You know the position you built. Do you know the one AI describes?
A buyer asks an assistant for a shortlist. The answer arrives assembled from sources you own and sources you do not, with total confidence and no citation. AIWorthy measures what that answer says about your company — against the record your company files.
The reassurance most firms give themselves
Type your own company name into an assistant and it will tell you about you. Where you are, what you do, sometimes who runs the place. It sounds like it knows you.
That test is the most common way a marketing team checks its AI exposure, and it measures the wrong thing.
Ask an assistant about a named company and it will usually produce something. Ask the same assistant to recommend firms for the exact service that company provides, without naming it, and the answer is frequently a different set of companies entirely. The first question tests recall. The second tests discovery. Only one of them resembles what a buyer does.
Recall is what happens when someone already knows you exist. Discovery is what happens when they don't — which is every buyer who hasn't heard of you yet.
The buyer journey gained a new author
Brand teams have always shaped perception through positioning, content, public relations, social channels, customer experience, and the website. Buyers weighed those signals and formed a view.
Now an AI system may form the view first.
It searches across sources you own and sources you do not. It decides which evidence matters. Then it gives the buyer a concise, confident version of your company — often before anyone visits your site or speaks with sales.
The question is no longer only whether they can find you. It is also:
- Did the system understand what we do?
- Did it place us in the right category?
- Did it preserve the distinction we spent years building?
- Did it recommend someone else for the exact need we serve?
AI did not create your brand inconsistencies. It made them visible to buyers.
Some of what AI gets wrong about a company originates inside the company.
Where a company's website and its regulatory filing describe different things — an office listed in one and absent from the other, a name that changed, a figure that moved — an AI system will resolve the conflict on its own. Sometimes it qualifies the difference. Sometimes it asserts one version with a specificity neither source supports, and gives a different answer on the next ask.
Instability is itself the finding. A source being read does not change between asks. A model filling a gap does.
Where a company's own public records disagree, an AI answer will pick one and state it plainly. That is a governance question before it is a marketing one.
Why a conventional brand tracker cannot see this
The lost opportunity leaves no footprint.
If a buyer asks an assistant for a shortlist and your company is absent, there may be no impression, no visit, no form fill, no lost-opportunity record. The pipeline never forms.
Being present is not the same as being understood.
A firm can be cited frequently and recommended rarely. It can appear for its own name and disappear when the buyer describes a need. A single visibility score flattens those into the same problem when they require very different responses.
Your site is only part of the evidence.
AI systems draw from journalism, directories, forums, social platforms, partner sites, databases and old pages alongside the content you publish. When those sources conflict, the answer may expose a governance or reputation problem rather than a technical one.
A single check can create false confidence.
Platforms use different retrieval systems, different sources and different response patterns. Some vary their answer from one ask to the next; others repeat the same sources every time. A screenshot is an anecdote, not a measurement.
What the work requires
Measuring this properly is slower than running a tool, and the difference is in four places.
Questions derived from a public record.
Every question is built from a dated, verifiable source — a regulatory filing, a public disclosure, or another documented account of what the firm actually does. Not generated by a language model, not pulled from a keyword list. If we ask whether a firm surfaces for a particular service, it is because the record shows the firm performing it.
The question set frozen before any answer arrives.
The questions are fixed and their digital fingerprint published before collection begins, so nothing can be quietly rewritten once results come in.
Every platform, asked more than once.
Assistants disagree with each other and some disagree with themselves between asks. Measuring one platform once tells you almost nothing about what a buyer will see.
Every source recorded, and every count read by hand.
Which pages a platform consulted is often more useful than whether it named you, because it separates a positioning problem from a firm the system has never encountered. Automated matching produces candidates; a human confirms them before any number is reported.
Two baseline studies are in publication. Both panels, both screens, and the collection conditions are already published — including the questions as sent and the fingerprints of every frozen artifact.
What you receive
A document, not another dashboard.
- The questions, and the public sources they were derived from
- The full AI transcripts
- The firms recommended instead
- The sources each platform cited or surfaced
- A hand-verified analysis of what the result does and does not show
- A clear statement of limitations
Then we stop. Your team or your agency decides what to change. Because AIWorthy did not perform that work, we can independently measure the result afterwards.
You know the position your company intended to build. We show you the position AI can actually see.