If we told the model who you are before asking about your category, it would mention what we just told it and the measurement would be worthless. The record is used afterwards, when deciding whether an ambiguous mention is you.
What happens to the discarded mentions?: They are counted apart and the dashboard tells you how many. A denominator that silently shrinks is worse than no denominator: if forty of three hundred answers were discarded, you have a right to know before reading the percentage.
What if the system gets the separation wrong?: You can correct any answer by hand, and the correction is marked as human review rather than hidden. The recomputed metric says how many answers carry review, because a figure corrected by a person is not the same as an automatic one.
Does publishing the markup actually help?: It helps crawlers read your identity without having to infer it, and that reduces homonym confusion at the source. It is not magic and it does not fix a site with no content, but it is one of the few things you can declare yourself rather than waiting for somebody to work it out.

Start by saying who you are

And check how many of the mentions you thought were yours really were.

Free scanner Back to the catalogue

You declare who you are

You declare who you are not

You publish the signal

Measure before you pay anything

It is not for you if...

Start by saying who you are